SCIENCE

By Chris O'Neal -- Last week, Appro announced the award of an additional supercomputing cluster from Lawrence Livermore National Laboratory (LLNL). Its peak performance is 33.18 TFLOPS with a total of 13,864GB of memory available. To take a closer look, Supercomputing Online interviewed Maria McLaughlin, Director of Marketing at Appro. SC Online: What is the unique value proposition of Appro? Why do the leaders in the supercomputing field such as LLNL come to your company rather than say, Dell? McLaughlin: High performance computing (HPC) is Appro’s focus. Appro offers high-value, end-to-end HPC cluster solutions that are flexible, powerful and scalable. By providing industry-standard and open source technologies, we deliver HPC solutions with a superior price/performance ratio. In addition, Appro offers a choice of Intel or AMD processors, high performance network connectivity solutions with multiple I/O options and outstanding professional services. This is evident with Appro’s new supercomputing cluster deployment at LLNL. With the addition of the Minos cluster, Lawrence Livermore National Laboratory will have a total of 20 SU, 2,880 nodes or 23,040 cores of supercomputing power available for programmatic computing projects. SC Online: What area of the market is your strongest niche? I’m interested here in where you are really gaining ground as opposed to where you would like to gain ground? McLaughlin: Appro has always been focused on the HPC market but the HPC market is not a niche market anymore and it is growing tremendously over the last years. HPC covers all computers that are used for highly computational or data-intensive tasks. IDC now uses these terms to cover all technical servers used by scientists, engineers, financial analysts, and others. According to IDC data, the server market is now $52 billion representing 4.2 CAGR in server factory revenue over last 4 years and HPC at all-time high is now $10 billion representing 20.9% CAGR in server factory revenue over last 4 years. Within the HPC market, Appro is gaining ground in the following verticals: * Oil and Gas * Financial Services * Government Labs/ Defense * Education * Electronic Design Automation * Digital Content Creation * Manufacturing SC Online: What parts of the server market do you not serve? McLaughlin: ISP’s, web serving, file server and others are examples of markets that we do not serve in high scale. Since are focal point in HPC, we do everything that we can to absolutely provide the best products and services to serve the successful vertical markets. HPC business is very dynamic and the need for additional performance for the need of doing more/better science is making this market grow tremendously. Appro will always be in the look out to deliver the best in class HPC cluster solutions to our customers. SC Online: How long have you had a blade out on the market? Please tell its brief history. McLaughlin: APPRO HyperBlade was launched in 2003. The launch was in conjunction with AMD Opteron availability announcement. Shortly after that, Appro announced the HyperBlade Intel Xeon based cluster server at the Intel Developer Forum. HyperBlade marked a significant milestone for Appro as the company delivered high-density blade architecture by using commodity x86 components to address the high-performance and technical computing markets. Appro also partnered with Raytheon, HPC Solutions Group to market HyperBlade cluster solutions and Raytheon's value added services to address government/defense HPC vertical markets. HyperBlade is a unique platform; it provides a combination of advantages of blade servers and rack-mounted servers. HyperBlade offers fully integrated cluster solution using a hybrid blade/1U form factor. Like blades, HyperBlades plug into a common backplane, eliminating cable clutter. Like 1U rack mounted servers, HyperBlades are also independent servers; vertically mounted, they double the traditional server density, making them ideal for HPC server consolidation in the datacenter. SC Online: What's new about this blade in the past few months? McLaughlin: We have been updating HyperBlade for years now. Appro is an Intel Premier Provider and AMD – Platinum provider. By saying this, we have been reviewing the HyperBlade platforms to provide the best server density for HPC server consolidation efforts with great price/performance. We offer the Dual processor servers based on AMD Opteron 2000 series and Intel Xeon Processors 3000, 5100 and 5300 series. One of the latest upgrade with the Appro HyperBlade is the launch of the HyperBlade Servers based on Intel Xeon processors 3000 series. This server is unique positioned for a special group within the HPC market. These HyperBlades offer two boards per chassis nearly cutting the chassis and rack infrastructure cost in half. This cost-effective blade server delivers essential server-class features to reduce size, cost, power consumption and complexity. In addition it offers a small form factor