insideHPC.com

Oak Ridge National Lab’s Jaguar supercomputer just completed its latest upgrade.  The Cray XT3/XT4 system was previously chewing through 119 Tflops.  With this upgrade, the new peak performance rate has jumped to 263 Tflops.  If you’ve been following our coverage on the Top500 rumor mill, we failed to quote this system for the list of newly procured/upgraded systems.

Where does this new upgrade put Jaguar on the list?  Well, currently the system runs at 119 Tflops peak and roughly 101 Tflops max [as quoted by the Nov07 Top500 list].  If one does the math, you’re looking at a system with ~85% efficiency on linpack.  Now, at 263 Tflops peak, the max should come in right around 223.55 if they maintain the same efficiency [which is often very difficult].  If the Top500 list were not to change at all this June, this would place Jaguar in the number two slot.

Regardless of how the list will change this summer, Jaguar will most certainly be competitive within the Top10.

Share This

If you’re tracking or using Grid Engine, you’ll be interested in knowing that binaries for the latest beta are available. From news at Sun’s HPC Watercooler

Grid Engine 6.2, which has undergone significant changes in qmaster to significantly improve its scalability in challenging environments, adds powerful features to the core system, introduces multi cluster support for the Accounting and Reporting Console (ARCo) and comes with a new module extending the scope of Grid Engine to a new domain of use cases: the Service Domain Manager (SDM), aka. project Hedeby allows to dynamically (re-)assign computational resources on demand.

You can download the binaries here.

Share This

On Monday HPCwire carried news from the MathWorks that they have extended the integration of PCT:

The MathWorks … announced the integration of its Parallel Computing Toolbox with its widely used MATLAB optimization toolboxes to help further simplify the development of parallel applications. Parallel computing capabilities are now integrated inside the optimization solvers of MathWorks Optimization Toolbox and Genetic Algorithm and Direct Search Toolbox, enabling users to solve computationally-intensive optimization problems on multicore computers and computer clusters without significantly changing their existing programs.

Also new in this release is support for job scheduling apps PBS Pro and TORQUE.

Share This

HPCwire carried news earlier this week that the DOE is accepting proposals for HPC hours to be awarded in 2009 as part of the INCITE program

The U.S. Department of Energy (DOE) announced today it is accepting proposals for a program to support high-impact scientific advances through the use of some of the world’s most powerful supercomputers at four of DOE’s national laboratories. Through the Innovative and Novel Computational Impact on Theory and Experiment (INCITE) program, DOE’s Office of Science plans to award approximately 680 million supercomputer processor-hours at its laboratories in Berkeley, Calif.; Chicago, Ill.; Oak Ridge, Tenn.; and Richland, Wash. for large-scale, computationally-intensive science projects in 2009.

Share This

Earlier this week Sun announced that it was plugging AMD’s beleaguered quadcore Opterons into its gear. It announced:

…the availability of its first Sun Fire and Sun Blade systems powered by Quad-Core AMD Opteron processors, bringing new capabilities, increased performance and expanded scalability to customers that purchase or upgrade to these quad-core systems. The Sun Fire X4140, Sun Fire X4240 and Sun Fire X4440 servers, the newest systems to join Sun’s extensive x64 (x86, 64-bit) server line, give customers industry-leading energy efficiency, density and scalability powered by Quad-Core AMD Opteron processors and a choice of operating systems, including the Solaris 10 Operating System (OS), OpenSolaris operating system, Linux, Windows and VMware.

The Sun Fire X2200 M2, Sun Fire X4100 M2, Sun Fire X4140, Sun Fire X4200 M2, Sun Fire X4240 and Sun Fire X4440 servers powered by Quad-Core AMD Opteron processors are available now. The Sun Fire X4600 M2 and Sun Blade X8440 servers powered by Quad-Core AMD Opteron processors are expected to be available by the end of the quarter

Share This

IBM has revved its Cell-based blade again, following an update to the QS21 last August (and which the company will continue to sell).

The QS22 uses the PowerXCell 8i variant of the Cell processor, which has mods for double precision math and up to 16 times the memory of the previous version.

From Ashlee Vance’s coverage at The Register

The QS22 ships with a pair of Cell chips and up to 8 DIMM slots, depending on the model. Customers will also find a pair of Gigabit Ethernet ports and options for InfiniBand connections. The system has no internal hard drives, although an 8GB flash drive will be available in the second half of the year.

Additional coverage at EE Times provides more details on the specific changes

The new chip uses 65nm process technology to reduce the power consumption of the previous 90nm chip while maintaining the same 3.2 GHz frequency. That allows IBM to get two of the chips on to a single board while keeping board-level power consumption under 250 W required by IBM’s BladeCenter servers.

