Breaking down the Reliability Barrier for SSDs
It is well-known that while SSDs don't suffer from mechanical wear like their hard-disk brethren, continued writing and rewriting breaks down the flash layer over time, making it difficult to mount SSDs in high data load environments.
The problem, naturally, is more acute on consumer-grade laptop devices than the newer enterprise-level systems coming out. Nevertheless, even the top-level SSDs have had trouble catching up to the reliability and durability benchmarks set by enterprise HDDs.
But that may be changing. A new round of SSDs is about to hit the channel with claims, at least, of increased reliability -- enough, manufacturers contend, to put them in the most demanding environments.
One of the latest is from Israeli start-up Anobit Technologies. The company's Genesis device features the more rugged multi-level cell (MLC) design, plus a proprietary gizmo called a Memory Signal Processor that provides additional error correction beyond the standard ECC component common to most SSDs. The drive is available in 200 or 400GB Serial ATA designs and provides sequential read/write speeds of 220/180 MBps. Anobit is targeting high-duty cycle applications like relational databases for initial deployments.
Another newcomer is the tachIOn drive from Virident Systems. In addition to 1.3 GBps sustained read bandwidth, the drive features end-to-end error correction and a combination of global and local wear-leveling, which the company says is suitable for such data-intensive applications as rich media, Web 2.0 and data analytics. The unit is designed for 1 or 2U servers and is suitable for Tier 0 build-out deployments in conjunction with the Virident GreenCloud software stack that provides for pooled resources through capacity defragmentation and resource management tools.
OCZ is also out with a new enterprise-class drive, the Deneva, which it's targeting at Tier 0 applications using the MLC and eMLC capabilities of the Sandforce 1500 SSD processor. The company has also added an emergency power-loss module called a supercap, along with enterprise-level encryption and ECC protection. Throughput maxes out at 285/275 read/write and the unit can deliver 50,000 IOPS with up to 4 KB random writes.
Claims issued in press releases are one thing -- actual field experience is another. So the final determination as to actual reliability can only be made through real-world deployments.
However, it's probably safe to say that SSD reliability is improving, and so should the comfort level in placing them in mission-critical environments.