August 2015 OpenSFS NEWSLETTER

OpenSFS_EOFS_theme-banner_final ISC Showcases Lustre Worldwide

The International Supercomputer Conference 2015 recently finished up in Frankfurt, Germany. It’s the largest annual supercomputing show outside of North America, and every year it does a great job of showcasing advances by European companies and products in the field of supercomputing. This year, the topics ranged widely under the stated rubric of diversity for the event, including the Top3 Exascale Projects, HPC in the Global Automotive Industry, Future HPC System Design Concepts, HPC and Cloud Computing in Financial Services, and even the Potential Impact of Quantum Computing On HPC.

OpenSFS participants were in full force with a spate of Lustre-focused announcements:

  • DDN’s product announcement (“ DDN Rolls Out World’s Fastest Entry-Level Lustre Appliance” – included new customers CYFRONET and The Wroclaw Centre for Network and Supercomputing in Poland, Public Health England (PHE); 
  • Intel previewed a number of new features coming in its Enterprise Edition for Lustre (IEEL) 2.3, including support for Multiple Metadata Targets in the Intel Manager for Lustre GUI. (“ Intel Previews Enhanced Lustre File System for HPC“). Intel is also reportedly preparing to roll out new security, disaster recovery, and enhanced support features in Intel Cloud Edition for Lustre 1.2, later this year. 
  • At the ISC15 Vendor Showdown, Charlie Wuischpard, GM of Intel’s HPC Group Presented Intel’s Scalable System Framework that included four components: 1) compute; 2) memory/storage; 3) fabric and 4) software including Lustre. ( Here’s a roundup of twitter feeds from the event) In the Intel Booth, Mark Seager, CTO for the HPC Ecosystem at Intel, presented on CORAL and Intel’s HPC Scalable System Framework: the Path to Exascale. Dr. Seager presented more details on the CORAL Aurora and Theta systems that both include Lustre as the parallel file system. Argonne National Laboratory and DOE announced selection of Intel and Cray to deliver Aurora back in February 2015. Argonne’s previous system was Mira using BlueGene/P and GPFS technologies.  Having Argonne transition from GPFS on Mira to Lustre on Theta and Aurora systems is a major design win for the Lustre community.
  • At the HPCAST24 meeting prior to ISC15, Bill Mannel of HP and Charlie Wuischpard of Intel announced an HPC Alliance that has HP leveraging the Intel Scalable System Framework for future generation of HPC products in general and collaboration on Lustre future storage products in particular.
  • Bright Computing Announced Integration with Intel Enterprise Edition for Lustre with an integrated combination of GUI management for Lustre, Intel’s IEEL and Bright Cluster Manager. Intel previewed a number of new features coming in its Enterprise Edition for Lustre 2.3, including support for Multiple Metadata Targets in the Intel Manager for Lustre GUI. 

And finally at ISC, the annual OpenSFS-sponsored Lustre Birds of a Feather (BoF) session was titled “EOFS & OpenSFS: Challenges & Opportunities for Lustre File System within HPC & Big Data” and focused on gathering the worldwide community of Lustre developers, administrators, and solution providers to discuss new challenges and corresponding opportunities emerging within HPC and Big Data. Though the format is open and free-form, a succession of speakers helped to push the agenda: Frank Baetke (HP),  Hugo Falter (ParTec),  Mark Seager (Intel).

ISC15Mark Seager and Hugo Falter celebrating OpenSFS+EOFS collaborations.

In Other OpenSFS News:

New OpenSFS offer! Get Involved with Lustre for FREE. OpenSFS is providing a limited time offer to subsidize your first year membership–up to $5,000! You can apply the money to any membership level. Each membership level gives your technical and business teams the ability to ask questions directly to Lustre experts and test out ideas. To sign up immediately: https://opensfs.org/get-involved-with-lustre-for-free/

OpenSFS Board meetings, open to community participation

OpenSFS prides itself on being deeply committed to the philosophy of open source. We bring together a wide range of organizations and individuals to solve problems that move open source file systems forward. The weekly OpenSFS Board meetings are open to all OpenSFS participants.  https://opensfs.org/about/board-meetings-and-activities/

  • Meeting day: Wednesday
  • Meeting frequency: weekly
  • Time: 12-1pm Pacific
  • Dial in:  877-709-0823
  • Passcode: 4840841#

LUG 2015 Presentations and videos

Now available online, wide selection of Lustre technical topics:  https://opensfs.org/lug-2015-presos/

OpenSFS Technology News:

Lustre 2.8 – Coming Soon!

OpenSFS works closely with the Lustre community to agree upon the content for the community Lustre roadmap. Lustre 2.8, slated to be released soon, includes three new important features for the community: 

  • RHEL/CentOS 7.1 Server Support
    Lustre servers can now run on RHEL/CentOS 7 systems. This is a big improvement with security and maintenance for Lustre servers. RHEL 6 is based on a 10-year old kernel, RHEL 7 on a very recent one.
  • Distributed Namespace (DNE) Phase 2
    The “full version” of DNE II, with striped directories. Striped directories can provide massively improved single directory metadata performance. Includes asynchronous updates, which improves performance and reliability.
  • Multiple modify Remote Procedure Calls (RPCs)
    Allows multiple modifying metadata operations at once (create, delete, update) from a single client. This radically improves the metadata performance possible from a single client, up to a factor of four times.
For technical comments or for mechanisms to join the Lustre community, please see the lustre.org site, central to the Lustre technical community and supported by OpenSFS and EOFS.