Researchers at the San Diego Supercomputer Center (SDSC) and their collaborators have released version 3.3.0 of the Rocks Clustering Distribution for immediate download (www.rocksclusters.org). Rocks supports all commodity processors: x86, Opteron, Itanium, and now EM64T Xeon.
Major new features of this release include: support for Intel's new 64-bit Xeon EM64T platform, inclusion of LAM/MPI as an alternative to MPICH, and two new Rolls for Visualization and Infiniband.
"The clustering community is quickly moving towards 64-bit CPUs with the introduction of AMD's Opteron just over a year ago and most recently with Intel's EM64T Xeons in the last few months. With strong support from key vendors, Rocks is able to rapidly add support for the most relevant commodity HPC platforms," said Philip Papadopoulos, Program Director for Grids and Clusters at SDSC. Early access to EM64T hardware, in the form of equipment donations, was provided by both Dell and Intel.
Like the evolution to 64-bit, clusters are moving beyond their HPC-only niche. Rollls are optional extensions to the Rocks base distribution to support different cluster types or provide enhanced functionality. Users simply insert a Roll CD at cluster build time to incorporate the desired extensions.
The Visualization Roll allows researchers to build their own tiled-display visualization clusters with the same ease as building a traditional Rocks computational cluster. Core software on this Roll includes: VTK, GTK, DMX, Chromium, nVidia libraries, NCSA's pixel blaster, and an open-source port of Doom. Development of the Visualization Roll leverages expertise from NCSA's Visualization group and OptIPuter partners at the Electronic Visualization Laboratory (EVL) of UIC and The National Center for Microscopy Imaging Research (NCMIR) at UCSD.
"Rocks plays a critical role for running our Sun Opteron/nVidia 5x4 tiled display.", said David Lee an Application Engineer for National Center for Microscopy and Imaging Research (NCMIR). "With Rocks, administration requirements are very low, requiring only the occasional maintenance from systems administrators.", adds Lee.
The Infiniband Roll supports IB hardware from InfiniCon Systems. InfiniCon, AMD, and SDSC partnered earlier this year to add support for IB into Rocks as an alternative cluster interconnect. With the inclusion of IB, Rocks now supports Myrinet, IB, and Gigabit Ethernet as a cluster interconnect.
While development of Rocks core distribution is lead at UCSD, the wide range of
supported capability is enabled by a world-wide group of dedicated collaborators. For example, The Korea Institute of Science and Technology Information (KISTI), have also released a native Korean version of Rocks 3.2.0 with native Hangul localization. Titled "KRocks" and available for download at their site (rocks.cluster.or.kr), KRocks includes translations of the user guides and interactive setup screens. An updated localized version of 3.3.0 will be released in the coming weeks. "We are pleased to be working with our PRAGMA partners at SDSC to bring their software to our Korean community, and we look forward to the inclusion of KISTI developed tools in future versions of Rocks", said Jysoo Lee, the Director of KISTI.
This release also includes refreshed packages for the Grid Roll (Globus and other services packaged by the NSF Middleware Initiative), the Grid Engine Roll (SGE version 5.3p6), the AMD Roll (32-bit compatibility libraries for Opteron and EM64T) and the Condor Roll (version 6.6.0).
The Rocks development community includes the Cluster Development Group at San Diego Supercomputer Center, Scalable Systems in Singapore, the HPC Group at University of Tromso in Norway, the SCE Group at Kasetsart University in Thailand, and the cluster development group at KISTI in Korea. Development of Rocks is funded from NSF, and aided by generous equipment donations from Sun Microsystems, Dell, AMD, Infinicon Systems, and Intel. The Rocks project was started at SDSC in early 2000, with the goal of "making clusters easy". Today, the Rocks user base includes five Top500 computers, and several hundred clusters around the globe. The Rocks Register, a web page where Rocks users voluntarily register their deployed cluster, shows that Rocks powers an aggregate of nearly 70 TFlops of peak computing. More information on Rocks, including documentation and complete access to the source code, can be found on the project's homepage (www.rocksclusters.org).