BlueArc Clips Clustering Team

One engineer is gone, but NAS player says its clustering strategy remains on track

February 27, 2003

3 Min Read
Network Computing logo

The confusion over NAS vendor BlueArc Corp.'s commitment to building clustering software -- following the dismissal of a key member of its clustering team -- turns out to be a storm in a teacup.

Of the 12 engineers working in this group, one has been fired due to "productivity issues," according to a BlueArc insider. The company declined to provide the name of the engineer or the circumstances surrounding his departure. However, the team's chief architect, Francesco LaCapra, is still on board, and the group is in better shape than ever, BlueArc officials say.

BlueArc's position on clustering hasn't changed. "You're always better off with a single product rather than stringing lots of boxes together when it comes to total cost of ownership and reliability," says Geoff Barrall, co-founder and CTO of BlueArc.

Using the analogy of network switches, Barrall points out that nobody would install multiple 10/100-Ethernet switches to obtain gigabit-per-second throughput. Rather, they'd install a Gigabit Ethernet switch. "You go with the fastest technology available," he says.

BlueArc's highest-capacity box, the Si8900, offers 2-Gbit/s Fibre Channel to disk and supports up to 228 Tbytes of storage. In other words, 228 Tbytes is a helluva lot of storage in a single system, so clustering a couple of these babies together probably isn't necessary for most organizations. [Ed. note: 100 Tbytes is roughly equivalent to 25 billion pages of text, which is quite a lot of data, you'll agree.]Meanwhile, BlueArc's Si8300 supports up to 7 Tbytes of storage capacity, while the Si8700 supports up to 98 Tbytes of storage and delivers 25.6 Tbytes in a single 42U enclosure (see BlueArc Unveils SiliconServer Family).

Still, Barrall says there will be some customers out there, like Lawrence Livermore National Laboratory (LLNL), that require high-availability features like clustering (see Livermore Labs Back for More BlueArc).

The first iteration of BlueArc's clustering software, code-named Agathos [ed. note: Greek for "good"], will be released in the second quarter of 2003. Right now, BlueArc offers an active/standby feature, which means one of its systems can fail over to a second server. The new software will allow any BlueArc server to fail over to a second system -- in industry-speak, n-way clustering.

In the first quarter of 2004, BlueArc expects to ship its single file-system image feature, or global file system. This is the Holy Grail of file system technology: It provides one giant file system that can store every file by every user in an organization, making it much easier to manage files in a large company. (For more on global file systems, see our report, Next-Gen File Systems.)

Meanwhile, NAS startup Spinnaker Networks Inc. says it's providing new clustering features with version 2.0 of its operating system software, available this month. The upgrade provides three primary features: high availability, meaning any server in a cluster can fail and the others pick up the load; nondisruptive data movement among servers in a cluster; and asynchronous mirroring (see Spinnaker Cuts Its Crew and Spinnaker Shoves Off).The two majors in the NAS market -- EMC Corp. (NYSE: EMC) and Network Appliance Inc. (Nasdaq: NTAP) -- currently offer basic failover capabilities. EMC's recently released Celerra NS600, for example, provides built-in high availability features via dual data-mover boards. NetApp is using InfiniBand as a cluster interconnect for its FAS900 series systems (see EMC Darts Into Midtier NA$ and NetApp Uses InfiniBand

Stay informed! Sign up to get expert advice and insight delivered direct to your inbox
More Insights