Making life easy and straightforward.

Our clusters can be supplied fully operational, leaving you more time to perform experiments, run applications, analyse results and reap the benefits of having a cluster. Why spend precious time agonising over the installation and configuration of an entire cluster?

Linux HPC Solution

Our linux clustering solution has been designed and developed with the end goal of making life as easy and straight forward as possible for both users and administrators. We have based our solution around the award winning Rocks HPC OS, which was voted the “Most Important Software Innovation” for the HPC community for the last five years in a row*.
* HPCwire editors and readers choice awards
 
Our cluster architecture positions a central or headnode as the focal point of the cluster. This node is configured to have full control over the entire cluster. The rest of the nodes (known as compute nodes) are all monitored and maintained via this headnode. All user management, node updates, job submission and monitoring is performed on this headnode, reducing the burden of administrating a large system.

Scalability

Once a headnode is installed and configured, the addition of nodes to the system at a later date is a very straight forward procedure, allowing customer to purchase a small system initially and scale up to a large system over time.

Ease of Management

All the compute nodes are installed from the same central system image which is stored on the headnode, resulting in a consistent package configuration across the entire cluster and simplifying the addition of packages to your system. All user configuration and policies are also just configured on this single headnode and automatically synchronised across the entire system.

Queuing/Scheduling Software

Initiate and manage serial and parallel batch jobs remotely (create, route, execute, modify and/or delete jobs). Define and implement resource policies that determine how much of each resource can be used by a job. These are a few of the features of our queuing/scheduling software! Customers have a choice here regarding their preferred queuing/scheduling software. We can offer customers either SunGridEngine or a combination of the torque batch system (previously openPBS) and the Maui Scheduler.

Monitoring Software

Both hardware and software monitoring tools enable users and administrators to observe every feature of the entire cluster. Visual representations of node activity and job throughputs are available through a web interface (using the ganglia monitoring system). Basic hardware management is available by default (Fan speeds and CPU temperatures), but to access a more advanced selection of features, Viglen recommends the addition of an IPMI module to your node.

Parallel Development Libraries

All our linux HPC systems will have both MPICH and LAM-MPI parallel libraries installed as standard, built using the standard GNU compilers. Should any customers require commercial builds of any of these libraries (using the Intel, Pathscale, etc. compilers) there is no problem facilitating this either.