Enhanced performance for workload-optimized HPC solutions

The IBM Intelligent Cluster increases the power and flexibility of HPC by combining new processors, networking components, management tools and GPUs with a streamlined delivery process

Email the article Download this pdfDownload pdf

IntelligentThe IBM® Intelligent Cluster—previously called the IBM System Cluster 1350—incorporates a range of new technologies and capabilities that will help organizations deploy HPC clusters optimized for their specific workloads. The integration of the latest Intel processors, along with availability of graphics processing units (GPUs), can help organizations achieve outstanding performance for a wide range of applications while controlling costs. Meanwhile, the addition of new networking components enables organizations to configure connectivity for their precise needs. 

IBM is also making purchasing and deployment simpler than ever. Organizations can select preconfigured Intelligent Cluster solutions for ease of deployment, cost reduction and a single point of service and support. Once the cluster is installed, new management tools facilitate rapid provisioning so business and research groups can move between workloads quickly and maximize the value of their cluster. 

Intel processors deliver intelligent, scalable performance
The Intelligent Cluster includes IBM System x® rack, BladeCenter® or iDataPlex™ servers equipped with the latest generation of Intel® Xeon® processors to deliver exceptional performance and price/performance. The Intel Xeon processor 5600 series offers intelligent performance that automatically adjusts processing and power usage to meet workload requirements. The result is significantly improved application performance and energy efficiency compared with previous-generation processors.

IBM eX5 rack and blade servers can be equipped with the Intel Xeon processor 7500 series, which combine exceptional raw compute power with high memory bandwidth and support for large-scale memory capacity. The new processors enable organizations to achieve extreme scaling within each node for running demanding workloads on a compact system. IBM MAX5 memory expansion technology for eX5 servers provides leadership in memory capacity for the Intel Xeon processor 7500 series.

GPUs boost performance and control costs
Whether trying to improve the precision of seismic models, increase performance of financial analyses or accelerate graphics rendering, organizations are increasingly looking to GPUs to enhance cluster performance. The IBM Intelligent Cluster offers IBM iDataPlex servers with NVIDIA GPUs to help deliver even better price/performance for x86-based clusters by augmenting the microprocessor computations. With up to two individually replaceable GPUs per server, the iDataPlex systems can achieve up to ten times the performance on certain HPC workloads compared with servers without GPUs. In addition, the GPUs are incorporated in an energyefficient iDataPlex design that helps drive down power and cooling costs.

New networking options provide greater choice
In addition to existing networking options from Blade Network Technology, Brocade, Chelsio Communications, Cisco, Force10 Networks, Mellanox, QLogic and Voltaire, the Intelligent Cluster offers new, leading-edge networking component options, providing greater performance plus increased flexibility for meeting organizational needs. Mellanox InfiniScale IV QDR InfiniBand Switches and ConnectX-2 QDR Host Channel adapters deliver up to 30 percent better performance at a 40 percent lower cost than previous solutions.

Organizations also can choose from new Ethernet switching options. Force10 Networks 1 and 10 Gb Ethernet switches offer dense, wire-speed performance and low latency, while IBM DataCenter Networks Ethernet switches leverage Brocade technology to deliver data center–level performance, optimum flexibility and operational efficiency with simplified management. Organizations also can choose Mellanox 10 Gb Ethernet adapters, which provide low latency, support for data center Ethernet (DCE) and advanced quality-of-service features.

Management tools accelerate deployment
Key management tools included with the Intelligent Cluster help accelerate cluster provisioning. Administrators can use the included xCAT (Extreme Cloud Administration Toolkit) plus Moab workload management tools to dynamically provision clusters of any size within minutes instead of hours or days. These tools help maximize the value of the cluster by accommodating multiple business groups or research teams who need to run distinct applications and OS versions. Now administrators can prepare the cluster for the next group without the time-consuming processes of recoding and retesting.

Preconfigured solutions make it simple to capitalize on HPC
To speed up cluster delivery, IBM has integrated the Intelligent Cluster into the IBM Intelligent Solutions Factory—the streamlined process that IBM has used for years to build and deliver preconfigured solutions, such as the SAP NetWeaver Business Warehouse Accelerator and the InfoSphere™ Balanced Warehouse. The predefined, pretested Intelligent Cluster solutions incorporate a full array of software and hardware components from IBM and its partners, giving organizations everything they need to run their workloads in one integrated solution.

Once a preconfigured solution is selected, organizations can simply choose the correct size—small, medium, large or extra large—to fit their needs. Additional requirements can be accommodated by IBM solutions architects. With preconfigured Intelligent Cluster solutions, organizations no longer have to spend the time and resources researching, integrating and testing components. IBM eliminates the risk and accelerates the time to value, so organizations can capitalize on the power of HPC right away.

For more information about IBM Intelligent Clusters, visit:

To learn more about the Intel Xeon processor 5600 and 7500 series in IBM servers, read Srini Chari, “Tick Tock: Sustained, Smart and Steadfast Journey for Faster High Performance Computing” April 2010.