lassedesignen - Fotolia
New Violin Systems CEO talks up ambitious storage roadmap
New Violin CEO talks up plans for newly architected, NVMe-based flash array that will focus on enterprise software capabilities such as low RPO and deduplication.
Despite scaling down its business operations, flash storage pioneer Violin Systems still plans an ambitious product roadmap that includes a newly architected array code named Tanqueray.
New CEO Todd Oseth said budget constraints will prevent Violin from finishing Tanqueray by its original 2019 target date. The new system is now targeted for August 2020. Violin cut about 30 Calif.-based staffers over the summer and relocated its headquarters from San Jose to Colorado Springs last month.
Violin Systems continues to sell and support all products, including the X-IO line that the company got through its acquisition of Axellio's storage division in 2018. Oseth said Tanqueray would combine the best of the Violin and X-IO product features and use standard hardware components.
Oseth recently became Violin's third CEO in less than three years. He said former CEO Mark Lewis left by choice in August. Oseth joined Violin as COO in April 2018, shortly after Lewis was named CEO. Lewis replaced Ebrahim Abbasi, who left due to health issues.
Abbasi became CEO in April 2017 when Quantum Partners, a private investment fund controlled by billionaire George Soros, rescued Violin from Chapter 11 bankruptcy. Violin's ownership is now 90% Soros and 10% Axellio, according to Oseth.
Oseth formerly worked under Lewis at EMC, where he was responsible for enterprise features as VP of infrastructure software. He also served as CEO at FalconStor, Intermap Technologies, Neterion and Sanz; and COO at McData, which was acquired by Brocade. During an interview with TechTarget, Oseth discussed the changes at Violin Systems and the company's technology roadmap.
How did the recent spending cuts come about at Violin Systems?
Todd Oseth: It was done at the board level, saying spending at the rates that we were spending at just was not going to be maintainable. The first reductions were always part of our plan. Any time you have two companies that merge, you're going to have an abundance of people doing similar things. We had duplications of a huge number of positions. Then the second [set of reductions] was: Now we have to get things more operationally focused. We had to minimize some of the capabilities of the company so that we can really focus on profitability.
Why did Violin move operations to Colorado Springs?
Oseth: X-IO was in Colorado Springs. If you compare operating costs between San Jose and Colorado Springs, everyone would find themselves out of San Jose. In San Jose, electricity is three times what electricity is in Colorado Springs. The dollars per square foot are almost 12 times. So it makes a lot more sense for investors and for everybody to find the lowest cost operating facility that we could possibly find.
How did the budget cuts affect product development?
Oseth: The rate at which we're bringing out Tanqueray has slowed down quite a bit. Many of our execs decided they wanted to have a faster rate of development, and our investors were not prepared to do that. So, we ended up just slowing things down.
Will Tanqueray integrate code from the Violin Systems and X-IO products?
Oseth: It won't be the exact code base, because they are totally different architectures. Our new generation is a new architecture that will utilize all the enterprise features that Violin had for years with a lot of these lower-cost approaches of X-IO.
Can you give examples?
Oseth: A lot of what makes Violin an exceptionally fast array has to do with the way that we handle timing of writes and garbage collection. All of that technology will continue to move forward in future arrays. On the X-IO side, they've been masters at using more off-the-shelf product instead of custom. Violin's systems -- the XVS 8s, as well as all the 7000 Series-- every board in that system is custom, making it very costly for our company. It's very difficult to be cost-competitive when everything is custom. So we're moving to software differentiation, and software is going to be the storage.
You will see an updated [XIO] G4 product early next year that will be all NVMe based. Once that NVMe is in place, we'll then start getting experience in all of the other storage procedures prior to Tanqueray coming out in the middle of next year. With Tanqueray, our software will run in the cloud, as well as on our platform. So you will be able to migrate seamlessly back and forth between the cloud and an array.
Will the Tanqueray software run in a virtual machine in the cloud?
Oseth: You bet. Fundamentally, we are believers that the cloud is not the answer. A hybrid cloud is an answer, because you're going to find there are certain applications that you need to run locally, and there are other applications that are more economical to run in the cloud.
Will Violin still make its own SSDs?
Oseth: We call them DIMMs -- Violin in-line memory modules -- but they were very, very expensive to make. We're not doing that anymore. We're too small. When you're under $50 million [in revenue], you've got to really look at the economics and your operation. You have to make sure you understand where all the cash flows are going and the expense of staying up to date with custom products.
How would you characterize Violin's customers versus X-IO's customer base?
Oseth: X-IO had a high performing box that was much cheaper than Violin. So their customers were not necessarily Fortune 100, whereas Violin customers were the Fortune 100. They were big banks. They were big organizations that really knew where their applications must have performance.
Has Violin retained those customers?
Oseth: We've lost no one, and we continue to support 'em all the way through. We've got people that are at their sites working with them. We treat 'em like kings.
Wasn't one of Violin's main selling points for the banks and large customers the custom hardware that gave the arrays higher performance?
Oseth: Correct. The custom hardware did give it an edge for that point in time. But the coolest thing about Moore's Law is that, as time goes on, those functions could now be done inside processors instead of custom hardware. So we're not losing performance. NVMe by itself is a plus to everybody because now the hardware architecture itself has reduced latency. The old SAS architecture has much more latency associated with it than the NVMe [architecture]. NVMe is where the future is going. Everyone that's building storage arrays is doing everything on NVMe.
What functionality will you offer that no other vendors have?
Oseth: We're going to really focus on recovery point objective. We believe that one of the biggest challenges that these large enterprise people have are viruses. If you can turn back time easily, so that these viruses will not infect your entire array or all of your storage pools, that is a huge advantage.
Everybody has RPO. It's a matter of doing it at one-second intervals or one-minute intervals or five-minute intervals. And most of the industry is at the one [minute] to five [minute] range right now. A virus can go pretty far inside your environment in five minutes. We believe we'll have the lowest RPOs in the industry.
On the hardware side, do you plan to use Intel's Optane technology to boost performance?
Oseth: When Optane gets to be a little bit more economical. We've looked at it. We've actually had some Optane in the lab. Right now, it's still more cost-effective to put a battery on the DRAM or put a battery backup system on the entire array.
Is Violin's architecture set up to be able to work with Optane and other new memory technologies?
Oseth: Yes. We're abstracting ourselves away from the hardware, so let the better technology win. We don't care. Our arrays will run on fully loaded Optane systems or on an SSD made with flash.
Do you think customers need the level of performance that Optane or other new types of memory promise?
Oseth: It is totally safe to say that 10% to 20% of the entire market really needs performance. Many of the rest still are moving from rotating media to flash, and there's a couple of orders of magnitude performance [improvement] just in that move. So, if you were one that was satisfied with rotating media, the movement to flash is a huge win.