Scan all the news and blogs about EMC’s Project Lightning and Thunder announcements this week and you certainly get a lot of sturm und drang. Bloggers are shouting everything from EMC announcing another “me too technology” to the entry of a major market disrupter that will put a lot of cool SSD startups like Kaminario out of business.
I’m sure you expect me to say there isn’t much to this announcement, but, frankly, we’re talking about EMC so I’m not going to pretend it doesn’t matter. As Steve Duplessie says in The Bigger Truth, “EMC could put a wad of gum in a box and sell $300M worth.” Before you decide that Lightning and Thunder represent the perfect storm that will overwhelm all those other SSD solutions out there, here are a few things to keep in mind:
It’s cache for a cash cow EMC’s VFCache is just that, a cache. It’s not a true SSD storage device like the K2 or any other number of SSD solutions out there, but more like a performance band aid for EMC’s cash cow, which is disk storage and management. Since it’s a cache, it requires intelligence to determine which data is getting the greatest number of hits and to place that data on the cache. In many cases it will be right, but in a number of cases it will be wrong, especially if your applications are doing a lot of random reads. It also takes time to determine what belongs in the cache, so until it does you won’t get that performance. So in some cases you’ll get great performance, in others you won’t. And since it’s a cache, it’s a duplicate of data that’s already stored on hard disk storage. If you have an application whose immediate and consistent high performance is an absolute necessity, straight SSD is probably a better solution than a slower hard disk fronted by an SSD cache.
It’s a Tiering Solution and it’s EMC specific At some point VFCache is meant to integrate with EMC’s FAST technology, which means it’s meant to be part of an EMC proprietary, though widely-used, intelligent tiering solution. Like a cache, this requires intelligence and data movement. Without true application awareness, the type of storage tiering FAST does has limited usefulness. It also adds complexity and overhead that can have a negative impact on performance.
It’s a READ cache Fast reads are what VFCache and Project Lightning are all about. If you need fast writes, this is DEFINITELY not the solution, as VFCache’s write-through cache sends all writes to disk before it allows that data to be made available for reads in the cache. If your application has a lot of random writes, write-through caching can slow down both reads and writes. For fast writes, the best solution is still a K2 hybrid or DRAM-based solution that takes full advantage of the clear superiority of DRAM over Flash, and certainly over disk (LOL), for writes.
It’s a Server Card Server cards have a lot of performance advantages since they sit right on the PCIe bus close to the server processor. The drawbacks are the lack of “Utilities”—scalability, reliability, availability—that only a robust shared networked solution can offer. We’ve already delved into the advantages of a shared SSD solution for clustered applications and the infinite scalability, availability, and manageability advantages of Kaminario’s SPEAR architecture, so I won’t blather on about them again here. Check our last few blog entries for more details. Just keep in mind again that a true resilient enterprise solution is not about performance alone. You need those “ilities” or, believe me, you’re asking for trouble. To be fair, Project Thunder will be a shared appliance, but again, it’s meant to be either a caching or FAST solution involving automated intelligent data movement and placement, not a dedicated SSD solution like the K2.
It’s Version 1.0 OK, I’m going to get smug here and remind you that the K2 is a seasoned product built from the ground up for SSD performance, scalability, and endurance. The same with the FusionIO cards that provide the Flash in the K2. EMC is just getting out of the gate with this. Expect some issues and revisions.
Project Thunder? Pretty Quiet for Now One of the first questions asked at EMC’s announcement was when we could expect Project Thunder to ship. EMC’s answer? We can’t tell you that. “Early customer engagements” and beta testing will start sometime in the second quarter. Mind you EMC has even put the decision on interconnect technology for Project Thunder on hold for now, and remember that the first announcement of Project Lightning occurred almost a year ago. If you judge the proximity of a storm by the time elapsed between lightning and thunder, this storm is still far away.
But don’t listen only to me. You know my biases. Take a look at some of the blogs out there, including InfoStor, RayOn Storage, Storage Soup, The Bigger Truth, The Storage Architect, Wikibon, and others to get more of an in-depth feel for what EMC’s announcement is all about and decide for yourself. Just remember that from EMC’s point of view, as Mr. Gelsinger said at the announcement, HDD is here to stay for decades to come, really, and Project Lightning and Thunder are architected according to that perspective. At Kaminario, we think the future belongs to SSD and we’ll keep building products for that future.
Tags: application awareness, cache, DRAM, EMC, FAST, Flash, FusionIO, hard disk, HDD, K2, Kaminario, Project Lightning, Project Thunder, reads, SSD, Steve Duplessie, Storage Soup, Storage tiering, The Bigger Truth, The Storage Architect, VFCache, Wikibon, write-through cache, writes