Almost every major data storage vendor, including EMC Corp., Hewlett-Packard (HP) Co. and IBM, is incorporating flash memory into their storage arrays in the form of solid-state drives (SSDs). Yet many enterprise data storage managers are hesitant to use flash-based SSDs because they're not sure where to use them or how to justify the expense.
To begin, some IT managers have trouble even placing SSDs, also referred to as NAND flash, in the technology stack. "They have to think of it like storage cache that's faster than hard disk storage but slower than DRAM," said Jim Handy, SSD analyst at Los Gatos, Calif.-based market research firm Objective Analysis.
The high price also poses a problem. "The cost of NAND flash collapsed in 2009," Handy said. On average, he figures SSD is still 20 times more costly than hard disk storage on a cost-per-gigabyte basis, although prices are expected to continue to drop.
However, if you look at SSD in terms of cost per input/output operations per second (IOPS), the cost equation looks decidedly different. According to Mark Teter, chief technology officer (CTO) at Advanced Systems Group (ASG), "flash is almost 140x faster than the fastest HDD [hard disk drive]. To match the performance of flash using HDDs you would need to aggregate the output of many HDD devices."
As it turns out, a little NAND flash goes a long way. Teter often recommends organizations add some solid-state drives for their high-performance workloads and use low cost, slow Serial ATA (SATA) for everything else.
Still, enterprise data storage managers can't just throw SSD into the storage mix. Most storage controllers will be quickly saturated by an SSD. "The current crop of storage controllers can only handle a limited amount of disk IO. A flash drive as storage will easily swamp today's controllers," Teter noted.
Flash-based SSD uses: caching tier for servers or storage tier
But when you deploy a small amount of flash storage in the right place at the right time and for the right application, flash produces dramatic reductions in the cost of performance as measured in IOPS.
There are currently two general uses for flash-based SSD in the enterprise IT stack: as a caching tier for servers or as a storage tier. Teter currently recommends the use of flash for dedicated caching since it offers more efficiency and flexibility.
The best way to take advantage of this is through a file system or caching appliance, such as Storspeed Inc.'s SP5000 NAS caching appliance, Dataram Corp.'s XcelaSAN, Gear6 Inc.'s Cachefx appliance or Avere Systems Inc. FXT Series.
Conventional file systems generally don't work well with flash storage. The exception is the latest Linux kernel. "The Linux 2.6.30 kernel is chock full of next-generation file systems, including NILFS [New Implementation of a Log-Structured File System] that's showing great promise with SSD drives. At ASG, we've been testing NILFS," Teter said.
By using flash as a caching tier, it can work directly with next-generation host file systems that already know what to do with it. Here, flash can provide application performance acceleration; at a minimum, you could put indexes on flash to boost database performance.
Four ways to use flash storage
Given the price and performance advantages of flash, Teter suggests four ways to use flash in the enterprise IT storage stack today:
1. As cheap IOPS. Organizations that need high performance for processing transactions or other high-volume work requiring top performance can use a few flash units to replace multitudes of underutilized high-performance hard disk drives.
2. Swapping costly HDDs for low-cost HDDs for most applications. Replace most of your costly, high-performance HDDs with lower cost, slower, but still adequate HDDs while using a small amount of flash as cache for those few applications that require truly high performance.
3. To get extreme performance at increasingly lower costs. Where organizations need extreme performance, or where they need to differentiate themselves and gain a competitive advantage through performance, they can use SSD flash instead of conventional HDDs. If cost per IOPS is critical, SSD flash will give them the lowest cost per IOPS now, and it will only get lower in the future.
4. For significant green storage savings. Enterprises constrained by power availability or those facing high energy costs can replace energy-intensive HDDs (especially in performance situations) with a mix of flash -- which consumes far less energy -- and lower performance, large capacity SATA drives, which use less energy than high-performance HDD.
Solid-state drives have the potential to replace high-performance HDDs. At that point, storage managers can use low cost, slower SATA HDD for the bulk of their mass storage needs and deploy SSD where performance is an issue.
About the author:
Alan Radding is a frequent contributor to TechTarget sites.
Discover the differences between flash-based solid-state disk vs. DRAM-based solid-state disk
Learn about SSD trends in enterprise storage
We tell you how to add solid-state storage to your enterprise data storage systems