The claim is that this chip is quite a bit faster on business workloads that might make the Cell platform more attractive to a bigger (busines) audience. From coverage at Australian Computerworld

“The new Cell chip is up to 20 times faster on some common financial calculations than Intel’s quad-cores,” said Dan Olds, an analyst with the Gabriel Consulting Group. “The real advantage is being able to run a lot of workloads a lot faster than you thought possible. It could potentially be a game changer for some companies.”

The new blade will start shipping in June.

Share This

Douglas Eadline has an informative article that reviews some of the basic tenets of parallel programming, including the lesson that concurrency is a property of an algorithm, but having it is not the same as realizing a parallel implementation.

If the independent parts are run on separate cores/nodes then we will say the program is executed in parallel. And, just because your program has some concurrency, it does not mean that it will run efficiently in parallel. The distinction is subtle, but very important when real hardware is used. Since the goal is to make your program run faster, we need to ask the question, Does making all the concurrent parts of the program parallel increase execution speed? The answer to that question is a definite “maybe.” Why? Because in some cases running concurrent parts of your program in parallel may actually slow down your the overall application!

It’s good to review the basics every now and then. Check it out.

Share This

From Department WTF: last week the Stanford faculty Senate concluded that the university’s AHPCRC contract, awarded to a consortium led by Stanford last year, is “ok.” They guess. But don’t try it again.

Professor Steve Monismith, the chair of the committee, told the Senate on May 1 that the complex agreement negotiated with the Army addresses the openness concerns raised by 60 faculty members in a letter they signed in the spring quarter of 2007 questioning the appropriateness of the contract.

Specifically, Monismith said, no classified research will be conducted on the Stanford campus or by Stanford faculty; Stanford researchers will not be required to obtain security clearances; and there will be no restrictions on academic publishing of research results.

Ah, openness. But not everyone is happy…what about ethics?

The research committee did not address the ethical questions of military research raised by the faculty letter. “We had this debate last year in the context of tobacco funding,” Monismith said. “We hope that we all remember that each of us is free, subject to our own ethics and conscience, to seek support from anyone that we choose that’s legal, and to form our own findings and conclusions.”

Or “recruiting” for the researchers who might work — not in uniform — but on research in the military-industrial complex? (Shudder)

Bernard Roth, the mechanical engineering professor who led the opposition to the Army contract, said he was still not satisfied with the details of the contract. He said he worried, for example, that the center’s summer programs for high-school science students would serve as a recruitment tool, “so they would get into the pipeline to work at Army research labs.”

Defense research labs have created some of the fundamental technology we use every single day. Right, let’s not encourage that.

So they let it pass this time, but not without a passing slap

If the Army were to attempt to classify university research, Stanford would find the move unacceptable, the research committee reported, “and would likely make it necessary for Stanford to terminate the agreement, despite financial and other hardships to researchers that could result.”

Here’s a thought: this we have a (mostly) free market and a (mostly) free society. If you don’t like the terms, don’t do the business.

Share This

UnivaUD, the newly formed conglomerate aimed at developing software products for the complete distributed computing lifecycle, has announced that it has added two industry veterans to its staff.  Gary Tyreman and Bill Bryce have both joined UnivaUD from Platform Computing.  Tyreman was named Vice President and General Manager of HPC.  Bryce was named Directory of HPC product management.

For more info on their move, read the full post here.

Share This

SC08 will once again bring the Cluster Challenge to university teams from around the world. Teams of undergraduates will compete on the exhibit floor, using clusters of their own design, to run a workload of real-world problems.

The contest is simple, but the idea is big: users of all skills, and organizations of all sizes, can now own tools that just a few years ago were limited to just a few of the world’s wealthiest R&D organizations. The Challenge aims to demonstrate just how accessible all this horsepower really is.

The Cluster Challenge, now in its second year as part of the technical program at the annual Supercomputing conference, is billed as an event to showcase the computational power that even inexpensive clusters have and, according to the website, the ability of potential users around the world “to harness open source software to solve interesting and important problems.” The competition pits teams of undergraduates against one another to see who can build and configure a cluster that accomplishes the most work using real computational codes in the least amount of time.

So what kind of clusters are we talking about here? ; In order to level the playing field, there are tight physical constraints on the systems. The cluster compute hardware (processors, switch, storage, etc.) must fit into a single 42U rack powered by two 120-volt, 13-amp circuits. According to Gorda, even with this physical limitation the systems built last year were substantial, “At the event one team achieved an amazing 420 GFlops of performance on HPL (Linpack). That score would have put them in the Top500 in 2003, only four years earlier.” For even more perspective, a system like this would have been number 1 in 1995 and still in the top 20 as late as 1998.

The winning team in last year’s Challenge was out of the University of Alberta, supported by vendor partner SGI. This team included 5 U of Alberta undergrads, 1 high school student, faculty coach Paul Lu, and variety of other coaches and contributors.

What was the key to Alberta’s victory? Paul Lu points to the deep commitment of team, “As a coach, I learned that you cannot really motivate a team that does not already want to win. The student team members were motivated and talented.” Lu also credits their vendor partner, “…SGI was great with engineering support (when needed) and, for example, in providing some last minute, expensive memory upgrades when we determined it could make a big difference with the GAMESS application.”

