New Zealand"s Massey U gets Linux supercomputer

23.11.2004
Von Juha Saarinen

New Zealand"s Massey University took delivery last week of its new Double Helix cluster computer and says it packs anywhere between five to ten times the computational power of the university"s existing supercomputer, the 64-node Helix 1.

Whereas the Helix 1 cluster, commissioned about three years ago, consists of 32-bit computers in normal PC cases sitting on shelves, the 64-bit Double Helix is mounted in a 19" rack, and takes up less than a tenth of the space of the old setup.

Peter Schwerdtfeger, director of the theoretical and computational chemistry research center at Massey University, says the new cluster probably beats anything currently available in New Zealand in terms of raw computing power.

Insite Technology in Christchurch built the cluster for Massey University, worth about $300,000 including the gigabit ethernet network used to connect the 19 slave nodes with the master node with a managed Cisco switch.

Schwerdtfeger and server business development manager Oliver Mould say the nodes are equipped with dual AMD Opteron 250 processors that run at 2.4GHz and have 1MB of L2 cache each. The Opteron 250 also uses AMD?s HyperTransport bus, running at 800MHz. Mould says Double Helix can be upgraded with 10Gbit/s Infiniband for interconnecting the nodes.

Each slave node has 4GB of RAM, with the master controlling node having 16GB RAM. Insite mounted all of that plus dual 120GB IDE hard drives on a Rioworks HDAMA server motherboard, and squeezed the lot into 1U 19? Chenbro RM311 chassis.

Double Helix runs the open source Rocks Cluster Linux distribution. Schwerdfeger says the OS and new set-up is considerably easier to manage than the old one ? changes only need to be made to the Master Node, from which they are then automatically propagated to the Slave Nodes.

Massey says the extra computational power will be welcome for its student and researchers in the areas of bioinformatics and computational chemistry. Users of the existing Helix 1 cluster are currently experiencing long queues during busy times, sometimes up to 19 days and Massey hopes Double Helix will drastically reduce the waiting times for students? computational tasks.

Double Helix will also enable Massey?s researchers to participate in overseas projects where its academic counterparts already enjoy access to high-power super computer clusters.

Asked if 64-bit cluster computing would have users outside academia and render farms for film studios such as special effects wizards Weta, Mould says that commercial customers running SQL Server databases, for example, would benefit from it. Large amounts of high speed memory means big databases will run much faster; Mould mentions a Microsoft financial risk analysis benchmark that ran in 13 hours on AMD?s 64-bit Opteron platform instead of 40 hours on existing 32-bit computers.