Linux is fast becoming the operating system of choice for high performance super computers. Linux is quick, robust, cost effective and supported by almost all hardware manufacturers. There are a range of open source software projects dedicated to Linux clusters; such as Beowulf, LAM-MPI and LVS (Linux Virtual Server) which make it easy to link physical machines together to form a single super computer.
Since a Linux cluster can be developed from common, off-the-shelf computers utilising free/open-source software, a cluster "super-computer" can be built and implemented at a fraction of the cost of other systems of similar computing capacity.
Typically clusters are used to run complex analysis programs such as LS-DYNA and NASTRAN or for hosting Web sites, proxy servers or databases in a high traffic environment.
Adelix provide the following:
Cluster Hardware
Cluster Software
Case Study 1
Adelix Ltd built an 8-machine LS-DYNA cluster (shown on the right) for a blue chip technology and engineering company in August 2002. The RedHat based Linux cluster replaced their existing Onyx mainframe and just a single machine was nearly twice as fast as the entire Onyx system! Each node comprised dual AMD Athlon processors, 1 GB of RAM, SCSI RAID array and a gigabit network connection. By using our custom built cluster instead of a branded Windows-based cluster, we estimate the company saved themselves over £200,000.
Case Study 2
Although nothing to do with Adelix, we thought this was a great demonstration of an LVS cluster. JANET (UK Joint Academic Network) have deployed 40 squid cache servers in 3 LVS clusters. The cluster has 300Mbit/s of aggregated uplink bandwidth and peaks at ~90million URL requests per day (that's roughly ~900GB of content served each day!).