2011 Agenda Abstracts

Best of Fast

A Study of Practical Deduplication

Bill Bolosky, Member, Microsoft Research

Abstract

We collected file system content data from 857 desktop computers at Microsoft over a span of 4 weeks. We analyzed the data to determine the relative efficacy of data deduplication, particularly considering whole-file versus block-level elimination of redundancy. We found that whole-file deduplication achieves about three quarters of the space savings of the most aggressive block-level deduplication for storage of live file systems, and 87% of the savings for backup images. We also studied file fragmentation finding that it is not prevalent, and updated prior file system metadata studies, finding that the distribution of file sizes continues to skew toward very large unstructured files.


Emulating Goliath Storage Systems with David

Leo Prasath Arulraj, Software Development Engineer, Amazon

Abstract

Benchmarking file and storage systems on large file-system images is important, but difficult and often infeasible. Typically, running benchmarks on such large disk setups is a frequent source of frustration for file-system evaluators; the scale alone acts as a strong deterrent against using larger albeit realistic benchmarks. To address this problem, we develop David: a system that makes it practical to run large benchmarks using modest amount of storage or memory capacities readily available on most computers. David creates a "compressed" version of the original file-system image by omitting all file data and laying out metadata more efficiently; an online storage model determines the runtime of the benchmark workload on the original uncompressed image. David works under any file system as demonstrated in this paper with ext3 and btrfs. We find that David reduces storage requirements by orders of magnitude; David is able to emulate a 1 TB target workload using only an 80 GB available disk, while still modeling the actual runtime accurately. David can also emulate newer or faster devices, e.g., we show how David can effectively emulate a multi-disk RAID using a limited amount of memory.

Birds of a Feather

Hiring CIFS Talent

Chris Hertel (ubiqx)

Abstract

SMB/CIFS/SMB2 is the most widely used network file-system on earth, but the engineering talent pool is small and those who are familiar with protocol internals are scattered around the world instead of being conveniently located in Silicon Valley.

This BoF will be an open forum to discuss how companies can find and hire SMB/CIFS/SMB2 developers, and how they help current employees and new hires learn to handle these difficult protocols. The BoF will also be an opportunity for SMB/CIFS/SMB2 engineers to discuss the marketplace and future opportunities.


Distributed Content Caching

Chris Hertel (ubiqx)

Abstract

Is PeerDist (BranchCache) the de-facto standard for Distributed Content Caching? In this BoF, we will discuss the barriers to wide-scale adoption and deployment in a multi-platform world. Technical aspects including design, implementation, and testing issues may also be covered.

 

CIFS/SMB/SMB2

SAS Standards and Technology Update

Harry Mason, Director Industry Marketing, NetApp

Marty Czekalski, Vice President SCSI Trade Association; Interface & Emerging Architecture Program Manager, Seagate

Abstract

SAS has become the backbone of enterprise storage deployments.  SAS has rapidly evolved by adding new features, capabilities, and performance enhancements.  This talk will include an up-to-the-minute recap of the latest additions to the SAS standard and roadmaps.  It will focus on areas of enhanced connectivity solutions, MultiLink SAS, status on 12Gb SAS development status, and a new transport investigation of SOP (SCSI over PCIe).

Learning Objectives

  • Attendees will learn how SAS will grow and thrive, in part, because of the Advanced Connectivity Roadmap, which offers a solid connectivity scheme based on the versatile Mini-SAS HD connector in addition to SAS Connectivity Management support.
  • Attendees will learn that MultiLink SAS improves how slot-oriented Solid State Drive (SSD) devices can be configured to boost I/O performance. When running at 12Gb/s, a single slot will be capable of providing up to 96Gb/s of bandwidth (full duplex).
  •  The latest development status and design guidelines for 12Gb/s SAS will be revealed
  • Attendees will learn the details of the standardization activity and architecture for SCSI over PCIe (SOP).

File Systems and Thin Provisioning

Frederick Knight, Standards Technologist, NetApp

Abstract

New operations to manage Thin Provisioning  have been added to or updated in the ATA and SCSI standards recently.    This session will explain these capabilities and their storage APIs so that file system developers and application developers will gain an understanding of how and when to use these new features to enhance the efficient use of their storage subsystems.

Learning Objectives

  • New operations to manage Thin Provisioning  have been added to or updated in the ATA and SCSI standards recently.    This session will explain these capabilities and their storage APIs so that file system developers and application developers will gain an understanding of how and when to use these new features to enhance the efficient use of their storage subsystems.
  • Understand Thin Provisioning concepts
  • Understand ATA Trim capabilities and commands
  • Understand SCSI Provisioning Management capabilities and commands

Storage Data Movement Offload

Frederick Knight, Standards Technologist, NetApp

Abstract

SCSI Operations that allow storage devices to offload data movement by the host into the storage device have recently been updated and enhanced to provide new capabilities.  This session will explore these new capabilities, and how they are being used today.  File system and application developers will gain an understanding of how these features operate and how they can be used to improve performance

Learning Objectives

  • SCSI Operations that allow storage devices to offload data movement by the host into the storage device have recently been updated and enhanced to provide new capabilities.  This session will explore these new capabilities, and how they are being used today.  File system and application developers will gain an understanding of how these features operate and how they can be used to improve performance
  • Understand what copy offload is
  • Understand how copy offload operates
  • Understand the SCSI commands used to perform copy offload

Data Integrity from Application to Storage

William Martin, Engineer Consultant, Emulex

Abstract

Data integrity failures to high-visibility applications have prompted vendors to add data integrity mechanisms to databases, file systems, and storage devices. The Data Integrity model being developed by SNIA's Data Integrity Technical Working group (DITWG) presents a model of how data can be protected from the application to the storage device. The Data Integrity model defines the building blocks used in operating systems for protecting data and providing true end-to-end data integrity protection. This model utilizes the "protection Information" feature (also known as DIF) defined in the SCSI block device command set (SBC) standard as one of the forms of protection provided in the complete stack. However, additional interfaces are necessary to extend this protection all the way to the application.

Learning Objectives

  • What are the risks for Data Integrity
  • Where are the risks for Data Integrity
  • How does the SNIA Data Integrity Model protect from those risks

SMB 2.2:  Bigger.  Faster.  Scalier - (Parts 1 and 2)

David Kruse, Principal Development Lead, Microsoft

Abstract

This session comprises two parts that will take a detailed look at new extensions to the SMB 2 protocol. These new developments target improving file server availability and client-server performance. The sessions will give you an overview of what is new in SMB 2.2, and then examine in detail specific areas of the protocol.

SMB 2.2 Multichannel adds new levels of network scalability and support for modern interconnects. Persistent handles adds fault tolerance and continuous availability to the protocol with the 'persistent handles' feature adds fault tolerance and continuous availability to the protocol."

We will also discuss auxiliary protocols which live side by side with the SMB 2.2 protocol to provide end-to-end reliability and manageability.


SMB 2.2:  Bigger.  Faster.  Scalier - (Parts 1 and 2)

Mathew George, Sr. Software Development Engineer, Microsoft

Abstract

This session comprises two parts that will take a detailed look at new extensions to the SMB 2 protocol. These new developments target improving file server availability and client-server performance. The sessions will give you an overview of what is new in SMB 2.2, and then examine in detail specific areas of the protocol.

SMB 2.2 Multichannel adds new levels of network scalability and support for modern interconnects. Persistent handles adds fault tolerance and continuous availability to the protocol with the 'persistent handles' feature adds fault tolerance and continuous availability to the protocol."

We will also discuss auxiliary protocols which live side by side with the SMB 2.2 protocol to provide end-to-end reliability and manageability.


Advancements in Backup to Support Application Storage on a File Server

Molly Brown, Principal Development Lead, Microsoft

Abstract

There are many compelling reasons for server applications, such as Hyper-V, to store their data on a file share, but this cannot be done if it compromises the application’s data backup and recovery strategy. This session will describe the new protocol, MS-FSRVP, designed for Windows Server 8 that allows an application server to drive the required coordination with a file server so that existing backup and recovery strategies continue to work as the application data moves from local to remote storage.

Learning Objectives

  • Understand the unique problems in the backup space when application data is stored on a file server
  • Understand the design goals for the new MS-FSRVP protocol
  • Get a detailed overview of the MS-FSRVP protocol

SMB 2.2 over RDMA

Thomas Talpey, Software Architect, Microsoft
Greg Kramer, Ph.D., Software Development Engineer, Microsoft

Abstract

A new protocol, SMB2 Direct, has been created which allows SMB 2.2 to operate over Remote Direct Memory Access (RDMA) transports such as iWARP, Infiniband and RoCE. This layering enables significant increases in performance for all SMB 2.2 file-based workloads and dramatically broadens the applicability of SMB 2.2. The presentation will outline the goals and motivations for the new approach, and will make a deep dive into the SMB2 Direct protocol itself, including early performance results.


SMB2 -  Advancements for WAN

Molly Brown, Principal Development Lead, Microsoft
Mathew George, Sr. Software Development Engineer, Microsoft

Abstract

This session covers advancements in SMB2 file service deployments over a Wide Area Network (WAN). This includes discussions on enhancements  for SMB2 file servers and SMB2 file clients in these scenarios, such as metadata optimizations and identification of potential file caching opportunities.


Accelerating SMB2

Mark Rabinovich, R&D Manager, Visuality Systems

Abstract

Global networks introduce a significant challenge when it comes to CIFS traffic, making it virtually unbearable for an end user. SMB2 is no less chatty than SMB, hence - it may be accelerated. We will show how to improve SMB/SMB2 traffic using various acceleration techniques. This presentation emphasizes on challenges introduced by SMB2 as in comparison with SMB. I will share our CIFS acceleration experience and performance statistics for accelerated WANs with SMB2 traffic

Learning Objectives

  • SMB2 traffic in WANs
  • SMB2 acceleration methods
  • SMB2 acceleration statistics
  • SMB2 acceleration examples

Samba Status Report

Volker Lendecke, Samba Team / SerNet,

Abstract

Samba is a rapidly evolving project that is part of the basis for many  NAS vendors. This talk will give an overview of the current development of Samba.
   
Learning Objectives

  • Clustering: The current status of the active-active clustered CIFS server with Samba and ctdb will be presented.
  • SMB2: With Samba 3.6 SMB2 is a fully supported protocol in Samba
  • Printer support: The printer subsystem has been overhauled. Reasons for this overhaul and implications will be presented
  • Samba/AD: Samba 4 strives to become an AD compatible Domain Controller, while Samba 3 is a solid file and print    server. Different ways to merge those two to go forward with Samba 4.0 will be presented.

