Wednesday, October 2, 2013

DEPLOYING FLASH IN THE ENTERPRISE

Flash technology is changing the way that enterprises approach storage. After years of use in the consumer market, flash has reached a price point and level of maturity at which it is being actively deployed to address the needs of business-critical applications. Hard disk drives (HDDs) have some nagging deficiencies that make provisioning storage for applications with high-performance demands difficult. Because HDDs are capable of performing no more than 300–400 random I/O operations per second (IOPS), a storage system capable of delivering tens of thousands of IOPS requires hundreds of disks—even when the capacity is not needed. Over provisioning disks to achieve performance goals is a significant capital expense and wastes rack space, power, and cooling. High-performance workloads increasingly require 100,000 IOPS or more, further exacerbating the problem. 

Flash is quickly emerging as the preferred way to overcome the nagging performance limitations of hard disk drives. However, because flash comes at a significant price premium, outright replacement of HDDs with flash only makes sense in situations in which capacity requirements are relatively small and performance requirements are high. Deployment approaches—including hybrid storage arrays, server flash, and all-flash arrays—that combine the performance of flash with the capacity of HDDs can be cost effective for a broad range of performance requirements. Some storage companies offers a full range of flash solutions, including server flash, hybrid storage arrays, and all-flash arrays. We’ve done a careful analysis of the cost of each solution at various combinations of performance and capacity to help you understand how to choose the best solutions to address your storage challenges based on your performance needs (IOPS and latency), capacity requirements, working set size (amount of hot data), budget, and data protection objectives.  

The fastest HDDs have access times of 3–4 milliseconds, resulting in latencies much slower than flash-based SSDs, which have latencies measured in microseconds and perform thousands of IOPS per device. HDDs alone may no longer meet the needs of latency-sensitive applications.  Because of clear performance advantages coupled with significantly lower power consumption, flash SSDs and other flash devices are beginning to take the place of high-performance HDDs. However, because SSDs currently cost more than 10 times as much per GB of usable storage, IT teams are still  searching for the best strategies to deploy flash technology to deliver performance where it’s needed while minimizing overall storage costs.

There are a number of options for deploying flash in the data center:
  • Hybrid storage solutions combine the performance of flash with the capacity of HDD by targeting hot data to flash using either migration or caching.
  • Server flash solutions may provide persistent solid state storage or cache data from HDD storage onto flash devices installed in servers, delivering extremely low latency for data accessed from cache.
  • All-flash arrays provide maximum performance and a high level of consistency for business-critical applications.

No comments:

Post a Comment