In this video, Ryan Waite, GM of HPC for Microsoft presents: Datacenter Efficiency - Optimizing Performance and Power. Recorded at the HPC User Forum in San Diego on Sept. 7, 2011.
There’s still time to register for the HPC User Forum series coming to Europe next week:
Registration is free of charge, including refreshments.
Power and cooling efficiency is an ever-increasing concern in HPC datacenters. Among the top-tier vendors, Dell in particular has been focusing on this area as a way to differentiate it’s product offerings. The company’s new PowerEdge Energy Smart Containment Rack Enclosures are a good example, as they provide a cost-effective stand-alone cold-air containment solution for raised-floor data centers.
Since a typical raised-floor implementation without containment distributes cold air into the whole room, it may be necessary to overprovision the airflow by 2-3x in order to provide enough air to the installed equipment. Even then, hot and cold air may mix, meaning warm air may be ingested by some systems or hot spots may form in some areas. Containment is a growing industry trend to address these issues but, unlike other containment systems (like aisle containment), the Energy Smart Rack can deploy one at a time, anywhere in the data center, without impacting facilities such as on-site fire-suppression systems. Just position it over a well-ventilated tile and install your IT equipment – the airflow from the raised floor is directed into the sealed front plenum for even distribution to all of the mounted systems. And, with the air delivery contained, the Energy Smart rack can support high-density and cloud-computing installations, up to 25kW of IT equipment at 25°C per rack.
Read the Full Story or download the whitepaper (PDF).
In this video, TACC’s Tommy Minyard presents: Power and Cooling at Texas Advanced Computing Center. The presentation was recorded at the HPC User Forum in San Diego on Sept. 8, 2011 as part of the Dell panel discussion on Power and Cooling. Download the slides (PDF).
The latest issue of The Exascale Report is out with interesting grumblings from the community about Intel’s recent announcement of exascale leadership. Personally, I came away from the press event (watch insideHPC’s exclusive video here) with a much stronger sense of clarity as to where the company wants to go by 2018, but others quoted in the article were highly critical:
Intel has yet to demonstrate a deep and credible plan to get to exascale. This [announcement] is about claiming ‘turf’ and is more of a strategic move for Intel – in an attempt to keep any of their major customers and partners from defecting to AMD or NVIDIA in search of longer-term exascale capabilities. Intel still has not been able to adequately address the interconnect issue – and that’s become a rather sensitive topic for them.”
There are also two really insightful audio interviews in this issue – one with IBM’s Don Grice (who is rumored to be the one taking over for Alan Gara as chief architect for Blue Gene), and one with Wilf Pinfold from Intel. Wilf is currently the Director of Extreme Scale Programs with Intel’s labs, but many of us know him from his role as the General Chair of SC09.
And also, for those of you who missed it, there’s a recap and pointer to the new exascale RFI issued by the E7 Consortium. Check it out at The Exascale Report.
Has the issue of optimized PUE largely been solved already for large datacenters? In this presentation, you’ll hear from the following individuals in a panel discussion moderated by James Hamilton of Amazon.
Recorded at the 2011 Data Centre Efficiency Summit held on May 24 in Zürich, Switzerland. A tip of the hat goes to Datacenter Knowledge for pointing us to this story.
Clipped from: www.datacenterknowledge.com (share this clip)Datacenter Knowledge takes a closer look at HP’s newly announced EcoPOD, an updated version of its modular Performance Optimized Datacenter that expands its capacity with a “double-wide” design that joins two 40-foot containers. The 900-square foot enclosure, which HP says can operate with a Power Usage Effectiveness (PUE) rating of 1.05 to 1.30, features a cooling system that can automatically adjust its operations based IT load and climate. Depending on environmental conditions and IT load, the cooling system automatically adjusts to use ambient air where possible, switching to direct expansion cooling mode when it is too warm to use outside air.
“The new modular offering features a server room on its lower level, with cooling equipment sitting atop the modular. Air can enter the unit through either side of the cooling unit, which also features Direct Expansion (DX) cooling. The system automatically adjusts to use ambient air where possible, switching to direct expansion cooling mode when it is too warm to use outside air.
Read the Full Story.
HP’s Marc Hamilton writes that the company has officially unveiled its long-rumored EcoPOD:
“While the EcoPOD serves a broad range of customer data center needs, I expect we will see a number of HPC cluster deployments in EcoPOD based on the early configurations HP’s HPC Competency Center has worked on this year. A single HP POD 240a can easily contain a complete PetaFlop cluster and with an industry leading PUE your PetaFlop cluster will cost substantially less to power than a similar cluster running in almost any existing data center.
It’s all part of what HP is calling Converged Infrastructure. Read the Full Story.
Today Platform Computing rolled out RTM 8 and Platform Analytics 8, which are designed to help administrators monitor, manage and analyze cluster usage. To learn more, I caught up with Louise Westoby, Senior Product Marketing Manager, Platform Computing.
insideHPC: Platform RTM 8 acts as a dashboard for the LSF product family. How does it change the way users currently interact with LSF? In other words, how did they get their work done before RTM came along?
