August 25, 2009
Protecting data is probably one of the more mundane tasks of the datacenter administrator. That's one reason why storage maker Spectra Logic has not exactly become a household name. The company specializes in tape and disk storage systems for data backups and archives. Based in Boulder, Colorado, Spectra Logic has been around for nearly 30 years, serving mid-sized and big enterprise users with a need for mission-critical data protection and HSM (Hierarchical Storage Management). But within the last 18 months, the company has put a special focus on high performance computing customers.
In particular, Spectra Logic wants to become the archive storage vendor of choice for large HPC installations, which typically need to manage petabytes of data. The company has a good start in making this a reality. Today their products are in 7 of the 10 largest TOP500 supercomputing sites, including the DOE national labs at Los Alamos, Sandia, and Argonne. In April, the company displaced Sun Microsystems at NASA Ames, replacing the 10 StorageTek 9310 silos with two of its top-of-the-line T950 tape library systems (and in the process, freeing up 1,400 square feet of space). It also has a presence at approximately half of the NSF TeraGrid sites and is looking to expand to the remaining half.
Spectra's entry into HPC was mainly an outgrowth of its federal government business, which represents about a third of the company's revenue. Over the years, it established accounts in essentially all US federal agencies. Its segue into the HPC side of these organizations was a natural one since many of these same agencies maintain in-house supercomputing infrastructure. Spectra also has HPC customers in the commercial space, in both the financial services sector and the oil and gas industry.
The company also maintains a customer advisory panel, and according to Molly Rector, Spectra Logic's vice president for product management, 50 percent of the members are HPC users. The HPC group is helping Spectra guide its multi-year roadmap so that future storage product capabilities satisfy the demands of the supercomputing systems being planned for the early part of the next decade.
The competition in the archive and backup business is spread across many storage vendors, but at the top end, it is more restricted. According to Rector, their main rivals for high end tape storage are Sun Microsystems (StorageTek products) and IBM. In the 90s, StorageTek dominated the big tape archive space. With its acquisition by Sun Microsystems and now Sun's acquisition by Oracle, the StorageTek story just seems to be getting more uncertain. Rector thinks customers are starting to look for alternatives and is hoping that the recent NASA AMES win is just the start of that trend.
Right now Spectra is also enjoying some of the advantages of being a privately-held company. Since it doesn't have shareholders to satisfy, it has more freedom with investments and company strategy. "Being a private held, smaller company, we have a lot more flexibility to architect and design our products to meet unique customer needs," says Rector. There's probably some truth in that, given that IT companies that are fully exposed to Wall Street expectations have often struggled to evolve their product offerings, especially during market downturns.
Spectra has also managed to avoid being the target of a buyout. Last week the company reported its third consecutive year of profits, and a 12 percent increase, quarter over quarter. Spectra attributed the positive results to higher sales of the company's very large and medium-sized tape libraries. Its main markets -- HPC, media and entertainment, and federal government -- have been among the most recession-proof.
If the company has a weakness in the HPC space, it's in its partnerships. Currently, SGI is the only system vendor reselling Spectra's tape storage gear for this market. The remainder is being sold directly via Spectra's own sales force, which consists of about three dozen individuals. "Usually in these big HPC sites, it is our resources that are selling to them," says Rector. Without a lot of company name recognition, it's important to get system vendors on your side to take advantage of as many opportunities as possible, especially for the numerous mid-range system deployments in the commercial and academic sectors. According to Rector, the company is actively working to expand its HPC vendor partnerships.
The US is Spectra's biggest market, but the company also maintains offices in Europe, where it has a handful of HPC deployments in the UK, Germany, Switzerland and France. Spectra also has deployments in Asia*.
At the big national lab HPC sites, Spectra is mainly selling its tape storage gear (T50e through T950) for nearline archive storage, that is, storage that is less frequently used than primary disk storage, but needs to be online for relatively quick access. For these big installations, the nearline storage will typically sit behind a DataDirect primary storage tier. The Spectra storage is slower, but denser, and not nearly as expensive to run power consumption-wise.
Spectra's newer storage offerings, the disk-based nTier product line, have been deployed for some finance and oil & gas HPC applications, but have yet to make it to the big government supercomputing sites. The obvious advantage of disk-based archive storage is faster access time, but when paired with a tape-based archive system and a Hierarchical Storage Management (HSM) application, users can realize a lot more flexibility in managing their data.
Spectra's latest solution integrates disk- and tape-based archive/backup hardware with HSM and deduplication technology via its BlueScale management software. Since deduplication basically substitutes a pointer for redundant data, it typically is able to offer a 15:1 or better compression ratio, depending on the dataset. That means that not only can you store 15 times as much data on the media, but also you can effectively send 15 times as much data across a network. Rector says they are starting to see some traction for deduplication with commercial HPC customers that maintain big tape archives in a central repository. The idea is that remote site data can be sent to the central archive very efficiently, since duplicated data doesn't need to be sent at all, just accounted for. All this can take place transparently once the policies are in place.
Spectra has also added features to the latest BlueScale software to make tape administration less labor intensive. To that end, version 10.6 adds a handful of new capabilities that enable tapes to behave more like disk arrays, including "hot spare" drives for remote failover, proactive notification of storage components reaching their lifetime thresholds, and auto-discovery of new media. In disk storage as well as other areas of the datacenter, replacing human administration with management software has become a universal trend driven by economics. "Our development goal is that the expectations you have of your disk, you should have of your tape as well," says Rector. "Those get closer and closer together all the time."
* The original version of this article incorrectly stated that Spectra was not present in the Asian market.
10/30/2013 | Cray, DDN, Mellanox, NetApp, ScaleMP, Supermicro, Xyratex | Creating data is easy… the challenge is getting it to the right place to make use of it. This paper discusses fresh solutions that can directly increase I/O efficiency, and the applications of these solutions to current, and new technology infrastructures.
10/01/2013 | IBM | A new trend is developing in the HPC space that is also affecting enterprise computing productivity with the arrival of “ultra-dense” hyper-scale servers.
Ken Claffey, SVP and General Manager at Xyratex, presents ClusterStor at the Vendor Showdown at ISC13 in Leipzig, Germany.
Join HPCwire Editor Nicole Hemsoth and Dr. David Bader from Georgia Tech as they take center stage on opening night at Atlanta's first Big Data Kick Off Week, filmed in front of a live audience. Nicole and David look at the evolution of HPC, today's big data challenges, discuss real world solutions, and reveal their predictions. Exactly what does the future holds for HPC?