Nominations are now being accepted for the SVC Awards, which acknowledge excellence in Storage, Virtualization, and Cloud Computing. Judged by a panel of HPC industry editors and analysts, these awards are a great way to recognize your customers.
Nominations for the 2012 awards program are due by Oct. 17, 2012.
Can the cloud really work for HPC applications? For many HPC users thinking of moving their workload to the cloud, the idea of having a user-friendly front end free from back end administration is a strong incentive. At the upcoming ISC Cloud’12 conference, a panel of experts will examine the hurdles of moving HPC to the cloud.
This panel will examine the potential obstacles to cloud adoption, such as security and trust, compliance, outsourcing, performance, virtualization, pricing, payment model, software licenses, choice of service provider or private cloud builder, network bandwidth, and integration of all of these into existing business or research processes.
The event takes place Sept. 24-25 in Mannheim, Germany. Attendees are encouraged to register by Sept. 15 to enjoy the preshow discount.
In this video, Chad Harrington from Adaptive Computing describes the coming wave of virtualization technologies in high performance computing and how the company’s Moab software helps customers manage HPC clusters of all sizes.
Recorded at VMworld 2012 in San Francisco.
insideHPC provided onsite coverage of Hot Interconnects 2012 in Santa Clara.
In this video from SIGGRAPH 2012, Nvidia’s Ian Williams presents on the VGX Hypervisor, the company’s move to bring graphics to the Cloud.
The new NVIDIA VGX technology allows for true hardware virtualization of the GPU, enabling a true PC and Workstation experience in a virtual desktop environment. This session will cover a comparison of graphics virtualization technologies available in the industry (both SW and HW methods) as well as accelerated remoting solutions.
We had a terrific time at ISC’12 filming dozens of videos with key HPC vendors. In case you missed them, here are the featured programs:
In this video, Josh Simons from VMware discusses how virtualization technologies are making inroads into traditional HPC. Simons has been working on the project for two years as a representative for VMware’s CTO office.
Recorded at ISC’12 in Hamburg. You can follow progress on this effort at the Josh Simons blog.
I’ve been shooting interviews non-stop with HPC users and vendors here in Hamburg. In the meantime, Day 1 of ISC’12 brought a virtual blizzard of news to my inbox . What to do? Here’s the HPC News with Snark for June 19, 2011.
By Timothy Prickett Morgan • Get more from this author
Chip giant Intel has been hogging all of the headlines lately in the server racket lately, and Advanced Micro Devices this morning is trying to get a word in edge-wise with some clock speed bumps on its “Bulldozer” family of processors for server with one, two, or four sockets.
Don’t get too excited, though. This is not a rev on the Bulldozer core or anything like that, but rather just the normal deep bin sorting that AMD always does a few months after the initial release of a chip as the wafer baking process matures and, generally, as a reaction to whatever Intel is doing at the time.
The Opteron 4200 (for servers with one or two sockets) and 6200 (for servers with two or four sockets) launched back in November, ahead of Intel’s “Sandy Bridge” Xeon E5 assault in March and April and got the jump on Intel at the low-end with its Opteron 3200s for single-socket microservers in March, ahead of the Intel “Ivy Bridge” Xeon E3-1200 v2.
AMD, under new CEO Rory Read and a whole new management team, is not trying to get the jump on Intel so much as get partner GlobalFoundries to predictably ramp its 32 nanometer processes and get on track to the next process nodes in the roadmap, so AMD can better compete against the current and future Xeons and forthcoming “Centerton” Atom processorsaimed at servers.
There are two new “Interlagos” Opteron 6200 processors and three new “Valencia” Opteron 4200 processors, and all of them are just deep bin sorts on the chips coming out of GlobalFoundries, Michael Detwiler, product marketing manager at AMD, tells El Reg.
The entry Opteron 3200s are not being refreshed at this time, and Detwiler was not at liberty to divulge if there are going to be other speed bumps in the current Bulldozer family of Opterons before the next-generation “Piledriver” cores and their related Opterons come to market, perhaps starting later this year. If recent history is any guide, this is the last hurrah for the Bulldozers and AMD is focused on getting its new cores etched and baked onto the future Opteron 3300 (“Delhi”), 4300 (“Seoul”), and 6300 (“Abu Dhabi”) processors for the existing C32 and G34 processor sockets and SR5600 chipsets.
Here is the current Opteron 6200 lineup and how it stacks up against the prior Opteron 6100s:
Updated AMD Opteron 6200s; new chips in bold italics
As is usually the case, it is debatable if the extra oomph in the chips that come out of the deep-bin sorts is worth the money that AMD is charging. (The same holds true for Intel’s Xeons.) But in this case, if you are a high freaky trading company or HPC center than needs an extra couple hundred megahertz and you have money to blow, these chips are possibly for you.
