SGI to launch Ivy Bridge Big Data cluster, NoSQL appliance
Silicon Graphics was, in the 1980s, a brand famous for high-powered graphics workstations. After some corporte restructuring and an asset buyout by Rackable Systems, Inc. in 2009, the Silicon Graphics International (SGI) brand's focus shifted to High Performance Computing (HPC) servers, clusters and appliances. As Intel today launches its Ivy Bridge generation Xeon processors, SGI is ready day-and-date with servers based on the new CPU familiy.
What many may not know is that SGI has for some time offered hardware dedicated to Hadoop, with Cloudera's Distribution including Apache Hadoop (CDH) pre-loaded, as well as an appliance built around MarkLogic's NoSQL database. These products are now also ready to go with Ivy Bridge-based Xeon CPUs and gigabit Ethernet or InfiniBand high-performnce inter-node networking.
Serious Hadoop
In an interview with ZDNet yesterday, SGI's VP of Product Marketing, Bill Mannel, explained the rather rigorous approach SGI takes with these products. SGI doesn't just build that hardware and install the software on it. With respect to SGI Clusters for Hadoop:
- After installing CDH on the cluster, SGI tests everything at the system level, not merely at the chassis level
- The company employs Cloudera-certified technicians, who will install the cluster for customers wishing this assitance
- SGI also has a professional services team, with Cloudera-certified developers who can do MapReduce and other Big Data coding and development
SGI says it holds the Hadoop Terasort benchmark record with its now previous-generation Sandy Bridge-based Cluster for Hadoop. As you might guess, the company expects to improve upon that record with their Ivy Bridge-based product.
No-nonsense NoSQL
On the NoSQL side, SGI has partnered exclusively with MarkLogic, so don't look for DataStax/Cassandra or MongoDB on their appliance products. The NoSQL appliance is available in 1/4, 1/2, full or multi-rack configurations and undergoes the same system-level testing as do the Clusters for Hadoop. SGI will also ship dev machines containing 1, 2 or 3 nodes for those customers dipping their toes in the NoSQL waters.
SGI has a number of U.S. Federal government agency customers for whom they have built very large Hadoop clusters. SGI has also supplied eCommerce giant eBay's PayPal unit with a large Hadoop cluster with which PayPal conducts fraud detection analysis.
SGI's undertakings in Big Data exhibit quite well the mainstreaming of NoSQL and Hadoop and the trend toward the latter's increasing embeddedness in turn-key solutions. In this case, it's not an Enterprise software vendor presenting the turn-key solution but a high-end hardware vendor instead.