EMC is announcing a major “dot” upgrade to their VFCache technology with version 1.5 In previous blogs here, here, and here I’ve been an early follower of the PCIe storage craze and Project Lightning becoming a product and solution. Chuck Hollis did a nice blog on this here. This announcement is pretty cool. Here’s some details;
“Inline” Dedupe (read & write)
VFCache will now feature a capability to dedupe not only reads but also writes. I’m told this is a derivative of the Data Domain dedupe algorithm. I assumed that because fast inline dedupe is fast for writes but not efficient as reads that there would be an issue. That does not seem to be the case although I have not personally benchmarked it. Knowing that read (or re-hydration) is probably a spindle bound issue the all flash technology.
New Hardware and Performance Upgrades
There’s now a 700GB SLC card plus some updates on supporting multiple cards. Also the caching algorithm got tuned up based on customer feedback and usage experience. For example; larger block sizing for Exchange environments.
Protection against card failure?
EMC VFCache offers performance with protection – any write to the card is persistent on the back end array. In case of a failure, all data can be loaded after replaying the card.
In previous rev VFCache did not have vMotion capabilities which presented a single point of failure issue. This has not bee resolved for Windows 2008 R2 and RHEL just to mention a few. Below are some detailed graphics showing the VSI plug-in, the hardware card, the VFCache software, and the vMotion capability.
Integration with EMC arrays:
In 1.0 EMC was both careful to disclaim that in 1.0 there was no specific integration with EMC FAST. I’m sure some within EMC didn’t let that stop them. Well now EMC VFCache 1.5 has that integration. An EMC array customer will get environment management. They will get alrets, statstics and control of the server side components of his storage (VFCache) as well, from a single point, thus improving the work process.
The integration post 1.5 will include deeper integration with the FAST algorythms, with more capabilites of hinting and tagging, which will enable to optimize FAST even further by placing the data correctly accorfing to the application behavior.
….more to come as this technology is rapidly evolving!