A CIFS Geek in Exile:  What I did on my Holiday

Christopher Hertel, Storage Architect, and CIFS Geek ubiqx Consulting, Inc.

Abstract

BranceCache is a distributed caching system implemented by Windows SMB2 servers.  BITS, according to at least one Microsoft Blog, is the "Earth’s most widely used file transfer service".  This presentation covers a new Open Source implementation of both BITS and BranchCache.

Learning Objectives

  • BITS protocol and how does it relate to SMB/CIFS/SMB2.
  • BranchCache, and how it relates to SMB/CIFS/SMB2 and BITS.

CTDB Status - Clustered Samba Growing Up

Michael Adam, Senior Software Engineer, Samba Team / SerNet

Abstract

CTDB is a highly specific clustered database and management software sitting between Samba and a cluster file system. It allows to create scaling CIFS/NFS clusters on Linux.    An early self-contained implementer of all-active service clustering, CTDB now slowly finds its way into the Linux distributions as a managed resource of the Linux cluster stack.    Initially, the problems that Samba is facing, when running on a file system cluster, are recalled as well as the design and history of the CTDB software.    An overview is given of the past year's bigger changes in CTDB, especially transaction handling and vacuuming.    The various modes in which CTDB can be run are described, and how major Linux distributions start to integrate CTDB in their cluster products.

Learning Objectives

  • Learn about clustered CIFS services with CTDB and Samba
  • Learn about new developments in CTDB / Clustered Samba
  • Learn about integration of CTDB into the Linux distributions

Experiences in Clustering CIFS for IBM Scale Out Network Attached Storage

Dr. Jens-Peter Akelbein, IBM Germany, Research and Development

Abstract

Clustering the CIFS and SMB2 protocol is enabling managing large scale data in a single name space scaling the bandwidth of access as well. IBM SONAS uses clustering across various nodes while scaling capacity indepently by a second tier of nodes. Experiences with clustering CIFS including the underlying clustered file system the past years led to improvements in regards of performance and stability. Utilizing SMB2 as the protocol leads to improvements beyond CIFS capabilities. Compared to traditional active-passive configurations larger clusters provide active-active configurations allowing flexible maintenance and management. This talk should give an insight on resolved performance challenges in applying clustered CIFS with different installations and workloads including improvements being made or currently applied to Samba and CTDB for being used in an Enterprise product.

Learning Objectives

  • Clustering and performance, how to achieve both
  • Requirements to clustered file systems by clustering NAS
  • Requirements to protocol implementations by clustering NAS
  • Benefits of clustering for managing Scale Out NAS devices

Hidden Gems in the NAS Protocols
James Cain, Principal Software Architect, Quantel Limited

Abstract

Having spent the last few years implementing SMB and SMB2 servers, the presenter has discovered that there are parts of these protocols that seem to offer untapped semantic richness. This session will propose theories and demonstrate practical working examples that test these theories. Examples will include: dynamically offering different representations of the same resource, extracting provenance from running software and avoiding NAS head state to support dynamic failover in a clustered file-system. All these examples will be built up from theoretical principals and demonstrated with working prototypes.

Learning Objectives

  • Understanding untapped NAS semantics
  • Exploiting COTS operating systems for innovative uses
  • Exploring how NAS installations can scale

Through the Looking Glass; Debugging CIFS/SMB/SMB2

Robert Randall, Senior Software Architect, Micron Technologies, Inc.

Christopher Hertel, Storage Architect, and CIFS Geek, ubiqx Consulting, Inc.

Abstract

While protocol suite tests are quite useful, there are other ways to understand  how Windows is interacting with your SMB client or server.  Built into the   kernel of Windows is a treasure trove of telemetry which provides a rich  context and clear complaints when redirectors are interacting with another  end point.  Watch as the treasure is revealed through simple examples  that demonstrate the value of knowing what the Windows kernel can tell  you about how your client or server is behaving.  Leave with step by step  instructions on how to use these valuable tools.  The tools are freely  available from Microsoft's web site.

Learning Objectives

  • Understand the tools required for Windows kernel debugging
  • Learn the steps required to turn on SMB/SMB2 redirector telemetry
  • Experience real examples of how the telemetry can help you to better understand your client or server and how it interacts with Windows

Lessons learned implementing a multi-threaded SMB2 server in OneFS

Aravind Velamur Srinivasan, Senior Software Engineer, Isilon Systems, Inc

Abstract

This talk will examine the lessons learned implementing SMB2 in the OneFS operating system and also highlight performance optimizations made in a multi-threaded SMB server implementation. In addition to these, the talk will also compare SMBv1 vs SMB2 in OneFS operating system to stress the performance benefits of using SMB2 over SMBv1.  SMB2 offers a server-side credit mechanism to throttle greedy clients. Different credit algorithms can cause weird client behavior in certain scenarios. We’ll examine some common mistakes to avoid.  The multi-threaded SMB servers have their own advantages and disadvantages. This presentation will throw light on the performance optimizations that can be made in a multi-threaded server implementation.  In addition to the aforementioned points, the presentation will also highlight the inherent performance benefits achieved by using SMB2 protocol over SMBv1, by presenting some performance numbers of using SMBv1 vs SMB2 in OneFS operating system.

Learning Objectives

  • SMB2 implementation in OneFS
  • The SMB2 server credit algorithm and potential pitfalls in its implementation
  • Performance optimizations in a multi-threaded SMB server
  • Performance gains achieved by using SMB2 over SMBv1 in OneFS operating system

Implementing SMB 2.1 In Likewise Storage Services
Gerald Carter, CTO, Likewise Software

Abstract

After completing support for SMB2.0 and MS Vista clients, server implementers must turn their focus to the additional SMB2 protocol features utilized by Windows 7 and Windows 2008 R2 clients.  This session will focus on experiences and knowledge gained from implementing SMB2.1 feature support in the Likewise Storage Services platform.  Topics covered will include protocol dialect negotiation beyond SMB 2.0, concurrent support for Windows 7 leases and legacy oplocks, multi-credit I/O support, and persistent file handles.

Learning Objectives

  • Understand the design and implementation trade offs involved in managing SMB 2.1 Leases
  • Explain the requirement changes to a server’s credit handling regarding large I/O requests
  • Discuss support for durable handles vs. resilient handles in SMB2 and 2.1 servers

Thinking Inside the Box: Embedded Active Directory / Storage Appliances Based on Samba

Kai Blin, Embedded Developer, Samba Team 

Abstract

In many SOHO setups, a central storage server or NAS device is already in use.  Existing Open Source software makes it very easy to also move the Active Directory  domain controller to the same machine, providing easy-to-use user management and  file/print services to SOHO customers. This talk will describe a proof-of-concept  implementation of an embedded Active Directory DC and SMB/CIFS file/print server  for SOHO setups that can be administrated using a web interface or existing AD  management tools. While the proof-of-concept implementation is limited to about  a dozen clients, the same system is useable on more powerful hardware for bigger  networks.

Learning Objectives

  • Using Samba to provide AD DC and file/print services
  • Scaling Samba down to embedded system constraints
  • Using a web interface to administer Samba

Moving an Enterprise Database Platform to run on CIFS/SMB/SMB2 File Access Protocols

Kevin Farlee, Storage Engine Program Manager, SQL Server, Microsoft

Abstract

There are a lot of considerations to go through when converting a performance-sensitive enterprise app designed to run against direct-attached or SAN hardware, and running it against Network Attached Storage. The Microsoft SQL Server Storage Engine team faced this problem when redesigning SQL Server to run over CIFS/SMB/SMB2 protocols to use NAS. I will discuss some of key issues we tackled: I’ll walk you through each of these stages, and discuss how this progressed in real life.

Learning Objectives

  • How are databases different from more typical NAS workloads?
  • Why would we do that? (What makes it worth the engineering investment?)
  • What are the considerations in moving to a new storage platform?
  • Code Changes – What product code had to change?
  • Testing – How does this impact automated test infrastructure?
  • Taking advantage of new capabilities

Cloud

Programming the Cloud

Fleur Dragan, Consultant, EMC

John Kilroy, Principal Software Engineer, EMC

Abstract

This talk will describe how cloud-based applications interact with stored data.  The traditional semantics of file systems are often not applicable or relevant.  We will discuss how language-specific idioms from several common development frameworks are mapped into the stored data abstractions present in the cloud.  We will also outline some of the practical implications of deploying such applications.

Learning Objectives

  • Understand how cloud applications interact with stored data
  • Semantic differences between traditional file systems and cloud stores (blob, NoSQL and key-value)
  • Comparison of language-specific idioms for interacting with cloud stores (Ruby vs Java/Spring vs Python)
  • Practical implications of variations in consistency models, latencies, and geographic distribution
  • Lessons learned from a real-world implementation

CDMI for Cloud IPC

David Slik, Technical Director, Object Storage, NetApp, Inc.

Abstract

In addition to providing storage services to end users, cloud storage systems enable cloud-aware programs to communicate between themselves in a distributed and asynchronous manner by using the cloud as a platform for Inter-Process Communications (IPC). This session discusses ways that CDMI enables cloud IPC, and the various use cases enabled by the use of a cloud in this manner. Special emphasis is given to CDMI Queues, which provide a first-in-first-out storage object, and are valuable for fan-in, fan-out and buffering messages and data exchanged between programs.

Learning Objectives

  • Learn how CDMI clouds can be used to author distributed Internet applications
  • Learn about the different use cases for cloud IPC that are enabled by CDMI
  • Learn about CDMI Queues and message passing
  • See a demonstration of how cloud IPC can be used to create distributed web applications

Open Source Droplet library with CDMI support

Giorgio Regni, CTO, Scality

Abstract

A year ago we started the Scality Open Source Program here at SDC by opening our Droplet project, a BSD licensed cloud storage client library.  Now Droplet has been extended to support CDMI in addition to the S3 protocol and it's the only cross cloud compatible C client library.  Developers around the world have contributed and worked on very promising projects including a cloud migration tool, an incremental backup agent with data deduplication as well as virtual machine image targets.

Learning Objectives

  • Access CDMI compatible storage
  • Develop cross cloud applications
  • Migrate data from one cloud to another
  • Integrate Droplet in existing applications

CDMI Federations, Year 2

David Slik, Technical Director, Object Storage, NetApp, Inc.

