Violin's arrays play in Concerto - but they can't hit the dedupe notes yet In a major architectural extension Violin Memory has revealed its new Concerto 7000 array. This uses Violin's existing 6200 hardware and gives it a separate high-availability controller front end, with software to provide business continuity, data management and protection features - but not deduplication. Let’s just recap …
Interesting announcement. I was expecting deduplication just because its so common in the flash space. Why is that taking so long? However, replication definitely helps. I'm also curious about the application-integrated snapshots. NetApp SnapManager was cool, but we couldn't manage it at scale. If Violin can do something similar for SQL and Oracle, and make it manageable, that would help them stand out.
Outside perhaps of Violin's existing niche customer base? The company's future is obviously not certain, this is brand new tech not battle tested, the cost is high etc etc.. I'm sure there are use cases where their stuff out shines everyone else but myself I don't see any way they can become truly competitive in the more mainstream market segments, they are just too far behind. On the same note I am not sure who would want to buy the company either.
I can't tell if this thing has legs for the long term, or is just a transitional technology like NFS cache accelerators used to be (Auspex etc.) until the generic stuff gets good enough to not need it.
Certainly, when I read "uses Violin's own inline memory modules (VIMMs) instead of SSDs, and run under the control of VMOS software" my brain sees "proprietary hardware and software, cha-ching might be expensive, me screwed if they go under".
Discuss.
If you take a closer look at Pure, ExtremeIO, Sun, EMC, Violin. Utilizing Intel based gear to provide Fabric access and HA has been the norm for many many years.
Violin originally provided this level of capability with their earlier 1010 and 3200 series hardware running the OpenSolaris heads. 6000 series internalized those two internal servers making them into blades.
This would be just an extension of the existing technology, allowing 4 Arrays to be virtualized - thus the external Control/Concerto heads.
Funny thing, looking and IBMs webpage.
F840 (the flash array) specs state a lot of performance.
V840 (the bundle of SVC and F840) states a lot of features, but ignores performance.
Given that that flash array offers WAY MORE performance than the bundled 2-node SVC cluster can deliver, it is obvious that introducing the SVC layer cripples performance.
Now Violin does the same? Violin that always had performance as key differentiator. I seriously find it hard to believe that Violin has invented something that will allow a X86 gateway to offer so much more performance, than anybody else. (Think SVC - and think NetApp.... Its not like Netapp has offered significant performance boosts over the last decade on a single controller solution - despite the x86 platform having been boosted significantly by more cores, cache and better chipsets).
Why would anybody pay BIG-$$$ for high performance and then cripple it using gateways - instead of buying less-$$$ good-enough performance?
IBM's web page shows performance specs for the V840 - 0.2 ms latency and 1.4M total IOPS. I suspect the total IOPS will get better with the new SVC hardware they just introduced. It makes sense that adding a software layer for data management functions adds a bit of latency and in IBM's case it's minimal.
Violin's making a smart move as I'm sure many opportunities have been stalled by the lack of features and scalability. This virtualization layer is unproven though so that's the next thing they'll have to overcome.
Flash is fast, yes, and now that it's pervasive people are looking for advanced features and easier integration into their environment.
The Register - Independent news and views for the tech community. Part of Situation Publishing
Biting the hand that feeds IT © 1998–2022