The Opteron 6284 SE (short for Special Edition and burning 140 watts of juice) runs at 2.7GHz, a 3.8 per cent jump in clock speed. But that extra 100MHz costs $1,265, which is 24.1 per cent more than a 2.6GHz Opteron 6282 SE running at 2.6GHz and costing $1,019. You really need those clocks on those 16 cores to pay that premium.
That said, you get 16 threads running at 2.7GHz, and that is still faster than the top-bin Xeon E5-2470 part from Intel for two-socket boxes, which runs at 2.3GHz with eight cores and 16 threads with HyperThreading turned on for $1,440. The faster E5-2670 processor from Intel runs at 2.7GHz with eight cores, in a 130 watt power envelope, and costs $1,723; the E5-2690 burns at 135 watts and runs at 2.9GHz across its eight cores, but costs $2,057. If you want a four-socket machine, a Xeon E5-4650, which runs at 2.7GHz and is also rated at 130 watts, costs a staggering $3,616.
With those price differentials, you can see now why some HPC shops looking for fat server nodes or those using heavy virtualization on their x86 iron are going for the Opterons.
Still, that new 16-core Opteron 6728 that runs at 2.4GHz is only 100MHz faster than the Opteron 6276, which is a 4.3 per cent bump, but it costs 25.5 per cent more. You are going to need the cash and a pretty good reason to justify that money. Perhaps the new AMD management is wishing they had not cut prices quiet so low on the Bulldozer Opterons and is trying to make up for it a little?
AMD is adding three new chips in its Opteron 4200 lineup:
Updated AMD Opteron 4200s; new chips in bold italics
That Opteron 4240 chip, which is a standard 95 watt part, runs at 3.4GHz, which is a respectable 9.7 per cent faster than the next SKU down the line, but it costs 81.6 per cent more. The Opteron 4276 HE (which is short for Highly Efficient and meaning it is a low-voltage part) is 20.7 per cent more costly than its nearest sibling in the Bulldozer family, but only adds 4 per cent on the clock speed.
AMD’s announcement calls the Opteron 4230 a regular part, but with it only consuming 65 watts and running at 2.9GHz, is looks like an HE part to El Reg and that is why we put it in that part of the table.
The Opteron 4200s don’t have SE parts, and there have been no updates to the Extremely Efficient (EE) very low voltage Opterons. If you want super low power consumption, then the 1.6GHz eight-core Opteron 4256 EE is your only option for a two-socket server. Depending on the workload and the memory requirements, a single-socket Opteron 3200 processor, which is less expensive but which has fewer cores in the 45 watt thermal band, might be a better option than an Opteron 4256 EE. ®
This article originally appeared in The Register. It appears here in its entirety as part of a cross-publishing agreement.
In this slidecast, Shai Fultheim from ScaleMP provides an update on the company’s virtualization software solutions for HPC. The company is preparing a big announcement at ISC’12. You can hear some of the details here or check out their booth #303 in Hamburg.
The innovative Versatile SMP (vSMP) architecture aggregates multiple x86 systems into a single virtual x86 system, delivering an industry-standard, high-end symmetric multiprocessor (SMP) computer. Using software to replace expensive custom hardware and components, ScaleMP offers a new, revolutionary computing paradigm.
Download the MP3 * Download the slides (PDF) Subscribe on iTunes * If Dropbox is blocked, download from this Google page.
In this audio podcast, Patrick Geoffray and Torsten Hoefler from the Hot Interconnects Conference lay out thier final Call for Papers.
Conference themes include cross-cutting issues spanning computer systems, networking technologies, and communication protocols for high-performance interconnection networks. This conference is directed particularly at new and exciting technology and product innovations in these areas. Contributions should focus on real experimental systems, prototypes, or leading-edge products and their performance evaluation.
Submissions are due May 20 * Download the MP3 * Subscribe on iTunes * If Dropbox is blocked, download from this Google page.
By Dan Olds, Gabriel Consulting • Get more from this author
Multicore processors drive everything these days from the biggest HPC cluster to the lowliest tablet – even smartphones. While parallel programming has come quite a way, there are still many apps that aren’t well-behaved at all.
They’re the worst kind of guests – acting like they own the whole damned house while paying absolutely no attention to the needs of other residents.
They’ll grab more memory than they need and never let it go. They’ll spawn enough threads to crowd out everyone else; it’s like inviting their deadbeat friends over to watch the Super Bowl at your house and eat your snacks. Operating systems and virtualization mechanisms attempt to control unruly apps, but they don’t have the ability to completely control and prioritize system resources.
Enter exLudus, and what they’re calling the industry’s first micro-virtualization solution, intuitively named MCOpt. What we’re talking about is a suite of software packages that provide dynamic workload containers, workload characterization, and a performance monitoring/management for Linux operating system instances. It works at a node-level, as a layer between the Linux kernel and the applications running on top of it.
