Last week, EMC officially announced VFCache, fka Project Lightning. To be short EMC puts an PCIe Flash Card in the server that handles read cache for legacy storage and adds write cache for some of the EMC arrays (VMAX, VMAXe, VNX & VNXe).
We have managed to get big volumes out of our arrays but the performance didn’t follow. The last few years all vendors have been working on adding SDDs in serious numbers in their arrays to address these issues(1).
Lately, lets say the last year, there was one vendor that managed to got up to 1.5 million IOPs out of a single server(2): FusionIO. This would eliminate the need of those SSDs in the arrays and customers can choose whatever SAN solution underneath that (I’m simplifying here!). The VFCache is EMC’s answer to that so that the customers would still be buying from EMC. Because FusionIO would be too expensive (to acquire) they take Micron and LSI cards in OEM (not a smart move IMHO!)
As usual when an Enterprise Vendor announces new technologies, the “independent”
analysts have a job stripping the fancy sales words from the actual content. And thats what they did. I don’t need to say the same they did so here are some thoughts(3):
* Chris Evans: EMC Enters The Market With “Me Too” Flash Products
* Martin Glassborow: Cache Splash / Soft Cache
* Mark Farley: New storage soap opera gets rave reviews
* Stephen Foskett: EMC VFCache Is One Small Step, But an Important One
As you might notice the analysts still have plenty of thoughts on what it will really bring to the table. Everyone agrees it is still a Gen1 product but it will challenge the server vendors to get up to speed(4).
But here’s the real reason of this blog: having an opinion is one thing, bashing and selling FUD on official vendor blogs is something completely different. This morning I ran into Tom Joyces blog: No need to band-aid poor performance on older storage arrays as an answer to VFCache. I think Tom has missed the internal note that bashing is so 2011. Leave that for the NetApp guys or Chuck Hollis (#lol). Not only does he make an ethical mistake here, he also makes one of the most critical mistakes in vendor marketing: Tom mentioned 6 times HP and 14 times EMC in this post. Everyone in marketing knows this is wrong!
Even if we take marketing and ethics out, Tom also makes content mistakes. When referring to the VFCache being a band-aid for older storage arrays he does not take into account that he doesn’t work for 3PAR but for HP. Or did he forget that we now have the “new” EVA (P6000) that brought nothing new to the table(5), has the crappy Thin Provisioning, awe-full Tiering and that we are buying HP IO Accelerator Cards (OEM from FusionIO) to get the necessary performance out of it? OUCH!
Dear Tom, I do appreciate your opinion but please: take a blog of your own to do this or even better: let the analysts, resellers and customers do that for you!
(1) EMC needed over 400 flash drives (only 70TB) last year to get an Spec sfs2008 benchmark of 500k IOPs.
(2) FusionIO did a 1.5 billion IOPs on one single server on a live demo at HPdiscover Vienna
(3) I only mentioned the ones that had an opinion of their own. There are plenty others that blogged about the announcement but copying an announcement is the same as linking to the vendors site.
(4) HP announced the Proliant Servers Gen8 yesterday with a catch phrase in between that is called: “Dynamic Workload Acceleration”. I might get back on this when more details are released.
(5) Disclaimer: HP P6000 (and the former EVAx400) are very decent products that have a purpose out there. But it is still outdated technology and just an in between product until HP manages to shrink 3PAR technology for the SMB market.