This is one of a series of posts discussing the presenters at Storage Field Day 9, occurring 16th-18th March 2016 in Silicon Valley, at which I am an attending delegate. See links at the end of this post to other presenting vendors. Details of the event can be found on the Storage Field Day 9 page at techfieldday.com and on the dedicated events page on this site at Storage Field Day 9.
I’ve written a lot about Violin Memory over the years, with a range of comments from complementary to not so complementary. For those not familiar, Violin was one of the pioneers of the all-flash market, starting off with the 1010 Memory Appliance in 2008 and moving to flash products in 2010. The basis of Violin’s technology is the VIMM (Violin Inline Memory Module), a custom designed flash module with PCIe interface that allowed the company to build appliances with high performance and low latency while eliminating the issues of spikes in performance that were (initially) seen with systems that simply substituted SDDs for HDDs.
Violin appliances were the ultimate all-flash dragsters, delivering all the thrills, but without the frills – in this case the features that we expect of all flash systems, namely data protection (snapshots, replication) and data optimisation (thin provisioning, compression, dedupe). The problem was resolved through a partnership with Falconstor that jointly delivered features for Violin (implemented as dual controllers sitting above standard 6000 Series appliances) and knowledge of the flash market for Falconstor, which has been used in their FreeStor platform.
Unfortunately for Violin, while they were re-engineering for features, the rest of the market was developing bespoke all-flash solutions based on commodity SSDs (or in the case of Hitachi, developing custom enterprise-class FMDs). The likes of Pure Storage, Kaminario, XtremIO, SolidFire, HPE 3PAR (and lately Tintri, Tegile and Nimble) focused on building systems not components. Today’s flash component market is dominated by Samsung, followed by Toshiba, SanDisk, Micron, SK Hynix and Intel. Samsung has a lead over other vendors and is the only supplier of 3D-NAND technology in commercially available products (although other vendors are quickly catching up). As a result they have been able to announce products like the PM1633a, a 16TB SSD based on TLC V-NAND (Samsung’s name for 3D-NAND). All-flash system vendors can take products from the component leaders (like Samsung) and put them into their products, without lots of complex hardware engineering. This leaves them time (and money) to develop features in software. Violin is tied to using products from Toshiba, a strategic investor in the company and of course re-engineering their hardware to take advantage of advances in NAND.
One other architectural issue I believe Violin has is that the back-end Flash Fabric Architecture (the feature that distributes I/O evenly across all available VIMMs) is designed for a flat, single tier implementation. Today we are seeing the availability of multiple NAND flash products built around the three axes of capacity, endurance and cost (and possibly performance). This allows all-flash system vendors to implement multi-media solutions, either using different flash products as tiers or cache. This results in cheaper more targeted products for the customer and is not a direction Violin seems to have been moving towards.
The Architect’s View
So there’s the problem. The all-flash market has matured past one-dimensional products and with the future offering NVDIMM, 3D Xpoint and other persistent memory products still in development, successful all-flash systems will actually return to being hybrid, albeit not based on HDDs. Getting back to the point of this post, what should we expect to see of Violin Memory at Tech Field Day? Somehow there needs to an evolution of the current technology to compete with today’s all-flash and hybrid offerings. These days, 1 million IOPS and ultra low latency isn’t enough. I hope we see new products announced and some idea of how the company is looking to move forward and address the ever more maturing all-flash marketplace.
If you are interested in a breakdown of Violin’s financials, I recommend checking out Justin Warren’s excellent analysis over at eigenmagic.com.
- Events: Storage Field Day 9
- Storage Field Day 9 Preview: NetApp
- Samsung, Toshiba and SanDisk Lead Global NAND Flash Market (Market Realist, retrieved 8 March 2016)
- Samsung Introduces World’s Largest Capacity (15.36TB) SSD for Enterprise Storage Systems (Samsung website, retrieved 8 March 2016)
- Violin Flash Fabric Architecture (Violin Website, retrieved 8 March 2016, PDF)
- SFD9 Prep: Violin Memory (eigenmagic.com, retrieved 8 March 2016)
- Violin Memory axes quarter of its staff after finances hit a bum note (The Register, retrieved 11 March 2016)
Comments are always welcome; please read our Comments Policy first. If you have any related links of interest, please feel free to add them as a comment for consideration.
Copyright (c) 2009-2016 – Chris M Evans, first published on https://blog.architecting.it, do not reproduce without permission.