Following on his talk "HPC Trends and Virtualization" a presentation given at Sun's HPC Consortium in Hamburg recently, Josh Simons goes further with his blog post "Virtualization for HPC: The Heterogeneity Issue, a subject to which he takes an even-handed approach, conceding at the outset that " ... while heterogeneity is either desirable or to be avoided, depending on your viewpoint, virtualization can help in either case."

Simons points out the shortcomings of a typical HPC cluster with each compute node running whatever distro was settled on as the site's standard OS. While such a configuration simplifies the job of the sysadmin, it penalizes the end-user, who is denied the flexibility to utilize the centralized facilities if their respective applications run on different OSes.

Simons also addresses the issue facing software providers whose product by design needs to run on more than just the OS it was developed on. Here the task is the continuing one of qualifying and requalifying the application to accommodate other supported operating system versions in the interest of maximizing revenue.

The solution to these unwieldy situations that Simons proposes is to establish a virtualized infrastructure in which cluster nodes run hypervisors on which pre-built and pre-configured software environments (virtual machines) are run. These virtual machines include the end-user's application and the operating system required to run that application, he writes.

What, then, would be the payoffs for such a system:

• End-users: End-users gain the complete freedom to run any application using any operating system because all of that software is wrapped inside a virtual machine whose internal details are hidden. The VM could be supplied by an ISV, built by an open-source application's community, or created by the end-user, he continues, adding that, because the VM is a black box from the cluster's perspective, the choice of application and operating system need no longer be restricted by cluster administrators.

• Cluster admins: Simons points out that, in a virtualized world, cluster administrators are in the business of launching and managing the lifecycle of virtual machines on cluster nodes and no longer have to deal with the complexities of OS upgrades, configuring software stacks, and handling end-user special software requests. This is rather like the job of running a shared, virtualized HPC infrastructure, similar to what has come to be known as a public cloud infrastructure like EC2.

• ISVs: And, in a virtualized world, ISVs are free to select a single OS environment for developing, testing, and deploying their application, Simons notes. Rather than basing their operating system choice on market share considerations, the decision could be made based on the quality of the development environment, or perhaps the stability or performance levels achievable with a particular OS, or perhaps on the ability to partner closely with an OS vendor to jointly deliver a highly-optimized, robust, and completely supported experience for end-customers, he concludes.

More Information

Virtualization for HPC: The Heterogeneity Issue

Sun HPC Consortium 2009 Videos and PDFs [...read more...]

		fullsource
Print Email Current Profile Archive Blog Generate newsletter

Other articles in the HPC section of Volume 143, Issue 4: Sun HPC Solution Ushers in Teraflop-scale Applications for Brazilian Computing Grid The Virtues of Virtualization (this article) See all archived articles in the HPC section.

Top 10 Most Popular Articles in Current Issue (Vol 167, Issue 4) Simplified Administration with ZFS Documentation for Common Sysadmin Tasks Solaris 10 OS and ZFS Cut Costs, Boost Uptime as Alternative to RAID Oracle Certified Associate (OCA), Oracle Solaris 11 System Administrator MySQL: An Introduction for Oracle DBAs illumos Meetup at FOSDEM Linux Adoption Trends 2012: A Survey of Enterprise End Users MySQL 5.0 and 5.1 available on Solaris 11 World Record x86 TPC-C Result Java Spotlight Episode 65: Deployment & Plugin Recent Blog Entries as of February 4, 2012, 9:06 am Oracle Enterprise Manager Enables Cerner Health Care to Cut Capital Expenditures by $9.5 Million Beefing Up Xeon Reliability, Availability, and Serviceability: Slide Set Java Spotlight Episode 54: Stuart Marks on the Coinification of JDK7 Getting PostgreSQL HA to Work in vFabric Data Director November 9th, Oracle Solaris 11 Launch, New York City The [System] Developer's Edge: 322 page pdf Oracle Global IT Upgrades to Sun Storage Products UNIX - Dead or alive? LibreOffice Adopts ODF 1.2 as an OASIS Standard The Growth Prospects for Tape vs. HDD and NAND as a Storage Devices Blog RSS Feed