Abstract

In addition to standardizing client-to-cloud interactions, the SNIA Cloud Data Management Interface (CDMI) standard enables a powerful set of cloud-to-cloud interactions. Federations, being the mechanism by which CDMI clouds establish cloud-to-cloud relationships, provide a powerful multi-vendor and interoperable approach to peering, merging, splitting, migrating, delegating, sharing and exchange of stored objects.    In last years SDC presentation, the basics of CDMI federation were discussed. For year two, we will review further refinements to making federations interoperable, demonstrate common use cases enabled by federation, and discuss the ongoing work within the SNIA Cloud Storage Technical Working Group to add federation as a formal part of the CDMI 1.1 that is standard currently under development.

Learning Objectives

  • Learn what CDMI Federations are, and how they help cloud users and providers
  • Learn about common use cases of CDMI Federations
  • Learn about updates to CDMI federation and ongoing standardized efforts
  • See a multi-system demonstration of CDMI Federations in action

CDMI Conformance and Performance Testing

 David Slik, Technical Director, Object Storage, NetApp, Inc.

Abstract

In order to deliver on the multi-vendor interoperability promise of the Cloud Data Management Interface (CDMI) standard, conformance testing and performance benchmarking tools are an essential part of the development and user community ecosystem. This session reviews the goals of conformance and performance testing, and provides an overview of the open source confCDMI and perfCDMI tools released by NetApp to assist in the validation and performance characterization of CDMI storage systems.

Learning Objectives

  • Learn about the value and goals of CDMI conformance and performance testing
  • Learn about the confCDMI and perfCDMI tools
  • See a demonstration of the confCDMI and perfCDMI tools

Use of Storage Security in the Cloud

David Dodgson, Software Engineer, Unisys

Abstract

Everyone is concerned with the security of their storage in the cloud, however; security in any particular case depends on what the user is trying to accomplish.  Someone storing pictures of their children in the cloud will have a different idea of security than someone who is generating payroll information.  Storage security needs to be implemented with an understanding of the different needs of different users.  Enterprises will want to use secure private clouds that are customized to their individual security requirements, while individuals will want public clouds to address their needs.  The most important security requirements will be those that satisfy the needs of the greatest number of users in a particular cloud environment.

Learning Objectives

  • Describe common use cases for storage in the cloud, from Infrastructure as a Service to Storage as a Service.
  • Compare how different use cases have different security needs.
  • Compile a list of security requirements for storage in clouds, both private and public, to satisfy these needs.
  • Show how these requirements might be implemented.

Authenticating Cloud Storage with Distributed Keys

Jason Resch, Senior Software Engineer, Cleversafe

Abstract

Cloud storage is different from traditional systems.  Typically the storage provider is not fully trusted.  Passwords are often reused, easy to crack, difficult to remember, and depend on availability of the authentication service.  Private keys offer a more resilient and secure method, but migrating, using and maintaining such keys is burdensome.  A new technique will be discussed which combines the best of both words: the ease of use of passwords with the security properties of keys.  Using this method in cloud storage systems user experience, security, and robustness  can be greatly improved.

Learning Objectives

  • Unsuitability of usernames and passwords
  • Difficulties in using certificates and keys
  • A new technique: Distributed authentication keys

Resilience at Scale in the Distributed Storage Cloud

Alma Riska, Consultant Software Engineer, EMC

Abstract

The cloud is a diffuse and dynamic place to store both data  and applications, unbounded by data centers and traditional  IT constraints.  However, adequate protection of all this  information still requires consideration of fault domains,  failure rates and repair times that are rooted in the same  data centers and hardware we attempt to meld into the cloud.  This talk will address the key challenges to a truly global  data store, using examples from the Atmos cloud-optimized  object store.  We discuss how flexible replication and  coding allow data objects to be distributed and where  automatic decisions are necessary to ensure resiliency at  multiple levels.    We will discuss the impact of using a virtualized  infrastructure inside the cloud, noting where this does and  does not change the resiliency characteristics of the  complete system and how it changes the design reasoning  compared to purely physical hardware.  Automatic placement  of data and redundancy across a distributed storage cloud  must ensure resiliency at multiple levels, i.e., from a  single node to an entire site.  System expansion must occur  seamlessly without affecting data reliability and  availability.  All these features together ensures data  protection while fully exploiting the geographic dispersion  and platform adaptability promised by the cloud.

Learning Objectives

  • Understand replication and coding as used in the cloud
  • Understand the basics of fault domains and resilient design
  • Separate the promise from the reality of cloud storage
  •  Understand resilience of virtual vs. physical infrastructure
  • Understand how automated, policy-driven storage increases data resiliency

Changing Requirements for Distributed File Systems in Cloud Storage

Wesley Leggette, Cleversafe, Inc.,

Abstract

File systems typically have centralized metadata servers that present performance bottlenecks as concurrent users and system size increase. These are unique challenges for distributed file systems. Cloud storage systems often store large unstructured content, and the streaming write access patterns typical of such systems allows for optimizations that cannot be made in traditional file systems.  A new technique that adapts principals from NOSQL and object storage paradigms - and uses information dispersal for both underlying storage and metadata - provides a viable solution for streaming write access patterns. This technique allows for distributed writes, no single point of failure, scalability of both system size and concurrent clients, and limits performance bottlenecks.

Learning Objectives

  • Learn how new access patterns for large content repositories allow for optimizations in file system design.
  • Understand the importance of providing reliability and scalability for both data and metadata.
  • Learn about optimistic concurrency on an underlying dispersed storage substrate, and how allows effective metadata management without complex distributed transaction systems.
  • Learn how this technique allows for distributed writes..
  • Learn how dispersal allows file system design to be simplified by eliminating the complexity of replication management.

Best Practices in Designing Cloud Storage Based Archival Solution

Jim Rice, Principal Engineer, EMC

Abstract

Cloud storage facilitates the use case of digital archiving for long periods of time by transparently providing scalable storage resources. With ever increasing amount of data to be preserved for legal and compliance reasons, cloud storage when designed correctly, can provide a low cost solution in a geographically distributed environment. This presentation highlights the key considerations while developing an archive product using cloud storage based on REST interface. It goes on to highlight the design choices while developing a file based archiving solution to cloud storage using EMC Atmos as an example. The aspects covered in the slides are – security, performance, using vendor neutral APIs, developing portable application irrespective of the backend cloud supported, taking advantage of geographically spread cloud storage nodes, faster searches and an efficient disaster recovery mechanism.

Learning Objectives

  • Considerations while developing an archival application with cloud storage
  • Security and performance aspects while designing an application for cloud storage
  • Leverage cloud vendor provided capabilities in your application

Tape’s Role in the Cloud

Chris Marsh, Market Development Manager, Spectra Logic

Abstract

There is no doubt cloud storage is having a profound impact on IT and how technologies are deployed and consumed. Tape is the strong, silent partner to the cloud – very much present and in use, but completely transparent to the end-user. Chris will discuss how cloud storage’s consumption model is built around ease of use, flexibility and cost savings, and why tape is one of the most logical and cost effective tiers for storing data in the cloud; particularly as the cost difference between tape and disk increases as data sets grow. He will review the key benefits of tape, reveal why it is quickly becoming the media of choice for cloud providers, and provide real-world examples of tape’s role in the cloud.

Learning Objectives

  •  Learn about the two key benefits of tape for cloud storage providers: cost savings (CapEx, OpEx, power & cooling) and multiple media platforms for reliability/continuity.
  •  Learn why tape has become a popular choice for cloud providers in relation to disaster recovery scenarios.
  •  A review of cloud storage provider best practices and how to utilize the right mix of media to guarantee that protected data is available to be restored when needed, including the need for offline

DATA MANAGEMENT

Long Term Information  Retention

Dr. Sam Fineberg, Information Management Chief Technologist, HP Software

Simona Rabinovici-Cohen, Research Staff Member, IBM Research - Haifa

Abstract

As more of the world’s information is digital throughout its entire lifecycle, we are faced with the age old issues of record keeping and preservation, applied to devices and formats that were never intended to last. Long-term digital information suffers from issues that didn’t exist in short-term or paper world, such as media and format obsolescence, bit-rot, and loss of metadata. The SNIA Long Term Retention (LTR) TWG has taken the lead on this issue for the storage industry. Working with key stakeholders in the preservation field, the LTR TWG is developing the Self-contained Information Retention Format (SIRF). SIRF is a low-level data format that enables applications to interpret stored data, no matter what application stored it originally. SIRF will be examined in a new European Union integrated research project, called ENSURE – Enabling kNowledge, Sustainability, Usability and Recovery for Economic Value. ENSURE creates a preservation infrastructure for commercial digital information built upon cloud storage and virtualization enablement technologies. It explores issues such as evaluating cost and value for different digital preservation solutions, automation of preservation processes, content-aware long term data protection, new and changing regulations, and obtaining a scalable affordable solution by leveraging cloud technology. The presentation will cover use cases, requirements, and the proposed architecture for SIRF as well as its potential usage in ENSURE storage services.

Learning Objectives

  • Recognize the difficulties in long-term digital preservation, and how they differ from short-term.
  • Learn best practices and techniques to mitigate the risks inherent in long term preservation, and how those can be applied to today’s storage systems.
  • Identify some of the related standards developed within SNIA and other organizations that are relevant for long term digital preservation.
  • Discuss the need, use cases, requirements, and proposed architecture of SIRF.
  • Discuss the latest activities in the development of SIRF.

Open Unified Data Protection and Business Continuity

Dr. Anupam Bhide, CEO/Founder, Calsoft

Abstract

A unified model for Data Protection and Business Continuity in complex enterprise systems  Today’s data center administrators are faced with challenges of managing and protecting complex enterprise systems comprising of physical and virtual components composed of heterogeneous hardware and software with complex interconnects. A typical enterprise system consists of    Applications –  SAP, Exchange, Share Point more  Middleware – Databases,  Web Servers,   Operating Systems – Windows, Linux, UNIX & Hypervisors  Servers –  Servers (Physical), Blades, Servers (Virtualized)  Network – Switches Physical & Virtual, Routers, Firewalls  Storage –  SAN, NAS, DAS, Cloud     All components in the enterprise system uses  persistent storage in the form of DAS, NAS or SAN.   The unified model presented and implemented by Calsoft explores a open independent framework for data protection using storage level snapshots. The framework uses SMI-S to interact with various enterprise system components and ensures a consistent state to perform data protection and disaster recovery.

