Premium Content

Access "Now, That's a Cluster!"

Published: 29 Oct 2012

Lawrence Livermore National Labs (LLNL) is pushing the envelope with a new storage cluster that mates 115TB of networked disk with a massive cluster of 600 dual Pentium 4 servers. Mark Seagar, assistant department head for terascale systems, says the project tackles several advanced storage issues. Key is Lustre, a Linux clustering technology under development by LLNL; Los Alamos and Sandia National Labs; and Cluster File Systems, Inc., Mountain View, CA. Using a metadata server, Lustre stripes data in the same object format that the simulations generate. The storage servers, 64 BlueArc Si7500s, handle block level translation, aided by firmware enhancements. The San Jose, CA-based BlueArc uses hardware-assisted protocol processing, which enables writes of 5GB/s over Gigabit Ethernet to the storage cluster. Ultimately, a second cluster for visualization will address the same data. Lustre provides data sharing (without data duplication) by both sets of applications and servers, difficult with today's technology. LLNL evaluated NFS technology, but Segar said ... Access >>>

Access TechTarget
Premium Content for Free.

By submitting you agree to receive email from TechTarget and its partners. If you reside outside of the United States, you consent to having your personal data transferred to and processed in the United States. Privacy

What's Inside


More Premium Content Accessible For Free

  • Big data storage challenges: Speed, capacity and HDFS integration

    Big data infrastructure and analytics are some of the hottest technology topics today, and it can sometimes seem impossible to dissect and digest all...

  • Moving target: Endpoint backup

    Mobile workers are now accessing, creating and modifying data on ultra-portable devices such as smartphones, tablets and phablets. Most companies ...

  • A lesson in flash caching

    Solid-state storage is proliferating as a replacement for hard disk drives, where it offers a quick shift into the fast lane of storage processing. ...