Green HPC Using Adaptive Clusters

Greener data centers continue to be a priority with Platform customers, especially with rising energy costs and growing environmental awareness. There are several tools that Platform is developing to help make everyone's data center "greener" and more efficient.  From simple mechanisms to power-off idle machines to smarter placement of heat-generating jobs within a data center to minimize air flow disruption between your hot and cold aisles, Platform has been exploring, implementing and delivering a number of innovative, earth-friendly technologies. One relatively new solution that may not immediately strike you as a green solution is Platform’s Adaptive Cluster product, which I recently helped set- up for an oil and gas exploration customer of mine...

As is typical in many industries, this particular oil & gas company runs a number of different software packages which require different operating systems. Since many of their applications run on either Windows or Linux, but not both, they had to supply the power, space, cooling, and maintenance costs to support one cluster dedicated to Windows and another one dedicated to Linux. Since neither cluster was running to full capacity, they wanted to learn how to maximize each cluster’s spare capacity and thereby reduce the environmental impact of an idle infrastructure.

My immediate recommendation was to leverage our Platform Adaptive Cluster product, which allows you to maximize a cluster’s spare capacity by running BOTH Windows and Linux operating systems (OS) on the same cluster and automatically provisioning application resources across both. With all the recent advances in virtualization, having a mixed OS cluster is nothing new, however, what is new about Platform Adaptive Cluster is its ability to dynamically change the ratio of running operating systems based on workload demand. Let's say that 75 percent of your day-time jobs require Windows, but at night, Windows OS activity decreases to just 25 percent.  Platform Adaptive Cluster automatically handles the provisioning and de-provisioning of Windows systems to make sure all job requirements get met - regardless of OS.

Having a solution that can dynamically change the make-up of the currently running OS without manual intervention was of much interest to my customer. Using Adaptive Cluster, they’ve been able to really capitalize on their new system by consolidating capacity during low activity hours, saving considerable money on data center overhead costs.  During peak hours, however, they’ve been able to get more compute power across all clusters thanks to the flexibility of Adaptive Cluster.  If it is of interest to you as well, please make sure to check out my next blog on how the provisioning process works and visit the Adaptive Cluster section on our web site at http://www.platform.com/private-cloud-computing/hpc-cloud