Cache not flash in the system

Don't be blinded by the flashier side of storage, says Dominik Mutterer

Datacentre operators looking for high IOPS often believe flash is a panacea for their problems. However, it's up to the channel to help customer find a better way to solve performance issues without using an expensive, oversized flash sledgehammer to crack the I/O nut.

Flash is most cost-effective when used intelligently, rather than extensively. But even more can be achieved with less by using technologies such as intelligent caching software instead of completely converting to flash.

Workload-intensive business applications are often restricted by the performance limitations of their existing hard disk drives. Since applications are used in real time, bottlenecks are noticed immediately and cause frustration for customers and end users.

To improve performance and negate some of the pitfalls of hard disks, channel partners are sometimes tempted to recommend flash technology to accelerate performance and improve IOPS.

The marriage between HDD and SSD, however, is sometimes troubled and it can be hard for customers to reap the expected benefits, namely a low price per gigabyte and vastly higher capacities for HDD, and high performance for SSD.

Intelligent caching software can be an effective way to unleash the performance of existing hard drive arrays with SSD-based flash cache.

When IT pushes hard disk drive arrays to reach their IOPS potential, data hotspots will appear. To improve read/write speeds and drop latency, a small SSD can be used as a front-side flash cache for a much larger disk array.

To add intelligence to keep the "hottest" data in flash memory dynamically and push cold data onto slow and inexpensive disks, intelligent caching software is needed.

Apart from the substantial improvement in performance for a minimal SSD investment, caching has several other advantages: wasted capacity by short-stroking hard disk drives can be eliminated or reduced, and since the most commonly accessed data now resides on flash, HDD wear is reduced, extending disk life.

In many customer datacentres, storage management and the assigning of business-critical workloads to faster drives or even flash drives is still manual. This requires constant monitoring and causes regular bottlenecks when the chosen settings need to be changed.

The result is far from ideal and it's expensive, with IT staff devoting resources to moving data from different tiered drives when their time could be better used on other tasks.

Software that manages hot and cold data intelligently and dynamically also lowers the total cost of storage ownership for datacentres and SMBs.

Advanced caching software offers both read and write controller-based caching on SSDs, dramatically enhancing the performance gains achieved by previous generations of caching software.

With the addition of write caching support, read/write-intensive workloads such as Exchange server, high-performance computing (HPC) applications, Web 2.0 and other I/O-intensive OLTP database system workloads may perform much better.

Flash has already become a prominent part of many multi-tiered storage options. But the performance improvements that flash offers can be expensive. Use SSD intelligently rather than extensively.

Dominik Mutterer is field applications engineer at Avago Technologies