If you are interested in taking part in this year’s Cluster Challenge, you still have time to get in on this incredible learning experience. You can get more information about the various Challenges sponsored at SC08 - including very detailed information about the Cluster Challenge - at the SC08 challenges web site. The deadline for entry into this year’s competition is July 31. And if you are a company interested in sponsoring a team, or you are part of a team in search of a vendor, get in touch with the Challenge chairs by sending an email to cluster-challenge@info.supercomputing.org.

Share This

ComputerWorld has posted an interesting peek into what could become the fastest supercomputer on the planet as measured by Linpack this June. Rumors have been circulating that RoadRunner was going to take a stab at the top spot. IBM’s fabled RoadRunner is in the process of receiving Phase3 of an upgrade to push its peak performance over a petaflop. 6,000 square feet of floorspace, 250 tons of equipment and lots of IBM logos have gone into building this behemoth. How sure is IBM that they will be able to do so?

We will break the petascale,” [Don] Grice [Chief Engineer of RoadRunner] told Computerworld.

According to sources inside IBM, they are quite simply contractually obligated to do so. This, of course, alludes to the fact that RoadRunner will be over 2x the current speed of its cousin, BlueGene/L at Livermore [currently 478 Tflops].

Computerworld has also posted a series of pictures of RoadRunner taken from within IBM’s lab in Poughkeepsie, New York. Have a look.

No coyotes were harmed in the making of this supercomputer.

Share This

Clustercorp has announced they have officially released their latest set of “rolls” for Rocks version 5 [x86_64]. The following were released today:

.: Intel Developer Roll: Intel C, C++, Fortran compilers; runtime components and developers tools [10.1]

.: PGI Roll: Portland Group’s C,C++, Fortran compilers [v.7.1-6]

.: Cisco-OFED Roll: installs/configures OFED 1.3 Infiniband stack and MPI as provided by the Cisco-OFED 1.3 source and RPMs

.: Absoft Roll: Absoft Pro Fortran v10.1 Compiler Suite and associated libraries

Head over to the Clustercorp website and download the rolls here.

For those looking for commercially supported Rocks, Clustercorp also announced the latest release of their Rocks+ rolls for Rocks version 5. For more info on Rocks+, read the full release here.

Please note that valid licenses are required for several of the aforementioned packages. Licenses can be obtained from the respective developer.

Share This

UPDATED: The University of Central Florida has landed its first supercomputer! Thanks to a $2.6million grant from the Army, UCF has purchased a new super from IBM. The new cluster tips the scales with 224 processors, 512GB of memory and 22.2TB of storage. There is also an upgrade planned for this summer to triple the compute and memory capacity.

This is a great opportunity for UCF and the simulation industry,” said Michael R. Macedonia, general manager of the Orlando unit of Forterra Systems Inc., a high-tech partner with the college. “People need to understand how important it is to have a supercomputer of this class in Central Florida. This will allow UCF to press the limits of science and attract new business to the region as well.”

The initial compute workloads will include defense simulation training [hence the Army dollars attached]. Future workloads will include medical simulation, civil engineering and nanoscience.

For more info on the new super, read the full article here.

Share This

AvarSYS has announced that is has made several key hires to in order to support its expanding high performance and technical computing business and solutions. These new hires will prove to be key in reaching into markets such as Petroleum, Scientific and Engineering, Pharmaceutical, Finance and Federal sectors. The list of n00bs includes:

.: Scott Hunter: formerly the HPC Professional Services Manager for LNXI

.: Doug Harless: formerly specializing in Oil/Gas and Federal sales for LNXI

.: Ron Blessing: formerly specializing in commercial sales for LNXI

.: Nate Furhiman: formerly a systems engineer with LNXI

.: Philip Carroll: formerly a Federal sales manager at Arrow ECS [IBM distributor]

.: H. West Richards: formerly a senior manager at several business consulting firms

Congrats to those with a new gig. For more info, read their full post here.

Share This

The University of Florida’s High Performance Computing Center has chosen RAID Inc for a new storage system totalling over 100TB.  UF has purchased six RAID Inc Falcon III [24bay, 4Gb FC to SATA-2] storage subsystems.

The High-Performance Computing Center at the University of Florida seeks best of breed providers that can push the technology envelope in support of our research projects, ” said Charlie Taylor, associate director of the University of Florida’s HPC Center. “RAID Inc. was the ideal partner for our challenging storage needs, and their Falcon III product was a key component in the success of this project.”

The center will utilize Lustre as the file system for the newly procured RAID Inc devices.  They intend on serving up scientific data to local HPC users as well as researchers at five satellite campuses.  Researchers at other Florida universities will eventually get access to the storage via the Florida LambdaRail.

For more info, read the full post here.

Share This