The vast majority of High Performance Computing Clusters (HPCC) run some version of Unix or Linux and the traditional HPCC users in scientific research and engineering are familiar and comfortable in that environment.

The price of high performance commodity compute resources has dropped to the point that many mid-sized organizations can now afford them, bringing HPCC to a new group of users.

Windows Server 2008 with HPC08 has matured and offers a familiar environment, enterprise integration, and performance and efficiency on par with Linux.

Rather than making your users choose between Windows and Linux on your new HPCC, there are methods to allow them to use both.

Types of mixed clusters

There are two common types of clusters that can run both Linux and Windows.  One type is a dual boot cluster.  In a dual boot cluster the entire cluster is running one operating system or the other at any time.  The other is a hybrid cluster.  A hybrid cluster may have a mix of operating systems running on the compute nodes simultaneously.

Dual Boot Clusters

Dual boot clusters typically have two different head nodes, one for Linux and the other for Windows.  The Linux head node can run Rocks+, PCM, xCat or other standard cluster management system.  The Windows head node will run Windows server 2008 with HPC 08.

The compute nodes usually have two different partitions, one for a Linux system and the other for Windows server 2008.  To transition from one OS to another a script is run on each compute node to change the grub configuration or change the master boot record on the disk.  When the node is rebooted, it boots the selected OS.  Minor (automatic) changes are also made on the head nodes so that the proper head node will supply services (DNS, scheduling, etc) to the cluster.

Variations on this technique include running a single physical head node with dual VMs for each OS,  diskless Linux compute nodes, controlling the compute nodes with PXE booting, etc. but the result is largely the same.  The entire cluster will be running one OS or the other.

A commercial version of a dual boot cluster is available from Platform Computing as Rocks+Hybrid.

Hybrid Clusters

Hybrid clusters are physically similar to the dual boot cluster, with dual head nodes, diskfull Windows installation on the compute nodes and diskfull, diskless, or mixed Linux installations.  The difference comes with the inclusion of a meta-scheduler that tracks the workload and makes intelligent decisions on how many nodes are needed of each type of OS at any time. 

These decisions can be based on workload requests, reservations, SLAs, or even as part of a complex workflow.  Users can submit jobs to each OS from a standard interface.   

Moab Adaptive HPC Suite™ from Adaptive Computing is an example of a complete hybrid computing solution.

Want more information? Contact us at This e-mail address is being protected from spambots. You need JavaScript enabled to view it !