Learning Objectives

  • Efficient Storage Provisioning and Storage Management
  • Protect Data from Corruptions and Attacks
  • Integration between Applications, Middle tier, OS /Hypervisors, Physical (Server) Layer and Storage
  • SNIA – SMI-S and a move to build an open standards for providing data protection at all levels.
  • To address the challenges faced in data protection of complex enterprise systems today

A Centralized Data Protection Application for cross vendor storage systems

Nishi Gupta, Tata Consultancy Services
Prateek Sinha, Tata Consultancy Services (bio pending)

Abstract

Data centers consist of Storage Products from different vendors to meet the storage needs of multiple servers hosting various applications like: Databases, MailServers. These mixed environments pose a challenge for administrators as they have to use different Data Protection Applications (DPAs) for different combinations of Applications, Operating Systems and Storage Systems.     Here we are presenting an approach to resolve the mentioned issue. A centralized DPA can be developed with a single user interface, which uses storage API’s in the backend to perform data protection operations on cross-vendor storage systems. It consists of pluggable modules for different functionalities in areas example Backup/Recovery, Archival, Compliance, Deduplication.     We will be sharing the results of the PoC and the challenges that were faced and proposing the need of Standards in Storage Data Protection APIs.

Learning Objectives

  • Understanding the need of Standardization in Data Protection APIs
  • Understanding how the API’s of inbuilt Backup/Recovery, Archival, Compliance, etc capabilities of various Storage Appliances can be leveraged to provide data protection for Applications (database, mail server, etc).
  • Understanding of existing gaps that are present in Storage API’s of different vendors in the area of Data Protection.

FCoE Direct End-Node to End-Node (aka FCoE VN2VN)

John Hufferd, Owner, Hufferd Enterprises

Abstract

A new concept has just been accepted for standardized in the Fibre Channel (T11) standards committee; it is called FCoE VN2VN (aka Direct End-Node to End-Node). The FCoE standard which specifies the encapsulation of Fibre Channel frames into Ethernet Frames is being extended to permit FCoE connections Directly between FC/FCoE End-Nodes. The tutorial will show the Fundamentals of the extended FCoE concept that permits it to operate without FC switches or FCoE Switches (aka FCF) and will describe how it might be exploited in a Data Center environment.

Learning Objectives

  • The audience will gain a general understanding of the concept of using a Data Center type Ethernet for the transmission of Fibre Channel protocols without the need for an FCoE Forwarder (FCF).
  • The audience will gain an understanding of the benefits of converged I/O and how a Fibre Channel protocol can share an Ethernet network with other Ethernet based protocols and establishes a virtual FCoE link directly between the End-Nodes.
  • The audience will gain an understanding of potential business value and configurations that will be appropriate for gaining maximum value from this Direct End-Node to End-Node.

Understanding Primary Storage Optimization Options

Jered Floyd, Chief Technology Officer and Founder,
Permabit Technology Corporation

Abstract

Selecting the right primary data optimization technology for your storage platform and integrating it into your existing software can be much less painful than it may at first sound. While compression and data deduplication are both now standard features for backup, few storage vendors have integrated either into primary storage. There are many challenges when trying to leverage technology designed for backup within primary storage, but newer technologies designed specifically for primary storage are much simpler to implement.  Combining deduplication with compression provides additive savings, allowing storage vendors to drive cost savings to their users while still maintaining data safety. In this session, Jered Floyd, CTO of Permabit, will compare the two technologies using real world case studies and will explore which is a better fit for different data types and system architectures.

Learning Objectives

  • Attendees will learn when deduplication and compression are best used together and when they are best used separately
  • Attendees will have a clear understanding of the benefits and drawbacks of the two technologies.
  • Attendees will fully comprehend how to avoid performance degradation and the impact the various data optimization solutions have on their clients’ data.
  • Attendees will leave the session with a clear understanding of how to select the right primary data optimization for their storage platform, having seen real world examples.
  •  

ETracker – Track files on your laptop and enhance your storage using email

Uttam Kaushik, Manager, Engineering, EMC

Abstract

As humans we rely a lot on computers for doing day-to-day activities. This is driving personal computer data growth at a phenomenal pace – from photos to videos, movies to songs; we want to keep it all “online”. To make space, data is burnt on DVDs or addition of external storage. This results in multiple copies of data at multiple places making it difficult to track files when needed. Small home business (SOHO) imperatively require to backup critical data and from time-to-time need to archive legal documents with proper tracking of changes done to such documents.     This presentation describes an application that can help in versioning, back-up and archive of files residing in a laptop to a public email system, like Gmail or yahoo mail, but can still be accessed from laptops seamlessly. It explains how to leverage Windows NTFS features for managing life-cycle of laptop files.

Learning Objectives

  • Leveraging public domain email system to  backup and archive files on personal computer
  • Difference in this product as compared to other popular tools like Gmail Drive or Gspace
  • How to use NTFS features and free email to facilitate low cost backup/archive solution

/etc

A Case Study: Unique NAS Issues and Solutions at The MathWorks

Ira Cooper, Senior Systems Software Engineer, The MathWorks, Inc

Abstract

The MathWorks is not the first company that comes to mind when one thinks of heavy NAS users. However, our testing environment relies heavily on NAS, and our needs are very different from those of most NAS users. As such, we face a unique set of issues and challenges. This case study will trace the progression of the MathWorks' NAS implementation -- from our start with off-the-shelf vendors to our current homegrown solution. We will detail the decisions we made, why we made them, and what ultimately drove us to develop our own solution. We hope that, by the end of this talk, you will have a better idea of what your clients are thinking and why.


Deep Dive into CIM Client Development with SBLIM

Brian Mason, MTS SW 4, NetApp

Abstract

CIM XML client development can be painful and slow without a good client side toolkit. Some developers choose to make their own, but why? SMBLIM is an open source client side library that simplifies client creation. It has libraries for: C++, Java and with a little help .NET. The presentation covers querying Object, Classes and Association using the programmatic API as well as WBEM Query Language (WQL). Function calls with intrinsic and user defined method and a deep dive into performance tuning queries. As a bonus a brief tutorial on how to use the Java version of the library in a .NET environment is included. All examples are done in Java.

FIBRE CHANNEL

16GFC Sets the Pace in Storage Networks

Scott Kipp, Senior Technologist, FCIA
Mark Jones, Director Technical Marketing, Emulex

Abstract

Storage area networks based on 16 Gigabit/second Fibre Channel (16GFC) will be deployed in 2011 and double the throughput of 8GFC SANs.  This presentation will give historical perspectives on the 7 generations of Fibre Channel and how 16GFC is different from other Fibre Channel speeds. The presentation will investigate applications that need 16GFC and the benefits of using 16GFC.  The presentation will also discuss the latest developments of 32GFC and hardware that is driving the need for more speed.

Learning Objectives

  • FCIA Roadmap from 1 Gigabit Fibre Channel (1GFC) to 1 Terabit Fibre Channel (1TFC)
  • Technical aspects of 16GFC
  • Benefits of 16GFC
  • iSER can be integrated in all the open-source as well as closed source multupathing solutions and achieve the require High Availability.

Fibre Channel over Ethernet (FCoE)

John Hufferd, Owner, Hufferd Enterprises

Abstract

The Fibre Channel (T11.3) standards committee developed a Standard called Fibre Channel over Ethernet (FCoE)   The FCoE standard specifies the encapsulation of Fibre Channel frames into Ethernet Frames and the amalgamation of these technologies into a network fabric that can support Fibre Channel protocols and other protocols such as TCP/IP, UDP/IP etc.  A “Direct End-to-End” FCoE variant has been accepted for the next version of the Standard  The tutorial will show the Fundamentals of these FCoE concepts and describe how they might be exploited in a Data Center environment and its position with regards to FC and iSCSI.  The requirements on the Ethernet Fabric for support of FC protocols will also be shown.

Learning Objectives

  • The audience will gain a general understanding of the concept of using a Data Center type Ethernet for the transmission of Fibre Channel protocols.
  • The audience will gain an understanding of the benefits of converged I/O and how a Fibre Channel protocol can share an Ethernet network with other Ethernet based protocols.
  • The audience will gain an understanding of potential business value and configurations that will be appropriate for gaining maximum value from this converged I/O capability.

FCoE: The Next Generation

Michael Ko, CTO Office, Huawei Symantec

Abstract

The FCoE standard allows host servers on the Ethernet network to access storage on the Fibre Channel SAN. A bridging element known as an FCoE Forwarder forwards frames between the two dissimilar fabric. As currently defined, all FCoE control and data frames must pass through the FCoE Forwarder. The resulting bottleneck problem is unavoidable when the bridging function is required. This is not the case when Fibre Channel devices are gradually phased out and replaced with FCoE ones. This talk will describe the different ways currently being standardized to resolve the bottleneck problem. In addition, this talk will describe an alternative solution which allows customers to retain the use of the FCoE Forwarder for the control plane but bypass it in the data plane.

Learning Objectives

  • A short introduction to FCoE
  • FCoE Forwarder (FCF) and its functions
  • Future FCoE using Control FCF and FDF
  • Future FCoE without FCF
  • Future FCoE with FCF

Open-FCoE Software Initiator(s) – Architecture, Management and Performance

Prafulla Deuskar, Storage Networking Architect, Intel

Abstract

Open-FCoE Software initiator(s) have been released for Windows, Linux and ESX (in progress). Open-FCoE stack has been certified with EMC E-labs and NetApp storage certification.  This presentation takes a look at the architecture of the Open-FCoE software stack for different Operating Systems, how it inter-operates with DCB and how it plugs into management frameworks.  We will also take a look at performance of the stack and how it compares with HW initiators especially with benchmarks which mimic real world applications like JetStress, SQL-IOSim etc.

Learning Objectives

  • Understand Open-FCoE architecture
  • Understand how to manage Open-FCoE initiator
  • Compare/Contrast Open-FCoE performance with HW initiator

This presentation will cover some of these new features which includes:


The Impossible Takes Longer" : Emulating Windows File System Semantics on POSIX

Jeremy Allison, Google

Abstract

Over the years Samba has moved from a thin layer of Windows emulation on top of POSIX to implementing something similar to the Windows "File System Algorithms" layer. If you have to emulate Windows completely on the wire, you need to emulate it completely on top of your platform. As most new storage platforms are Linux-based, learn how Samba manages to create the illusion of Windows on POSIX, and about some of the things that are really impossible to get right.

