STORAGE

  • 08/09/2011
    4:06 PM
  • Network Computing
  • News
  • Connect Directly
  • Rating: 
    0 votes
    +
    Vote up!
    -
    Vote down!

LSI SSD Caching Software 2.0: The Write Stuff

The second generation of LSI's caching software for its MegaRAID 9260, 9261 and 9280 series 6-Gbps SATA+SAS controller cards can now do writes and should "significantly accelerate application I/O performance of hard-disk-drive-based arrays by intelligently caching frequently accessed, or 'hot spot,' data to solid-state drives (SSDs)," says the company.
The second generation of LSI's caching software for its MegaRAID 9260, 9261 and 9280 series 6-Gbps SATA+SAS controller cards can now do writes and should "significantly accelerate application I/O performance of hard-disk-drive-based arrays by intelligently caching frequently accessed, or 'hot spot,' data to solid-state drives (SSDs)," says the company. Originally unveiled just over a year ago, MegaRAID CacheCade Pro 2.0 delivers up to 13 times more transactions per second and 82% cost savings compared to HDD-only arrays, LSI says.

The software uses SSDs as a dedicated pool of high-performance cache in front of HDDs to boost both read and write performance, reduce I/O latencies, accelerate application response times, and lower RAID access and rebuild times. The differences between the first and second releases of CacheCade are equally significant, says LSI. Against HDDs and the first generation of the software, version 2.0 improves performance on four popular applications: TPC-E (Microsoft SQL Server), 1,251 versus 356%; Neoload (Web Server Simulation), 533 versus 374%; SysBench (MySQL OLTP Server), 150 versus 77%; and Jetstress 2010 (Microsoft Exchange), 163 versus 5%.

LSI says it has received great traction from system integrators for the first release, but customers apparently were waiting for the write capability. While no OEMs have been announced for 2.0, both Dell and IBM are selling the first release of CacheCade.

According to a Demartek evaluation commissioned by LSI, "The MegaRAID CacheCade software provides outstanding performance improvements to the web server tested here with minimal configuration needed by an administrator. With one and two 32GB SSDs in the configuration, we observed more than 3x and 5x performance improvements, respectively, in overall throughput and even higher improvement in web server response time. These tests show that the MegaRAID CacheCade solution with SSDs can improve the performance of existing web servers or can provide the same performance with fewer web servers, lowering the total cost of ownership. The performance was significantly improved even when all of the web server content did not fit into the SSD cache."

There are very few systems that already do this the way that CacheCade does, an approach commonly called automatic data placement, says analyst Jim Handy, Objective Analysis. "CacheCade is a really great way of taking relatively common hardware and using software to make it perform to its best advantage." Most systems force the user to approach the problem with human judgment, called manual data placement, which IBM has found to significantly underperform systems using automatic data placement, he adds. "Because of this, I expect for all OEMs to be seriously considering which automatic data placement software to use: CacheCade, other competing software or something internally designed."

While the addition of write caching is very significant, that significance depends on the system's read/write balance, says Handy. "Some systems, notably PCs, have a very low write load--10% or less of the operations are writes--and a system that doesn't accelerate writes might be OK in this environment. Other systems have a higher write load. Online Transaction Processing systems used in banks, flight reservations, Internet commerce and such applications have write loads of close to 50%. These systems will experience a phenomenal boost when migrating from the older to the newer versions of CacheCade."

See more on this topic by subscribing to Network Computing Pro Reports Parallel Memory Optimized (subscription required).


Log in or Register to post comments