board for high performance and high density computing applications enabling greatest system-level innovation. The two socket HyperBlade server solution features one PCI-Express x8 slot enabling added I/O such as GbE, Infiniband, SAS RAID cards with 4 DIMM slots (maximum 8GB) of DDR2 533/667 memory. It delivers maximum flexibility, high-performance and value. SC Online: What can we expect for the rest of the year in terms of upgrades etc to this blade? McLaughlin: Appro will continue to upgrade the HyperBlade. We will offer the latest on AMD and Intel multicore technologies. In addition to the HyperBlade, Appro is working on a new blade design this year and we will be launching it sometime in the end of Q4. SC Online: Please give us a brief history XtremeServer series. McLaughlin: Appro introduced XtremeServers the 1U and 3U based on AMD Opteron processors in 2006. The new 1U and 3U XtremeServers provide innovative server technology designed to deliver outstanding performance and reliability with extreme memory scalability for compute and memory intensive applications. It is important to say that these new XtremeServers are the first to be built on motherboards designed by Appro. One of the important features in these servers design is that our motherboard comes with eight memory slots per processor instead of only four slots supported by other motherboard manufactures. This provides to our customers enormous system memory flexibility. The APPRO 1U XtremeServer allows up to 64GB of on-board memory while the APPRO 3U XtremeServer permits up to 128GB. The memory capacity for the servers represented a new paradigm for server memory availability for the HPC market! SC Online: Please provide a brief description of your company’s clusters at LLNL? McLaughlin: In June 2006, Lawrence Livermore National Laboratory awarded Appro the Peloton Project to supply clusters used in solving complex computational problems. The deployment involved cutting-edge technology, three Appro 1U Quad XtremeServer Clusters with a total of 16,128 cores with latest Dual-Core AMD Opteron Processors, technical expertise and execution ability and commitment. Appro delivered “Altas”, Rhea and Zeus” supercomputing clusters to LLNL last December. Lawrence Livermore National Laboratory is expanding the Supercomputing Projects using Appro Cluster Solutions. This additional cluster deployment, named “Minos” will consist of a supercomputing made up by Appro 1U Quad XtremeServer Linux Cluster with a total of 6 scalable units each with 144 nodes for a total of 864 nodes/6,912 cores based on the latest Dual-Core AMD Opteron Processors. Its peak performance is of 33.18 TFLOPS with total of 13,864GB of memory available. These clusters are estimated to be generally deployed by end of July 2007 and available to LLNL by August. The Minos cluster will join the already installed 4 SU, 576 node “Rhea” cluster and will be used for Stockpile Stewardship capacity computing workload: jobs with 256-4,192 way MPI Parallelism. Stockpile stewardship is the U.S. Department of Energy/National Nuclear Security Administration (NNSA) program to ensure the safety, security and reliability of the nation’s nuclear deterrent without underground testing. Minos will be one of the machines used by NNSA’s Advanced Simulation and Computing for stockpile stewardship calculations. This will entail running the integrated weapons design codes in support of the Reliable Replacement Warhead (RRW) design effort as well as laser plasma interaction studies for the National Ignition Facility (NIF). SC Online: What's on the horizon for the XtremeServer line? McLaughlin: All XtremeServer line will be supporting the new AMD Quad core processors that will be coming out in July/August timeframe. SC Online: What are HyperServers? McLaughlin: Our Hyperservers are based on Intel Xeon processors. Like the XtremeServers, our Hyperserver series offer powerful, flexible and reliable architecture for a great HPC cluster foundation. SC Online: Please provide a brief history of the HyperServers. McLaughlin: We actually had made the HyperServers available last month. They are brand new addition in our product line. Our financial customers are very pleased with them and we have a variety of platforms to choose from HPC to multi-purpose servers. This has been a very flexible line of servers. As you can see, we have been busy... and that is an exciting moment for us. SC Online: Anything else big planned for this year? McLaughlin: Yes. We do have big plans for this year. Watch out for Appro’s next generation blade server architecture and smart and easy to use cluster management solution. SC Online: What percentage of your business is Intel versus AMD and why? McLaughlin: We do fair percentage of our business with both partners. SC Online: What percentage is Linux versus Windows and why? McLaughlin: 90% Linux - the answer is simple: total cost of ownership. Supercomputing Online wishes to thank Maria McLaughlin for her time and insight.