Louise Westoby: Platform RTM is an operational dashboard that provides monitoring and reporting functionality for HPC administrators. It enables administrators to easily monitor and report on resource consumption as well as monitor resource allocation by user, group or project team. Before Platform RTM was available, administrators had to filter through log files and extract the usage data on their own. With Platform RTM, usage data is not only automatically extracted, but also it is done in real-time, providing information on the immediate condition of the cluster.
The ability to monitor cluster availability and performance is imperative when we’re running millions of design simulations to test our latest software releases,” said Steve MacQuiddy, IT Director Engineering Infrastructure, Cadence Design Systems. “Having the single Platform RTM dashboard allows us to simultaneously observe the entire cluster environment and it has not only made it easier for us to better balance our workloads, but it’s also helped us optimize throughput for our critical jobs during peak usage.”
insideHPC: Can you describe how RTM would help administrators to quickly resolve issues?
Louise Westoby: The operational dashboard included with Platform RTM provides visible status indicators, enabling administrators to quickly identify and correct problems. For example, this would include identifying idle capacity and capacity bottlenecks, and with the capability to monitor resource consumption in real-time, administrators can tune Platform LSF scheduling policies to optimize resource utilization, reduce job pending times, and improve user satisfaction and productivity.
insideHPC: Does RTM mask complexity, or does it help to remove complexity?
Louise Westoby: Platform RTM helps eliminate the complexity associated with monitoring and reporting on the status of the workload. Unlike typical monitoring tools that only monitor the infrastructure, Platform RTM is workload and resource-aware, providing full visibility into the utilization of Platform LSF clusters. It provides a single aggregated dashboard to monitor all workload scheduling facets, including global clusters, hosts, jobs, licenses, queues, users and applications. With its broad set of capabilities, Platform RTM can replace multiple tools or home-growth scripts in typical cluster environments with a single easy-to-use, monitoring tool. This results in improved productivity for administrators and users alike as well as reduced cost and complexity.
insideHPC: What capabilities does Platform Analytics enable for system resource planning?
Louise Westoby: Platform Analytics enables managers, planners and administrators to easily correlate massive amounts of historical workload usage data — jobs, resources and license usage data — from one or more Platform LSF clusters for future decision-making. Also, as with traditional analytics tools, external data sources can be easily combined with workload data to provide data views tailored specifically to an organization’s unique requirements, without the need to build intermediate data views. Platform Analytics turns this data into usable information, making it easy to identify changes in usage patterns. By understanding application and hardware utilization over time, planners can make better decisions, intercept trends, consolidate under-utilized assets more quickly, and ensure that spending is efficient and aligned to business needs.
insideHPC: Can you describe a specific use case where Platform Analytics helped a customer with resource planning?
Louise Westoby: Red Bull Racing uses Platform Analytics to track their cluster usage and identify potential problems that might interfere with running design tests, which are typically very time sensitive and critical to the success of the racing team at the next Formula One event. The software enables the design team to plan for peak resource usage at heavy test times so that the design process runs smoothly.
insideHPC: Why is home-grown scripting not a viable option any more for today’s datacenters?
Louise Westoby: Home-grown scripting is still a viable option for today’s datacenters. However, there are two key reasons datacenters are moving away from scripting. First of all, it is time consuming and introduces a significant amount of complexity in the environment. This is where a product such as Platform Analytics provides a particular benefit by automating the data extraction, correlation and analysis, without the need for complex scripts or intermediate steps.
Secondly, In recent years a new generation of HPC users have emerged that are highly skilled in their area of expertise, but are not computer scientists. They learned how to interact with computer systems – from gaming consoles to smart phones to laptops – using an interactive GUI, and therefore require an easy to use interface. This is why products such as Platform Application Center has become a critical part of Platform Computing’s HPC offering. The product’s easy to use interface makes it easier for users to run HPC applications without programming. For example, the included application-specific job submission templates eliminate the need for complex wrapper scripts, significantly reducing the amount of time it takes to integrate the application with Platform LSF while minimizing user errors during job submissions.
insideHPC: Do you have beta customers for these products? What has been their reaction?
Louise Westoby: The Beta program for these products was very successful. Cadence Design Systems and Simulia both participated in the program and, as you can see from the quotes they provided for the press release, were very happy with the results.
Google has announced plans for a European Data Centre Efficiency Summit, scheduled for May 24 in Zurich, Switzerland. The event will bring together industry leaders and innovators to discuss energy efficiency best practices for data centers. Attendance is free, but space is limited, so Register Now.
A Tip of the Hat goes to Datacenter Knowledge for pointing us to this story.
Rich Evans at Datacenter Knowledge points us to this video, which provides a rare look inside a Google datacenter in South Carolina. The video tour focuses on the extreme measures the company has in place to protect data, including the methodology for wiping and destroying failed hard drives. Read the Full Story.
Datacenter Knowledge just posted this story about Green Revolution’s new submersive server cooling system.