Learning Objectives

  • Internals of Windows filesystem semantics
  • Internals of POSIX filesystem semantics
  • Advanced Linux storage technology

Implementing Alternate Data Streams in Likewise Storage Services

Wei Fu, Software Design Engineer, Likewise Software

Gerald Carter, Director of Engineering, Likewise Software

Abstract

Modern SMB/SMB2 clients make use of alternate data streams for a variety of application purposes such as desktop UI enhancements, additional document properties, and location information for files downloaded from untrusted networks.  Expectations from end-users and client machines make support for data streams a highly desirable, if not required, feature in today’s storage devices.  Likewise-CIFS is the SMB/SMB2 file server component of the Likewise Open project's Active Directory integration effort and is part of the larger Likewise Storage Services platform.  This session will present both an architectural overview of the Likewise PVFS driver’s data stream implementation and as a case study about the effort required to add stream support into a pre-existing file server.

Learning Objectives

  • Understand the semantics and use of alternate data streams in SMB/SMB2 environments
  • Explain the stream and file object model used in the Likewise POSIX File System driver (PVFS)
  • Distinguish between the mechanism and policy of data stream storage on non-stream aware file systems such as Linux’s ext4

FILE SYSTEM

NFSv4 Protocol Development

Tom Haynes, Ph.D., Senior Engineer, NetApp

Abstract

The NFSv4 protocol undergoes a lifecycle of definition and implementation. We'll examine the lifecycle, what goes into the selection of new features, how these features are refined, and the impact these features will have on end users. We'll also look at how implementation experience will feed back into the protocol definition.

Learning Objectives

  • Understand the NFSv4 protocol delivery model
  • Understand how implementation experience impacts the final protocol
  • Understand the new features being delivered

Ceph Distributed Storage

Sage Weil, Co-founder, New Dream Network

Abstract

As the size and performance requirements of storage systems have increased, file system designers have looked to new architectures to facilitate system scalability. ¿½Ceph is a distributed object store, network block device, and file system designed for reliability, performance, and scalability from terabytes to exabytes.

Ceph's architecture consists of two main components: an object storage layer, and a distributed file system that is constructed on top of this object store. The object store provides a generic, scalable storage platform with support for snapshots and distributed computation. This storage backend is used to provide a simple network block device (RBD) with thin provisioning and snapshots, or an S3 or Swift compatible RESTful object storage interface.  It also forms the basis for a distributed file system, managed by a distributed metadata server cluster, which similarly provides advanced features like per-directory granularity snapshots, and a recursive accounting feature that provides a convenient view of how much data is stored beneath any directory in the system.

This talk will describe the Ceph architecture and then focus on the current status and future of the project.  This will include a discussion of Ceph's relationship with btrfs, the file system and RBD clients in the Linux kernel, RBD support for virtual block devices in Qemu/KVM and libvirt, and current engineering challenges.


Leveraging Btrfs Transactions

Sage Weil, Co-founder, New Dream Network

Abstract

Btrfs is a relatively new file system for Linux built on top of a copy-on-write btree abstraction. Unlike most other file systems, btrfs stores all metadata (and some data) in the btree, and uses a common transaction commit framework to ensure that the file system image is consistent on disk at all times. The Ceph distributed file system switched to using btrfs as the underlying storage for each object storage node because it could hook into the transaction framework to keep its data and metadata consistent at all times. The mechanism used to do this has evolved over the past few years. This talk will provide a btrfs design overview, including its copy-on-write and writable snapshot features, and then talk more specifically about how btrfs's architecture can be leveraged by applications.


RESTful File Systems

Presenter (Pending Confirmed Speaker)

Abstract

Pending


The Design and Evolution of the Apache Hadoop Distributed File System

Dhruba Borthakur, Project Lead, Apache Hadoop DFS, Facebook

Abstract

This talks describes the architecture of the Apache Hadoop Distributed File System (HDFS). It analyzes the evolution of HDFS by discussing why certain design decisions are made, what features are deemed more important than others and the type of applications that use HDFS. It contends that HDFS has been a creative but disruptive force in the world of general purpose file-systems.


GPFS - Scale-out File Storage

John Palmer, (Title Pending), IBM

Abstract

Parallel file systems, once peculiar to the world of engineering and scientific computing, are expanding their reach into mainstream information technology. Industry analysts predict the overwhelming portion of data will be in files as opposed to raw block storage within the next few years, and the majority of that will be in scale-out systems. This talk will present GPFS, IBM's parallel file system, and discuss how it is broadening its reach from supercomputing to the mainstream commercial world. This entails not only growing new features and supporting new workloads, but also subsuming many of the functions of block storage like copy services and disaster recovery.


GPFS-SNC: A Scalable File System for Analytics and Clouds

Prasenjit Sarkar, Computer Science researcher and Master Inventor, IBM Research

Abstract

 
GPFS-SNC is a scale out file system that leverages locally attached disk to provide high bandwidth to data parallel applications. In this talk, I will give an overview of the internals of GPFS-SNC and show examples of its application in MapReduce applications, data warehousing and cloud systems.


Linear Tape File System (LTFS)

Dr. David Pease, Senior Technical Staff Member, Manager, IBM Almaden Research Center

Abstract

While there are many financial and practical reasons to prefer tape storage over disk for various applications, the difficultly of using tape in a general way is a major inhibitor to its wider usage.  We present a file system that takes advantage of a new generation of tape hardware to provide efficient access to tape using standard, familiar system tools and interfaces.  The Linear Tape File System (LTFS) makes using tape as easy, flexible, portable, and intuitive as using other removable and sharable media (such as a USB drive).

Learning Objectives

  • Understand new tape hardware capabilities that enable efficient tape file system implementation.
  • Learn what LTFS is and who is behind it.
  • Understand LTFS capabilities and benefits.
  • Learn how to effectively use LTFS in larger storage solutions.

Interoperability Tools for CIFS/SMB/SMB2

Paul Long, Technical Evangelist, Microsoft

Abstract

Interoperating with Windows can be significantly simplified by learning how to use the tools and resources available to you. Tools such as Network Monitor, Protocol Documentation and Test Suites, and Spec Explorer can help identify interoperability problems and assist in testing implementations. Resources such as Microsoft Support for protocol documentation and Interop Plugfests are invaluable for insuring your software will interoperate properly with the latest version of Windows, in particular CIFS/SMB/SMB2 systems. This presentation will detail the available tools and resources to help you understand the offering.

Learning Objectives

  • Using Protocol Test Suites and Spec Explorer to improve interoperability
  • How to user Network Monitor to identify the correct protocol documentation for a trace.
  • How to use Unified Messaging (ETW) and NetSh to gather traces.

Panel: User mode vs Kernel mode filesystem development

Panelists:
Neal Christiansen, Principal Development Lead, Microsoft
Jeremy Allison, Computer Programmer, Google
Steven Danneman, Senior Software Developer, Isilon Systems
James Cain, Principal Software Architect, Quantel Limited

Abstract

Which is better and why? There is no right answer, hence a group of experienced devs talking about the pros and cons, the available interfaces and libraries on different operating systems and a lively debate about this timely topic.

Audience Participation actively encouraged!


Windows 8: Storage Provisioning and Management

Shiv Rajpal, Principal Development Lead, Microsoft

Abstract

Architecture/approach delivering major benefits for scalable deployments using Windows.


A Lightweight Layered Compressed File System with Hardware Acceleration

Shirish Phatak, VP of Technology Shirish Phatak, Altior, Inc

Abstract

We will discuss a lightweight layered compressed file system that can be layered over either a Linux or Windows native file store. The Compressed File System can achieve 3:1 storage efficiencies using hardware-accelerated data compression while preserving the native file system syntax and semantics. Intelligent application of hardware acceleration enables the compressed file system to run in real time without adversely impacting system throughput. By using sparse file allocation on the native file system, the compressed file system generates no additional meta data.


Windows 8 File System Performance and Reliability Enhancements in NTFS

Neal Christiansen, Principal Development Lead, Microsoft

Abstract

The NTFS File System is a robust, scalable, and performant file system that has been a foundation for Windows for many releases. With Windows 8 we have made even more improvements in the areas of:

GREEN

A method to vary the Host interface signaling speeds in a Storage Array driving towards Greener Storage

Dr. M. K. Jibbe, Technical Director, NetApp

Arun Rajendran, Software Engineer, NetApp

Abstract

This paper describes a method which we can effectively alter the signaling speeds of a Host Interface based on set performance criterion or user defined time of day criterion that are user definable. The end goals are considerable power savings by changing the signaling speeds to a lower supported speed as our background study indicates the same .Bringing down the MTBF of the components by operating them at nominal speeds and improving the operable life span of the system move towards Greener Storage, low power operation, minimize Heat dissipation and emission reduction. We are also achieving demand based Host interface bandwidth allocation to balance the throughput requirements of the application.

Learning Objectives

  • Problems and scope of the analysis and investigation ( Power Consumption differences, Power Bandwidth Ratio, Raw Data rate and Effective data rate)
  • How does the method of this paper make the optimal use of the available raw bandwidth by switching to lesser raw bandwidth if the data rate doesn’t utilize the initial higher bandwidth rate on offer?
  • How are the performance thresholds defined based on the specified performance metrics capable of the array?
  • How does the method of this paper proposal will deliver considerable power savings even when the system is active and performing IOs? (Link Speed Decison Change and Decison Logic)
  • How does the method of this paper address exception conditions and allows rollback to the last supported speed if the system can’t perform a link speed change?

Vibration Management System for Storage Performance

Gus Malek-Madani, CTO and Founder, Green Platform Corporation

Abstract

Gus Malek-Madani, Founder and CTO, Green Platform Corporation, will share 3 sets of test results that demonstrate how normal levels of data center vibration can degrade IOPS and throughput performance in HDDs by as much as 2/3.  These tests also show how this lost storage performance can be restored by mitigating vibration.

