Continue in 2 seconds

BlueArc NAS Storage Systems Power High Performance Computing Systems at Penn State University

Published
  • July 01 2003, 1:00am EDT
More in

REVIEWER: Vijay K. Agarwala, director, Graduate Education and Research Services group, a part of Information Technology Services at Penn State.

BACKGROUND: The Pennsylvania State University is one of the top research universities the United States. The University's Graduate Education and Research Services (GEaRS) group, a unit of Information Technology Services (ITS), strives to meet the research computing needs of approximately 300 researchers in a variety of academic departments and disciplines across the University. Penn State, one of the forerunners in high-performance computing at university-based centers, persistently explores new advancements in scientific computing technology by forging collaborative partnerships with various companies and institutions.

PLATFORMS: Currently, a single BlueArc Si7500 SiliconServer with 2 terabytes of aggregate disk has been deployed to serve two high-performance computing clusters, with a total of more than 300 machines. Lion-XL, the group's largest Linux cluster, consists of 176 Dell PowerEdge 2650 servers, each configured with dual Intel Xeon processors, 4 gigabytes of memory and an Ultra3 SCSI drive. The nodes are connected with Quadrics QsNet high-performance network. The older Lion-XE cluster has 128 nodes with dual Intel P3 processors and is connected by Dolphin's high-speed low-latency interconnect technology and SCALI's tuned MPI middleware. Both of these clusters use RedHat's distribution of the Linux operating system.

PROBLEM SOLVED: Prior to the induction of BlueArc SiliconServers, the GEaRS group ran separate file servers for each of the individual clusters, with each server having its own set of multiple disk enclosures. Using this configuration, we experienced situations where these NFS servers couldn't keep up with the demand placed by several hundred jobs running across our clusters, resulting in outages. Now, a single BlueArc SiliconServer serves both of the large clusters ­ with a total of more than 600 CPUs ­ and delivers better I/O performance than our previous implementation, as well as continuous uptime.

PRODUCT FUNCTIONALITY: The BlueArc SiliconServer has proven to be extremely reliable, and the GEaRS group has not had a single problem with it after our own acceptance testing and deployment. Additionally, as previously stated, the single BlueArc SiliconServer out- performs the two previous NFS servers combined. We see an average of 50 to 60 megabytes/second streaming in and out of BlueArc's Si7500 unit on a constant basis, with peak I/O reaching 120 megabytes/second.

STRENGTHS: In terms of the needs of the GEaRS group, the key strength BlueArc provides is far better performance and greater scalability at an affordable cost. As a result, we are able to deploy a much larger cluster without the need to deploy multiple NFS servers.

WEAKNESSES: Initially, we had an issue where certain workloads would cause the BlueArc SiliconServer to unfairly favor certain I/O processes over others, such as sequential reads and writes. However, BlueArc technical support was very proactive in solving this issue, and we haven't had a problem with the SiliconServer since we brought it up.

SELECTION CRITERIA: When evaluating storage and file servers, our selection process is driven primarily by scalability, reliability, cost and acceptance of the company's products by our peers in the high-performance computing world. We are interested not only in working with established vendors and their technology, but also entering into collaborative partnerships with emerging companies on the brink of changing the way we view and implement high-performance computing technology. The fact that BlueArc technology was beginning to show up in national labs such as Lawrence Livermore and large pharmaceutical companies was important to us in the sense that people with similar needs in HPC were also considering or deploying it. However, our own testing of the BlueArc SiliconServer, which showed the system's performance, scalability, reliability and cost benefits, as well as the ease with which it fit into our current environment, helped make our decision to go with this solution.

DELIVERABLES: A large number of research projects are being carried out on the two big computational clusters the GEaRS group runs, the results of which have appeared in the leading journals of their respective disciplines.

VENDOR SUPPORT: BlueArc's business style is nothing short of motivated, cooperative and flexible. Throughout the process, it has been clear that they value our partnership and input. Though our BlueArc account team was based in New York, the company's engineers worked flexibly with us from New York, California and the United Kingdom. As such, BlueArc is definitely a technology company that we want to continue to work with in the future.

DOCUMENTATION:BlueArc's documentation, as well as the technical support, is excellent.

Register or login for access to this item and much more

All Information Management content is archived after seven days.

Community members receive:
  • All recent and archived articles
  • Conference offers and updates
  • A full menu of enewsletter options
  • Web seminars, white papers, ebooks

Don't have an account? Register for Free Unlimited Access