“Green Revolution’s CarnotJet Submersion Cooling System resembles a rack tipped over on its back, filled with 250 gallons of dielectric fluid, with servers inserted vertically into slots in the enclosure. Fluid temperature is maintained by a pump with a heat exchanger, which can be connected to a standard commercial evaporative cooling tower. The company says its solutions will work with OEM servers with slight modifications (removing unneeded fans, applying a coating to hard drives).”
Of course, the use of dielectric fluid is not a new idea in HPC. The CRAY-2, which was the fastest supercomputer of its day, used immersive cooling back in the 1980s. To me, the most interesting thing about Green Revolution is that they seem to have addressed the server warranty problem.
Joe Pizzini writes that a Renderstream customer who works as a security expert at a Fortune 500 firm was able to crack 25 billion passwords in a single second using the company’s GPU solutions.
“As part of our security program, it’s continuously evolving to include the latest attack vectors and ensure that our organization is protected against attack. As part of that is called “penetration testing” which is a field where you attempt to hack the company in a manner that tests how effective the security controls in place are at handling and detecting a potential exposure. We leverage the large capacity of the graphics processors to crack passwords in a manner that greatly expedites how fast you can obtain them using graphics processors.”
Pizzini goes on to describe how the IT industry is wrangling with password complexity requirements to cope with the advancements in password cracking. To remain secure, the adoption of multiple two factor solutions using things like biometrics will be required because “secure” passwords with 50 plus characters are simply impractical.
SNIA Europe is pleased to bring you the Datacenter Technologies Academy in London, the 2011 edition of our highly successful series of events. Our 2011 program focuses on important Datacenter Technologies such as Cloud Computing and Services; Virtual Servers, Storage and Desktops; Contemporary Data Protection; Converged Ethernet and FCoE; Solid State Storage; Data Deduplication and of course IT Systems Energy Efficiency.
The Datacenter Technologies Academy is free to attend for end-users and qualifying channel partners, but registration will close once our maximum numbers are reached. Register now to secure your place!
By Chris Mellor • Get more from this author
Have we witnessed the end of an era? Has the great storage company growth boom ended with a series of big bang acquisitions: 3PAR, Archivas, Compellent, Data Domain, Diligent, EqualLogic, Isilon, LeftHand Networks, Ocarina, Storwize and more? Over the last five years a whole swathe of up-and-coming storage companies has been gobbled up by the majors eager to buy into technology and markets where they have limited or no presence.
The net result is that, in storage terms, EMC, Dell, HDS, HP, IBM, Netapp (less so) and Oracle – the big seven – are bigger than ever, with virtually a whole layer of substantial startups and early-stage public companies becoming operating units of the super seven.
The ones left behind, the too-large-to-buy (Symantec?), the refuseniks, the hard-to-swallow technologies, the mature niche occupants (CA) and the rejects, now have to grow organically and will have a harder job doing so because their previously independent competitors now have big bucks behind them for more development and greater sales and marketing resources.
If BlueArc, DataCore, FalconStor, Pillar, Sepaton or Xiotech ever thought that acquisition was an attractive option, it must now seem as if that door is closed and they have to grow the hard way, the organic way, the DIY way of building great product, keeping up with the market, and selling, selling, selling, which they can do as they have great products and energetic and enthusiastic execs running them.
There are a lot of early and mid-stage start-ups that could grow into substantial mid-tier publicly owned storage companies in the next five years though. Who are they and where are they in market niche terms?
The runners and riders for the next five years:
What about FCoE? It’s early, very early, and until there are complete standards plus compelling cost and simplified management attributes to running Fibre Channel over Ethernet, I can’t see it happening in any big way. If you want to run storage access over Ethernet you can already do it via iSCSI and that, you could argue, will benefit from Converged Enhanced Ethernet just as much as FCoE.
The net result of this could be to let Fibre Channel remain in its hygienic, protected, separate communications fabric as long as it wants, and bring iSCSI into the enterprise data centre: it is already knocking at the door. This matter is a tough issue for Brocade, Emulex and QLogic to deal with.
How about object storage and suppliers such as Caringo? My sense is that that market is just not as hot as it once was. Maybe cloud storage will give it a boost. Equally cloud storage could diminish object storage product sales to end-user customers. It is unclear what will happen.
All the companies I have listed have good technology and hard-driving management. Some are in faster-growing and less heavily contested market niches than others, and have more of a technology advantage over incumbent suppliers. These are the lucky ones. Short-term bets for start-up storage stardom? Here are six: DataDirect Networks, Fusion-io, Nirvanix, Riverbed, Virtual Instruments and Xiotech.®
Bootnote
This is a personal and speculative opinion. Don’t bet your farm or anything else on it. I have shares in CommVault and Overland, so my opinions there could be on the far side of being rosy …
This article originally appeared in The Register.
In this video, the site team from CSCS assembles the Cray XT5 named “Monte Rosa” at the Swiss National Supercomputing Centre in Manno in May 2009. As reported here, CSCS recently ordered a next-generation, multi-streaming Cray XMT system.
At tip of the hat goes to Data Center Knowledge for pointing us to this video.
Advertisement
insideHPC.com is a production of insideHPC, LLC. © 2006-2011 Sitemap
