Experiments in Storing Scientific Research Data in Cloud Cold Storage Services

Nowadays it is common for major cloud providers to provide cold storage services, which target to store data with relatively low access frequency for a long period, as a part of their public cloud services. The adoption of cold storage services is a viable alternative which reduces the TCO of maintaining large amount of data for a long period. In order to acquire the fundamental information which helps establishing best practices of cold storage adoption particularly for scientific research data, the presenter experimented on commercially available public cold storage services by storing large amount of data up to 1PB as well as storing and accessing the actual scientific research data such as observational data and simulation data. This presentation includes the survey of cloud storage services and some results of the experiments.

Learning Objectives:
1. Understand the specifications of commercially available cloud cold storage services from the viewpoints of functionality, performance, and pricing
2. Understand the performance characteristics (in terms of throughput and latency) of cloud cold storage services observed in experiments
3. Understand the performance and manageability characteristics of cloud cold storage services observed in experiments of storing large amount of data up to 1PB
4. Understand the performance, cost, and manageability characteristics of cloud cold storage services observed in experiments of storing and accessing actual scientific research data