With MCOpt, users can dictate the priority of apps and jobs, and the system will automatically adjust core and memory shares to ensure that the priorities are satisfied. It’s not a static fair-share scheduler; it works dynamically to constantly adjust resource shares and job timing so that SLAs are met and the system achieves maximum throughput.
It can do this because it’s monitoring how each job is using cores and memory. It can spot when a job isn’t using all of its allocated memory or core shares (or if it’s trying to use too much of either) and make adjustments on the fly to keep everything running smoothly and according to business priorities.
In our discussion, the exLudus folks talked about the Linux scheduler and how it can unpredictably cause job priorities to change during execution – which isn’t necessarily a bad thing. But it can make it difficult to pinpoint when resource contention is hindering overall performance. It also means that subsequent re-runs of the same set of jobs will result in different contention behaviors.
MCOpt can also save important work from falling victim to the Linux Angel of Death – the Out-Of-Memory Killer. When system RAM is oversubscribed, there’s a risk that the OOM Killer can swoop in (well, it doesn’t really swoop) and kill processes to free up memory.
MCOpt helps in two ways. First, it keeps apps from oversubscribing memory and thus prevents the OOM Killer from coming into play in the first place. Second, it can steer OOM Killer behavior to protect high-priority workloads. With MCOpt, subsequent re-runs of the same set of tasks will behave exactly the same way – meaning predictable application performance even under stress. (System stress, not personal stress.)
This high level of control can really help overall throughput. The company says that their tests show a 20 to 50 per cent increase in total throughput with MCOpt versus a stock Linux. This is a measure that includes the low overhead load of MCOpt, of course. They have some white papers and stuff here, plus free trial versions of their software too.
exLudus also made sure to point out that MCOpt doesn’t require any application or OS modifications – the MCOpt layer is transparent to both. Better yet, MCOpt can work with other cluster management and virtualization suites if needed – sort of as a subcontractor.
exLudus brings an interesting set of capabilities to the Linux workload management table. It’s like taking a previously unmanageable city traffic plan (Boston? The Bay Area?) and adding synchronized lights and a set of maniacally focused traffic managers. It’s definitely worth a look if you’re seeing signs of road rage between competing apps on your Linux systems. ®
This article originally appeared in The Register. It appears here in its entirety as part of a cross-publishing agreement.
In this video, Greg Papadopoulos from New Enterprise Associates presents: How to Design and Build Your Very Own Exascale Computer
Abstract:
A computer of any real size today is built around from thousands of individual servers, storage arrays and network switches. Mostly, the construction of these systems are left as an Exercise for the User, but that’s changing rapidly. Patterns around compute-storage-network virtualization are emerging, and are apt to coalesce, finally, into some coherent view of a interconnect-centered system. Optics will play both a defining and enabling role in this “re-integration,” and by 2020 it’s likely that any competitive large-scale system will crucially depend upon optical interconnects all the way to the processing chips themselves. In this talk, Greg Papadopoulos will look at the forces that have shaped the way we build very large systems today, and speculate about the future history exascale computers and the industry that creates them.
Recorded at the OFC/NFOEC 2012 Plenary Session in Los Angeles on March 8, 2012. Download the slides (PDF).
Today eXludus announced the first Micro-Virtualization solution that extends virtualization to the “micro” level so that users can also optimize the real-time allocation of micro-resources, such as cores and memory, in multi-core systems. The ability to manage resources at this level becomes especially important as the number of cores per system continues to grow.
The advent of multicore systems created a resource optimization gap that is very similar to what existed in multi-processor servers before server virtualization was invented,” said Dan Olds, principal at Gabriel Consulting. “The eXludus micro-virtualization software provides a similar capability at a micro level.”
Read the Full Story.
Today the SUNY Fredonia Technology Incubator announced that the Silicon Wolves Computing Society (SWCS) will be joining their incubator program. SWCS is a consumer-friendly high-performance computing system developer and manufacturer of the most advanced workstations and computer gaming solutions on the market. The high-tech start-up company recently relocated to the Incubator from Anaheim, Calif.
We are thrilled to be in Dunkirk and affiliated with SUNY Fredonia through the Technology Incubator,” said Ryan Wolf, President of SWCS LLC. “This is a great community, and a renowned university and we are excited about our opportunities for growth.”
SWCS is a computing solutions manufacturer and integrator that conduct research and development in highly specialized computer workstations, desktops, laptops and servers, and in particular, develop and configure dedicated computing solutions for Cloud, Virtualization, HPC and Reconfigurable Computing environments.
The Incubator is all about attracting companies which are poised for substantial growth, and helping them through the process while providing them access to the resources that a nearly 6,000-student institution like SUNY Fredonia can provide,” said Robert Fritzinger, Director of the Technology Incubator. “I see a strong fit between Silicon Wolves Computing Society, the SUNY Fredonia campus, and the Incubator and I am anticipating a successful partnership.”
Read the Full Story.
Advertisement
insideHPC.com is a production of insideHPC, LLC. © 2006-2013 Sitemap