Learning Objectives

  • Understand the Vibration Penalty on Spinning Storage
  • Understand how performance-killing vibration exacerbates the storage bottleneck in data centers
  • Understand the benefits of removing vibration
  • See empirical test results that substantiate the "Vibration Penalty"

Hot Topics

Advanced Format in Legacy Infrastructures – Disruptive or Transparent?
Curtis Stevens, Western Digital

Abstract

Since the launch of Advanced Format (AF) technology on hard disk drives in July 2010, many drives have been shipped and integrated using 512 byte emulation standards (AF 512e). As the industry prepares to introduce AF into long-standing legacy infrastructures, including enterprise systems, additional concerns for in compatibility and data loss have be raised. How real are these concerns? Do AF 512e and its 4K native derivative (AF 4Kn) have negative implications to legacy infrastructures? Take this opportunity to learn about the disruptive or transparent nature of AF from industry expert, Curtis Stevens.


Programmable I/O Controllers as Data Center Sensor Networks: Build and Deliver High-Performance Network and Storage Solutions

Sanjeev Datla, Emulex

Shaun Walsh, Emulex

Abstract

Developing and deploying high-performance network and storage applications requires the right tools, right data and the right perspective to bring out maximum optimization. Maintaining that high-performance requires predicting bottlenecks and other performance issues to proactively manage them before they halt your network. Emulex programmable I/O controllers are uniquely positioned at each source and endpoint of data flows in the modern data center, and can provide proactive information to help you prevent issues and reduce downtime through a solution architected from the ground up: OneCommand Vision.

In this session, we will present a brief introduction to the evolution of the next-generation programmable I/O controllers along with a framework of tools and best practices for building, monitoring, managing and deploying host and embedded applications that maximize your network and storage capabilities. We will discuss the key areas that can sap your performance across your application configuration, driver stack, network integration and storage back-end. Additionally, we will present two case studies from our target developer program for storage partners that focus on the tools, tricks and APIs required to make you network and storage applications sing for cloud content delivery, network and storage appliances.


The Role of InfiniBand and Automated Data Tiering in Achieving Extreme Storage Performance

Cynthia Mcguire, Principal Software Engineer, Oracle

Abstract

Extreme storage performance demands a unique and innovative approach to balancing system performance and data integrity. During this presentation, we will discuss the architectural trade-offs and advantages associated with implementing extremely fast storage systems. Including in this presentation will be a discussion on the use of InfiniBand in contemporary storage and system designs, and the associated software mechanisms necessary to take advantage of very low latency protocols.

KEY NOTE AND FEATURED SPEAKERS

Linux File & Storage Systems: Enabling the Latest Storage Hardware in Linux
Ric Wheeler, File System Team Manager and Architect, Red Hat

Abstract

This talk presents an overview of the newest features that we have done in the Linux world and then goes into some depth about how those features got from spec to enterprise customers of Linux distributions.

Recent enterprise class versions of Linux have support for relatively new features in the storage world. This talk will give an update on the status of some of these new features, which versions of Linux support them and how best to partner with the Linux community in testing and evaluating these features. In addition, the talk will give an overview of the Linux development process for storage and file systems and a summary of current work.


The Future of File Protocols: SMB 2.2 in the Data Center
Dr. Thomas Pfenning, General Manager, Microsoft
Jim Pinkerton, Partner Architect, File Server Technologies, Microsoft

Abstract

This talk will introduce SMB 2.2, the next version of the SMB2 protocol, which has been significantly re-designed to support the levels of performance, reliability, continuous availability, scale, and features that are required for server applications. The prior focus of SMB/SMB2/CIFS was in providing Information Worker desktops with shared access to their data. This objective continues with SMB 2.2 to further reduce chattiness and improve end-user responsiveness. However, due to the customer demand for large scale shared file storage in the data center, SMB 2.2 includes major new features to address requirements for transparent failover, bandwidth scale, and application consistent backups. This talk will go through the requirements that server application workloads pose for shared file storage, and describe opportunities for the SMB2 eco-system to move forward to provide shared storage for the virtualized data center, business critical databases, web serving, and other server application workloads.


Scalable Table Stores: Tools for Understanding Advanced Key-Value Systems for Hadoop
Garth Gibson, Professor, Carnegie Mellon University, and CTO, Panasas Inc

Abstract

Inspired by Google’s BigTable, a variety of scalable, semi- structured, weak-semantic table stores have been developed and optimized for different priorities such as query speed, ingest speed, availability, and interactivity. As these systems mature, performance benchmarking will advance from measuring the rate of simple workloads to understanding and debugging the performance of advanced features such as ingest speed-up techniques and function shipping filters from client to servers. This talk describes YCSB++, a set of extensions to the Yahoo! Cloud Serving Benchmark (YCSB) to improve performance understanding and debugging of advanced table store features. YCSB++ includes multi-tester coordination for increased load and eventual consistency measurement, multi-phase workloads to quantify the consequences of work deferment and the benefits of anticipatory configuration optimization such as B-tree pre-splitting or bulk loading, and abstract APIs for explicit incorporation of advanced features in benchmark tests. To enhance performance debugging, we customized an existing cluster monitoring tool to gather the internal statistics of YCSB++, table stores, system services like HDFS, and operating systems, and to offer easy post-test correlation and reporting of performance behaviors.


Leveraging the Cloud for Your Storage Needs
Bret Piatt, Director of Corporate Development, Rackspace

Abstract

Three of out every four U.S. CIOs surveyed by SNIA say they are already using or plan to use public cloud storage offerings. What are the factors that are leading to such a high adoption rate by today’s leaders?

The fact that some Cloud providers are able to make solutions available that previously might have taken a company’s IT staff many months to plan, finance and install has made this a no-brainer for many business leaders.

Rackspace’s Bret Piatt will discuss some of the new services and tools that Rackspace and other leading Cloud providers now offer to help companies manage their storage needs. Bret will also provide an update on OpenStack, a scalable open source cloud computing operating system supported by over one-hundred leading technology companies. He will also provide an update on cloud standards adoption, including the Cloud Data Management Interface (CDMI).


Evolving Enterprise Storage Models Resulting from the Flash Revolution

Andy Walls, Distinguished Engineer and Technical Lead, IBM

Abstract

We will examine how use cases for SSDs are affecting both the adoption of SSDs and server architecture. We will discuss whether SSDs will cause a significant shift back to direct attach storage at least for Flash. If so, we must explore how that storage will be shared and replicated and protected. How are SSDs evolving SAN based storage? Will the form factors for SSDs converge or is there a place for the plethora of formfactors that are appearing? We will show new use cases that exploit SSDs and potential ones in the future.


Apache Hadoop Today and Tomorrow
Eric Baldeschwieler, CEO, Hortonworks

Abstract

Apache Hadoop is having a profound impact on the data industry because of its ability to store, process and analyze very large data volumes in a very cost-effective manner. Eric Baldeschwieler, CEO of Hortonworks and former VP of Hadoop Software Engineering for Yahoo! will provide some insights into how Apache Hadoop is being used today, how it fits into current enterprise data architectures and what's planned for upcoming releases.

NFS

NFS High Availability in Windows

Roopesh Battepati, Principal Development Lead, Microsoft

Abstract

This session covers advancements in high availability for the NFS file services provided in Windows Server. This discussion is centered around using multiple NFS file servers in a failover cluster. The talk will cover briefly cover Failover Cluster Resource model, NFS Resource DLL and NFS Server HA architecture.


NFSv3 and SMB/SMB2 Interoperability in Likewise Storage Services

Evgeny Popovich, Senior Software Engineer, Likewise Software

Abstract

IT professionals are continually striving to reduce the management costs of storage systems and to provide seamless cross-protocol access.  NFS and SMB/SMB2 deployments struggle with three common interoperability problems: how to deal with multiple directories (NIS/LDAP/AD), cross-protocol access control, and differences in file locking semantics.  Likewise Identity Services provide administrators with a means to interact with directory services and the addition of an NFSv3 server to the Likewise Storage Services architecture makes it simple to solve the other two.  The result is a storage layer that does not require users mapping, provides equal access to the same user accessing files from different protocols, works around certain protocol limitations such as the 16 groups AUTH_SYS limitation, and provides support for cross-protocol locking.

Learning Objectives

  • Understand NFSv3 architecture as part of Likewise File Server
  • Learn how the Likewise storage architecture helps to solve NFS & SMB/SMB2 interoperability issues, like access control and locking
  • Understand issues arising from the user-space NFSv3 driver implementation, and their solutions

IETF NFSv4 Working Group -- What's Next?

Spencer Shepler, Performance Architect, Microsoft

Abstract

With the delivery of NFSv4.1 (RFC 5661) in January of 2010, the NFS version 4 community has been busy building and delivering NFSv4.1 products.  Fresh from that experience and emerging application requirements, the IETF NFSv4 Working Group has been busy identifying features and building out a proposal for the NFSv4.2 protocol.  The attendee will be provided an insider's view of the proposed feature set, the timeline involved, and up-to-the-minute status of where the working group is headed.  And most importantly, the attendee will learn what these new NFSv4.2 features will provide for the end application.


Scale-out NAS with NFS Referrals and pNFS

Brad Stone, VP Product Management, Nexenta Systems

Abstract

This talk describes the referral features added to the NFS standard and how to take advantage of the features for  scaling out NAS deployments. This feature will be compared with Microsoft's referral-based system, DFS.  The talk will also cover pNFS, providing parallel data access for NFS clusters.

Learning Objectives

  • What is pNFS and how to contribute to ongoing development
  • Benefits of pNFS for scaling NAS performance
  • What is NFS referrals and how it compares to DFS
  • Benefits of NFS referrals

PERFORMANCE

 

SMB2 - Advancements in Server Application Performance

Dan Lovinger, Principal Software Architect, Microsoft

Abstract

This session discusses SMB2 file services performance, focused on scenarios where the SMB2 client is running an application server workload like SQL Server. This includes extensive comparative analysis of different configurations and specific optimizations for application server workloads.


Performance Analysis of iSCSI & iSER in MPIO Environment.

Seikh Basiruddin, Member Technical Staff, NetApp

Abstract

iSCSI is an emerging storage network technology that allows block-level access to storage devices,  such as disk drives, over a computer network.  Since iSCSI runs over the ubiquitous TCP/IP  protocol, it has many advantages over its more proprietary alternatives. Due to the recent  movement toward 10 gigabit Ethernet, storage vendors are interested to see the benefits this large increase in network bandwidth could bring to iSCSI. In order to make full use of the bandwidth provided by a 10 gigabit Ethernet link,   specialized Remote Direct Memory Access hardware is being developed to offload processing  and reduce the data-copy-overhead found in a standard TCP/IP network stack.This analysis will  cover the performance benefit of using RDMA in iSCSI environment over the normal software iSCSI stack.  This presentation will also cover the benefit in a single path environment as well as multipaths environment.

