Who Says Data Center Storage Has to Be Inefficient?

Data center storage is evolving in ways no one could have imagined just a few years back. The storage subsystems are faster and larger, the interconnections are more sophisticated and flexible, and expectations of cloud capability are reaching near-mythic levels. Trouble is, this has been accomplished with a largely brute-force approach – exacting a penalty in terms of cost, space, power distribution, and cooling. Fortunately, new developments in data center architectures are paving the way for far more efficient storage.

This session will put the spotlight on several important industry trends that allow the data center to provide high performance and advanced storage at substantially lower costs and power than what had been the status quo. The first is shared I/O, which enables the processing subsystems to share the storage devices, thus eliminating duplication. The second is convergence, which allows the processing, communication and storage subsystems to be disaggregated, yet interact more efficiently than they do today. These two trends can be effectively implemented with a high-speed fabric built from PCI Express switching, which can eliminate most of the intermediate bridging devices and additional switch fabrics that contribute to the unnecessary cost and power.

Learning Objectives

How data centers can be architected to enable more efficient storage
How shared IO can enhance a data center's overall cost and scalability
How PCI Express will play a critical role in data center storage