SSDs and the Reliability Factor

Arthur Cole

The supposed superior reliability of enterprise-class solid state drives (SSDs) has always been taken with a grain of salt by most IT professionals.


The theory is that with no moving parts, the SSD will have a long shelf life because there is nothing to wear out.


The reality, however, is quite different. As Asst. Prof. Samuel Russ of the University of South Alabama points out, every time you inject high-energy electronics --that is, data -- onto the flash, you erode the gate structure. The problem is most acute on multi-level cell (MLC) flash devices because they store multiple bits on each cell, increasing the write/erase factor, but the problem also exists in single-level cell technology. In fact, he argues that inefficient load-leveling and repeated write/erase cycles can actually cause an SSD to burn out faster than a hard drive.


These issues drew a round of criticism at the recent Diskcon conference in Santa Clara, California, where even backers of the technology cautioned users against placing SSDs on highly active storage tiers, according to PCworld. Quite often, the specs issued for new SSDs are just that: performance as measured right out of the box. But as data starts coming in and out, critical parameters like write speed can drop dramatically within a few hours. There can also be wide differences in performance depending on whether the drive is handling sequential or random workloads.


This isn't keeping the design community from pushing the envelope, however. Pliant Technology recently came out with a pair of enterprise flash drives (EFDs) that can deliver 140,000 and 180,000 IOPS with sustained read/write rates topping 500/320 MBps. The company says that with certain log applications, the 3.5-inch EFD LS and 2.5-inch EFD LB can function in heavy workloads 24 hours a day for five years without failure. Both drives employ serial-attached SCSI (SAS).


SSDs are also winding up on all the major server platforms, with HP becoming the last vendor to join the club by adding Samsung 60 and 120 TB devices to the ProLiant G6 and G5 machines. The drives feature the more reliable SLC design and are also outfitted with Samsung's error-correcting code (ECC) as a means to increase the reliability factor. But just in case, the drives are hot-swappable so you can pull out the bad ones with minimal disruption to data operations.


It's tempting in IT circles to be constantly on the lookout for superior technology, and the introduction of enterprise-class SSDs was no exception. Many pundits touted SSD's superiority in terms of throughput, energy consumption and a host of other factors.


When measured head-to-head, however, it's clear that SSDs are not better than HDDs, only different. Both drives will excel in particular data environments, whether that's top-tier critical data or long-term archiving.



Add Comment      Leave a comment on this blog post

Sep 28, 2009 5:56 AM Joe Gleinser Joe Gleinser  says:

I think you mean 60 and 120 GB instead of "adding Samsung 60 and 120 TB devices to the ProLiant G6 and G5 machines."

Excellent info, as usual. I added you to my blogroll this weekend.

Reply

Post a comment

 

 

 

 


(Maximum characters: 1200). You have 1200 characters left.

 

 

Subscribe to our Newsletters

Sign up now and get the best business technology insights direct to your inbox.


 

Resource centers

Business Intelligence

Business performance information for strategic and operational decision-making

SOA

SOA uses interoperable services grouped around business processes to ease data integration

Data Warehousing

Data warehousing helps companies make sense of their operational data