DDN Tackles Enterprise Storage Needs as ‘Wolfcreek’ Looms
When it comes to keeping supercomputers fed with data, there are few storage makers that can keep up with DataDirect Networks. But increasingly, DDN is feeling pressure from enterprises that are struggling to keep up with the ongoing data explosion and mixed I/O workloads. That’s where DDN’s forthcoming high-end storage array for the broader enterprise market, codenamed “Wolfcreek,” comes into play.
Wolfcreek is DDN‘s next generation converged architecture for enterprise customers. The system borrows technology from DDN’s SFA12k line of servers, namely the proprietary memory controllers, and gains a new internal PCIe fabric and NVMe Flash drives. But perhaps more importantly, Wolfcreek will also have the enterprise features that SFA12k lacks, including support for virtual machines, compression, data replication, and in-line de-duplication.
The stats on the array–which DDN has been talking about for the past year and which is scheduled to ship in the second half of the year—are gaudy: 5 million I/O operations per second (IOPS), 60 GB per second in a 4U box.
Wolfcreek will make a big mark on the data center, predicts DDN chief marketing officer Molly Rector. “We can deliver up to 4x more IOPS using 70 percent less hardware, and there’s no feature tradeoff now,” she tells Datanami. “In the past, we didn’t have VM support, we didn’t have thin provisioning, snapshots, dedupe, etc. A lot of those features…a customer couldn’t get from DDN. So they had to make a choice.” Not any more.
The changing nature of application workloads is putting pressure on enterprises to adopt so-called converged architectures that can deliver high performance, no matter the I/O pattern. Enterprise customers today want to run big data applications like Hadoop, Web scale systems like Open Stack, and standard IT processes like backup for Microsoft Exchange, all against a single storage array.
“The needs of the enterprise [are] looking a lot more like HPC workloads and DDN’s been getting pulled into the enterprise increasingly,” Rector says. “What customers are often looking for is the capability to use one converged architecture to solve multiple application needs. They want to use multiple applications, some of which are new and write I/O in a very different way than the old-style ones. Some want to use RESTful interfaces, while others want to use POSIX interfaces.”
DDN is counting on its proprietary hardware to give it an edge in handling the divergent I/O patterns creeping into data centers. “We’re using NVMe as the flash component of this solution, then adding to it in the high bandwidth controllers, the DDN technology that we built in to be able to [deliver] fast throughput and mixed I/O, which is where most enterprise vendors fall down,” Rector says. “They can do one or the other, but not both well. It’s the combination of our smart software and how we handle I/O, and some very cool new technology (PCIe and NVMe) on the hardware side.”
The array will support several “embedded” solutions when it ships this fall, including Hadoop, Open Stack, and VM storage for VMware, Microsoft, and Citrix. The company will also be publishing an SPC-1 benchmark that shows Wolfcreek delivering 5 million IOPS, which is twice as many as the benchmark for the next fastest system, from Hitachi Data Systems. DDN hasn’t released pricing for Wolfcreek, but the company says it can deliver 30 percent savings over an equivalent-performing system from a competitor.
As technology advances, DDNs’ competitors will eventually catch up with the first-gen Wolfcreek system. However DDN has a few other proprietary aces up its sleeve, including the forthcoming Infinite Memory Engine (IME) that the company is in the midst of developing for its HPC customers.
IME, which has been in development for years and is currently in beta testing at 16 HPC sites, will optimize I/O by acting as a cache layer between the storage and the compute layers. In compute clusters that are running at only 60 percent efficiency because of malformed I/O, IME will be able to straighten out the I/O, and in so doing, get that cluster running at close to 100 percent efficiency, Rector says.
“I think you’ll see in the next couple years, every HPC environment will be running an I/O acceleration software package,” Rector says. “They can’t afford not to. They won’t get to exascale if they don’t. There’s no way. There’s not enough money… They have to do this. Then we’ll start to see it in the enterprise a year or two later. That’s the adoption curve we see.”
DDN also announced a new release of its object-storage system, WOS 360 2.0. The new release of the product features support for NAS storage protocols to complement its existing REST interfaces. It also brings a more complete security model, with support for encryption at all levels and a new authentication system. It also boasts a new Swift API, which will bolster WOS’ capabilities in the Open Stack environment and added a new high-capacity node that supports 800 TB in a 4U footprint.
This WOS release also brings a new UI that’s designed to deliver a unified view of customers’ storage portfolios. In previous releases, customers needed different GUIs to manage different storage technologies from DDN. That’s changing now, and WOS is the first to get the unified-view treatment.
That unified view is part of DDN’s software-defined storage strategy. “When a customer comes into our converged storage system , the data can sit in solid state, spinning disk, in the cloud or on tape, and all those mediums just look like mount points for the customer,” Rector says. “We’ve been able to integrate any kind of storage technology behind our software-defined layers.”
Rector and DDN co-founder and CEO Alex Bouzari are currently in the midst of multi-city road show to share the DDN vision and evangelize for Wolfcreek and IME, which are set to become available this fall. The Silicon Valley outfit–which claims to be the largest privately held storage company in the world–is clearly trying to raise the bar, in both the enterprise and HPC sides of the business.
As a privately held company, DDN has the freedom to innovate as a startup, Rector says, but at the same time, as an established company, it has the financial wherewithal to fund R&D. That gives it an edge against tier-one storage vendors like IBM, EMC, and NetApp, she says. “In a lot of cases those companies have not been focused on the HPC side of the market,” she says. “A lot of innovation has started to fall off, which gives an opportunity to a company like DDN to take the lead position in the market place as we’ve been able to invest and grow our capabilities.”
The bulk of DDN’s revenues still come from the supercomputing market. About two-thirds of supercomputing sites have DDN storage, Rector says, providing a stable source of maintenance revenue. But in terms of new revenue, 45 percent of sales are coming from the enterprise space. Analytics is the biggest driver of DDN sales, Rector says, accounting for nearly 80 percent of sales across enterprise and HPC sites.
DDN is in a solid position from a product point of view as it seeks to expand its enterprise reach. While it’s reliance on proprietary technologies will cause some enterprises to give pause—especially in this day and age, when open source is viewed by many as an advantage–it shouldn’t be a deal breaker at the end of the day if DDN can deliver the huge performance numbers that it’s claiming.