Learning Objectives

  • While using iSCSI over RDMA, it gives around 20% performance advantage over non-RDMA software iSCSI
  • Used UNH iSCSI initiator & target as its open source, and modified the same to adapt in different environment.
  • As today's iSCSI solution can't utilize the full bandwidth provided by 10 gig Ethernet, iSER can be used to get better utilization of the existing bandwidth.
  • iSER can be integrated in all the open-source as well as closed source multupathing solutions and achieve the require High Availability.

PLUGFESTS

 

Introduction to the SNIA CIFS/SMB/SMB2 Plugfest

To participate in the plug fest, click here.

Abstract

Every year at the Storage Developers Conference, a group of elite engineers hides out in a darkened room with long rows of tables, lots of equipment, plenty of caffeinated beverages, and a guard at the door. What's with that?

The annual SNIA CIFS/SMB/SMB2 Plugfest is an opportunity for CIFS/SMB/SMB2 implementers to test their products for compatibility, exchange results, and to work together to develop interoperable solutions. If your datacenter has a mix of products that all work together, this Plugfest is one reason why.

This session will explain how the Plugfest works, who is there, what they are testing, and how your organization can participate next year. It will also prepare you to ask questions at the Plugfest Open House reception, which follows immediately after this brief talk.

SECURITY

Using Protocol Fuzzing to Harden Storage Systems and to Protect Them from 0-day attacks

Mikko Varpiola, Senior Security Expert, Codenomicon

Abstract

Protocol fuzzing is a proactive method for discovering previously unknown flaws in software. Defects discovered through fuzzing, unless fixed, have a potential of exposing affected systems to Denial of Service (DoS) situations and Zero Day Attacks, which could increase liability, damage business reputation and cripple sales. This presentation explains how fuzzing can be used to harden interfaces of the modern storage system with hands-on examples of protocols such as SMB2 and NFSv4.

Learning Objectives

  • Fuzzing techniques: Random, Mutation, Generational
  • Applying protocol fuzzing to storage protocols
  • Integrating security testing and fuzzing into SDLC

Adding Role Based Access Control onto a Unix Storage Platform

Steven Danneman,Senior Software Developer,Isilon Systems

Abstract

The traditional Unix authorization model defines an all powerful root user who can perform any system task, modify any file, and change any system configuration.  This simple model produces several fundamental problems for a storage platform.  The root user, whether maliciously or accidentally, can cause catastrophic data loss.  They can also view and undetectably modify the contents of any file and thus need to be an extremely trusted individual.  Solving these problems requires partitioning the traditional root administrative rights among many different users and limiting within the file system the ability of any one user to view and modify all files.  This can be accomplished with Role Based Access Control.

Learning Objectives

  • The fundamental problems of the Unix root/other authorization model.
  • The generally accepted abstract model of Role Based Access Control.
  • Comparison between RBAC and traditional file system authorization like ACLs
  • How administrative actions are partitioned in Isilon OneFS using RBAC.
  • Difficulties overlaying a new RBAC authorization system onto the existing Unix process and file system authorization model.

Using Self-Encrypting Storage Devices Today and Tomorrow

Tim Markey, SandForce, Inc.

Abstract

This presentation will focus on the application of self-encrypting storage. After a brief overview of SED architecture, listeners will learn how self-encrypting storage based on TCG specifications can address their security needs.

The second half of the presentation will cover possible development of self-encrypting storage in near future.   - SED architecture overview  - Applications         

  • Transparent encryption with crypto erase (opal, enterprise)    
  • Simple protection mode (Opal, Enterprise)        
  • Multi user storage  (Opal, Enterprise)         
  • Secure OS on TCG Enterprise drive  - Future beyond TCG standards
  • Cache erase        
  • GPIO        
  • TPM
  • Tampering sensors

Learning Objectives

 

  • What is self-encrypting drive and what is under the hood
  • What is self-encrypting drive and what is under the hood
  • How SED is differ from software-based solutions in performance and threat models
  • How to use flexibility of SED to address specific practical needs of organizations: no-pain encryption for device sanitization, ATA-like use, dedicated storage areas, etc.
  • What to expect from the SED technology development in the nearest future

SOLID STATE STORAGE

HDDs and Flash Memory:  A Marriage of Convenience

Thomas Coughlin, President, Coughlin Associates

Jim Handy, President, Objective Analysis

Abstract

This talk will be based upon research by Coughlin Associates and Objective Analysis on ways in which flash memory and HDDs can be combined to create computer architectures with performance approaching that of SSDs and storage costs approaching those of  HDDs.  The talk is based upon a report from Coughlin Associates and Objective Analysis released in 2011 called "HDDs and Flash Memory:  A Marriage of Convenience" and will explore the ways that flash memory and HDDs can be combined in computers,  including tablets.  The talk will give our projections for growth of these hybrid and paired storage products and future developments.

Learning Objectives

  • Learn the ways that HDDs and flash memory can be combined in a storage system
  • Explore the advantages and uses of hybrid HDDs
  • Examine the methods of creating paired storage
  • Learn why flash memory will change the way we use DRAM in storage systems
  • Hear projections for the growth of hybrid HDDs and paired storage computers and storage systems.

Enhance NAND to Expand Enterprise SSD Market

Esther Spanjer, SMART Modular

Abstract

While early enterprise adoption of SSDs has been focused primarily on extreme performance and endurance using SLC NAND flash, proliferation into the enterprise mainstream will be fueled by the value proposition offered by lower-cost MLC NAND-based SSD solutions. Off the shelf, MLC NAND flash is not capable of meeting the endurance and retention requirements of enterprise server and storage applications. In order to meet life time and reliability requirements, SSD designers must use advanced technologies, have access to internal flash features, and incorporate system level design techniques to enhance the native capabilities of MLC NAND flash. This presentation talks about methods to achieve high endurance in MLC NAND flash, including a multi-faceted approach that incorporates advanced signal processing, optimization and adapt ion algorithms, and SSD architectural solutions to create compelling storage solutions that are bound to change how designers look at storage architectures.

Learning Objectives

  • Technical barriers of using MLC NAND flash in Enterprise SSDs
  • Technical solutions to enhance MLC NAND flash capabilities
  • Value proposition of MLC-based SSD in enterprise applications

Emerging Performance Tests for Solid State Storage Devices

Eden Kim, , CEO, Calypso Systems, Inc.

Abstract

Discussion of emerging Performance tests for NAND Flash based solid state storage devices used in Client and Enterprise applications.  This discussion focuses on device level synthetic tests that are tuned to more closely reflect workload characteristics observed in real world use cases.  Difficulties in successful capture and replay of user workloads (IO trace capture and playback) leads to the need for more finely tuned synthetic device level tests that can provide repeatable and comparable performance tests.

Learning Objectives

  • How to measure dimensions of SSD performance with synthetic tools
  • SSD test best practices

PCIe Solid State Storage Devices

Robert Randall, Senior Software Architect, Micron Technologies, Inc.

Abstract

The rapid pace of the evolution of storage devices is creating many new challenges in systems design and architecture.  PCIe GEN2 devices with MSI-X interrupt processing can deliver staggering throughput which can stress the traditional storage stacks on today's operating systems. There are also new standards, NVM-Express, T10 SOP and PQI, which seek to standardize the interface between the operating system and device (much like AHCI did for ATA) for PCIe attached solid state  storage devices.  Early adopters face the traditional trade-offs between legacy compatibility and new architectures which deliver higher performance but diverge from conventional thinking.

Learning Objectives

  • Understand the evolution of storage interconnects (from early parallel devices to many high speed serial).
  • Understand the emerging hardware technologies driving architecture decisions and protocol standards
  • Learn about new protocol standards from nvmexpress.org and T10.org
  • Learn about architectural challenges that PCIe storage devices pose to systems architects and storage software developers.
  • Understand today's layered storage device stacks, their cost, and the possibilities of new slim device stacks and what they offer in return for sacrificing legacy comparability.

Smart Hybrid Storage based on Intelligent Data Access Classification

M. K. Jibbe, Technical Director, NetApp

Abstract

This paper talks about how data access needs vary with application usage and how storage IOPS (Input/ Output Operations per second) can be increased by bringing hard drives and solid state drives together in to a logical volume group. In SAN (Storage Area Network) not always a drive group is accessed. Based on certain application demand the drive group access varies. We consider this varying IO patter (Input/ Output), this pattern is observed and controller firmware learns to identify the drive groups which need more bandwidth. Based on the IO load requirement solid state drives are attached to the drive group. This is done dynamically to improve cache at drive group level.

Learning Objectives

  • This paper classifies the need for data access in to three major categories (1. Mission-critical data or High performance or Sensitive data, 2. Reliable data, 3. Reliable & Sensitive Data)
  • The mode of operation of this paper in term of a) User Classification of data b) Storage Pool creation c) Different RAID levels used by this paper d) Data path virtualization layer to receive SCSI IOP from initiators e) Intelligent data pattern learn logic engine with smart data access classification inside CFW f) Intelligent data pattern learn logic engine with smart data access classification inside CFW
  • IO transaction using SSD for Disk drive groups performance boost for different RAID levels
  • Advantages of this method ( a) Better reliability b) Dynamic Performance boost c) Cost Vs Performance advantage d) Usage of Hybrid drives with NAND flash integrated for disk caching can boost the performance further.)
  • How user can select the data access needs and how the SSDs are allocated to existing disk groups based on the learn cycles from artificial intelligent Data access engine a.k.a Artificial Intelligence Data Access Classification Module

Hybrid Redundancy System New Approach to SSD Redundancy

Avraham Meir, CTO, Anobit

Abstract

Solid State Drives (SSDs) utilize a large number of flash devices. Flash redundancy might be needed to ensure high data reliability and availability.  Meanwhile, legacy Redundant Array of Independent Disks (RAID) implementations are often employed to ensure greater overall reliability and performance at the array system level. By combining SSD-level redundancy with RAID-level redundancy, higher overall reliability and performance can be achieved than when utilizing these techniques separately.

