GridIron TurboCharger
GridIron TurboCharger
Improving application performance is all about having the right data in cache, just before it’s accessed. To do this requires more than simply knowing when or how often a particular data object is read, it’s knowing which data objects are read together or in succession. The GridIron TurboCharger is a SAN-attached read-cache appliance that is deployed between application servers and block storage. Compared with caching appliances which interface with file data, block storage may be a better fit with the big database applications GridIron is targeting. This multi-tiered caching appliance continuously profiles access behavior of applications by collecting 23 different characteristics about data usage to develop what they call a “heuristics-driven map of the data space”. This kind of intelligence enables the cache to be loaded with the data that will be accessed next, producing the 2-10X increase in application performance, 50x increase in IOPS and 1000x decrease in I/O latency that GridIron claims.
Tuesday, April 5, 2011
TurboCharger appliance models provide 2.5TB or 6.5TB of SSD capacity, 1.6GB/s of bandwidth and 100,000 IOPS through 2 front-end and 2 back-end 8Gb FC ports. Appliances can scale linearly, accelerating up to 64TB of primary storage. GridIron uses custom FPGAs to drive the performance for the realtime analytics needed to produce this access intelligence. This knowledge of data access patterns also enables the system to keep appropriate data sets in RAM longer and reduce flash wear issues. As a “SAN proxy device” they can accelerate all servers and storage on the network and can be implemented without impacting hardware, software or data processes that are in place.
Storage Swiss Take
There are a lot of different ways to improve application performance, some of which seem to just throw hardware (and money) at the problem. When storage I/O is the bottleneck these ‘solutions’ can result in an enormous waste of resources as more spindles and now more solid state storage are added. Instead, GridIron has taken a ‘work smarter’ approach by learning the data access behavior of applications to solve the question of which data to cache and when. And, the resulting efficiency has another benefit, it also saves cost by reducing the amount of hardware and software infrastructure required.
Eric Slack, Senior Analyst
SNW Briefing Note
It is April and that means that Storage Switzerland is once again at Storage Networking World (SNW). This spring we find SNW in Santa Clara, California, the home of much of the storage industry. As a result this year we have a record setting number of briefing meetings as well as a speaking session on solid state storage specmenship.
GridIron is not a client of Storage Switzerland
Adding Intelligence to Read Caching Improves Application Performance