Learning Objectives

  • Analyze the performance/endurance/cost characteristics of legacy RAID schemes
  • Analyze the performance/endurance/cost characteristics of internal SSD redundancy schemes
  • Analyze the performance/endurance/cost characteristics of combined RAID+SSD redundancy schemes

Speeding Up Cloud/Server Applications Using Flash Memory

Sudipta Sengupta, Senior Research Scientist, Microsoft Research

Abstract

Flash is a non-volatile memory technology that sits conveniently in the huge gap between RAM and hard disk in terms of both cost and performance. With its properties of low power consumption, physical ruggedness, and small size, flash has enabled new experiences with many consumer electronic devices. However, it is only recently that flash is seeing widespread adoption in desktop and server applications, in the form of Solid State Drives (SSDs). The new applications of flash involve different storage access patterns (vs. in typical consumer devices) and pose new challenges to flash, due to its device properties, to deliver sustained high throughput and low latency.     We advocate that innovation at the system/application software layer when using flash memory can lead to several factors of improvement in performance  over simply using it as a drop-in hardware replacement for existing storage technologies. The key to deploying flash in the data center lies in (i) designing the software in a flash-aware manner so as to exploit its unique properties and work around its constraints, and (ii) identifying applications that can utilize the sweet spot between cost and performance. As an example of (i), we will present FlashStore, a high throughput, low latency persistent key-value store, that illustrates some guiding principles for designing software for flash -- exploiting fast random reads, minimizing random writes, using RAM space efficient techniques to index data on flash, and recognizing its non-volatile property. As concrete examples for (ii), we will present and evaluate two cloud/server applications that can benefit from a flash-based key-value store – (a) game state backend for Xbox LIVE online multiplayer gaming, and (b) ChunkStash, a flash-assisted inline data deduplication system.

Learning Objectives

  • Flash memory aware software design
  • Exploit benefits of flash memory
  • Design around peculiarities of flash memory
  • Identify applications for flash memory to exploit sweet spot between cost and performance
  • New cloud/server applications for flash memory 

SSDs in the Cloud

Dave Wright, CEO, SolidFire

Abstract

This session will cover three different methods of using solid state drives to provide persistent, high-performance primary storage within the cloud. It will explain the use of solid state as cache, as a storage tier, and as a full data storage solution, covering the advantages and disadvantages of each method. The speaker will also discuss how advances in SSD technology are enabling strides in storage efficiency, as well as performance.

Learning Objectives

  • How to best leverage SSDs within a cloud storage infrastructure
  • How customers are using SSDs in the cloud today and how they are benefiting

How Scale-Up and Scale-Out Flash-Based  Databases Can Provide Both Breakaway High Performance and Breakaway High Availability for  Enterprise and Cloud Datacenters

Dr. John Busch, Founder and CTO, Schooner Information Technology

Abstract

We present emerging storage and database  software technologies providing optimal scale-up through  ultra-high flash and multi-core  parallelism and optimal scale-out through synchronous replication, exploiting commodity hardware advances to yield 10x performance and 90% reduction in downtime, and providing key new building blocks for greatly improving data center QOS and TCO.

Learning Objectives

  • Information and Data Management Technologies :  understand how highly parallel and concurrent software coupled with advanced hierarchical storage management exploit flash memory and multi-core processors for optimal  scale-up and cluster-wide synchronous replication for optimal scale-out
  • Scalable and Distributed Storage Systems : understand the scalability, consistency, and availability trade-offs in scale-up and  scale-out architectures, and the key enabling technologies to concurrently optimize them
  •  Large Data Storage and Management: understand how cluster-wide synchronous replication simplifies large data multi-node cluster management by providing fully consistent data, eliminating data loss, and enabling automatic and transparent fail-over and recovery

STORAGE MANAGEMENT

“Windows Server 8” and SMB 2.2 - Advancements in Management

Jose Barreto, Principal Program Manager, Microsoft

Abstract

This session covers advancements in SMB2 file services management. This includes details on specific implementations of industry standards like Web-Based Enterprise Management (WBEM), Common Information Model (CIM) and Web Services-Management (WS-Man). It also includes discussions on management of Windows and Non-Windows systems providing SMB2 file services.


Microsoft SMI-S Roadmap Update

Jeff Goldner, Principal Architect, Microsoft

Abstract

Microsoft has been working to add SMI-S support to our products. Recent progress is evident in the announcement that System Center Virtual Machine Manager (VMM) 2012 will use industry-standard SMI-S storage providers for active management of storage arrays for configuring virtualized environments. This session will detail the progress around SMI-S support by Microsoft and discuss further work to integrate SMI-S into Microsoft’s management infrastructure.


Implementing a SMI-S Provider from Checkbox to Industrial Strength

Steve Peters, Storage Management Software, PMC-Sierra

Abstract

Data storage continues to grow at a rapid pace and managing the data becomes increasingly challenging.  Complying with SNIA's SMI-S must be more than a check box. The presentation will chronicle the development and evolution of a full featured SMI-S provider and a web based GUI that manages a PCIe RAID card.

Learning Objectives

  • Implementing View Classes
  • Implementing caches to speed access
  • Why implement indications

Proxy Providers Versus Embedded Providers (SMI-S)

Srinivasa Reddy Gandlaparthi, Software Architect, NetApp, Bangalore India

Abstract

The implementation of SMI-S providers for managing Storage Arrays or Controllers often involves selection of type of providers (proxy or embedded). This presentation compares the advantages and dis-advantages of Proxy and embedded providers, design considerations for selecting any one of them along with various   methods of implementing embedded and proxy providers. Issues in managing large number of objects, association traversal related issues and typical capabilities required in the providers to overcome this issue are also discussed. The Management client design consideration while managing proxy and embedded providers is also discussed.

Learning Objectives

  • Design considerations for selecting proxy or embedded SMI-S providers by the Hardware Vendors
  • Resolving issues in managing large number of objects in proxy providers
  • The Management client design consideration while managing proxy and embedded providers

TESTING

RESTful Fault Injector

Jim Rice, EMCs

Abstract

With the advent of cloud storage, REST (Representational State Transfer) is becoming the common method for access and utilization of cloud storage systems. This presentation explains a product development accelerator that can help in reducing the cycle time in development of REST based cloud storage and ISV application’s by using a fault injector.    A RESTful fault injector helps ISV to develop reliable applications for cloud storage. This tool facilitates simulation of error scenarios by injecting REST and HTTP error codes to validate handling of those errors by the application. This presentation explains the high level approach, the implementation details in Windows, tuning to enable usage of this accelerator across various cloud storage systems and configurations to simulate different error scenarios. The details of the implementation are elaborated by taking EMC Atmos cloud storage as an example.

Learning Objectives

  • Get knowledge about cloud storage testing and debugging.
  • Usage scenarios of the error injector to reduce development time.
  • Http responses and error codes returned by REST API.

Challenges of Testing Unified Storage

Peter Murray, Product Manager, SwiftTest

Abstract

Testing iSCSI targets can be difficult because teams often have to rely on existing, OS-based initiators. Testers could be more productive if an OS-agnostic synthetic initiator was available.    This talk presents an extensible and open-source framework for building synthetic SCSI initiators based on iSCSI. It provides ways to build iSCSI-based tests as well as SCSI-based tests. At the lowest layers it uses Ronnie Sahlberg's iscsilib toolkit and uses C++ to build an iSCSI transport layer along with SCSI Request classes that can be easily extend. Examples of extending the basic SCSIRequest class are given along with test code that uses it. The source code is available under the GPL. It could be extended to SAS and FibreChannel.

Learning Objectives

  • Testing iSCSI/SCSI Targets with synthetic initiators
  • Using open-source tools in testing
  • Automated testing of scalable storage

An Extensible Open-Source ISCSI & SCSI Test Tool

Richard Sharpe, Architect, Scale Computing

Abstract

Storage systems that support multi-protocol file, block and object storage are challenging to test.  These system are more complex to test than single access method systems.  Both functional and load testing require new strategies to ensure correct operation and performance.

Learning Objectives

  • Functional Testing Challenges
  • Load/Performance Testing Challenges

VIRTUALIZATION

Advancements in Hyper-V Storage

Todd Harris, Sr. Software Development Engineer, Microsoft

Senthil Rajaram, Senior Program Manager, Microsoft

Abstract

Hyper-V is a virtualization solution included as part of Windows Server 2008 and Windows Server 2008 R2. It provides the ability to expose virtual storage to a virtual machine in a number of different ways, including the use of Virtual Hard Drive (VHD) files. The talk will include discussions on different storage configuration options, workloads, and performance for Hyper-V.


Supporting Virtualization and Large workloads on NAS Storage

Dennis Chapman, Senior Technical Director, NetApp

Abstract

This presentation examines the hosting of enterprise level hypervisor and application workloads on storage provided by NAS servers. It will present a brief overview of the two main file protocols NFS & CIFS/SMB. Then a discussion of the use of NAS storage by a hypervisor and its guests. Next the use of NAS storage by a large database. A discussion on configuring namespace to more efficiently support the hypervisor and application workloads. Discussion on the use of array value-add such as snapshots, dedup or cloning with NAS. Finally a discussion of future trends in this area.

Learning Objectives

  • Basic understanding of current NAS protocols
  • Unique value NAS protocols provide for virtualization workloads
  • Future trends in support of these workloads using NAS

Benefits of ARI support in Virtualization

Sivakumar Subramani, Senior Project Leader, Wipro Technologies

Abstract

As per PCI specification, a single physical adapter can support only up to eight individual functions as only three bits are allotted for identifying a function in BDF (BUS / DEVICE / FUNCTION) value used to refer any PCI device. PCI SIG group has come up with a new method called ARI (Alternative Routing ID) to interpret the Device Number and Function Number fields within Routing IDs, Requester IDs, and Completer IDs, thereby increasing the number of functions that can be supported by a single Device. Alternative Routing-ID Interpretation (ARI) enables next generation I/O implementations to support an increased number of concurrent users of a multi-function device while providing the same level of isolation and controls found in existing implementations. While ARI obviously benefits the virtualized operating environments where each Function can be uniquely assigned to a guest OS, ARI also benefits non-virtualized environments where, e.g. due to increased process improvements, a large number of I/O Functions can be integrated into a single Device. This ARI is used in both Multi function adapters and SRIOV (Single Root I/O virtualization) functionalities to support more number of functions on a single physical function. This paper will analyze the benefits that can be achieved by using ARI in Multifunction and SRIOV (in virtualized environments like KVM, VMware).

Learning Objectives

  • Testing iSCSI/SCSI Targets with synthetic initiators
  • Using open-source tools in testing
  • Automated testing of scalable storage