Clustering offers significant price/performance advantages for many high-performance workloads. Linux clusters can further extend these advantages by harnessing low-cost servers and Open Source software.

Today, many businesses are building their own Linux clusters using commodity hardware, standard interconnects and networking technology, Open Source software and in-house or third-party applications. They discover that this activity consumes considerable resources to assemble, integrate, test, manage and support the cluster. The cost of overcoming the complexity of assembling these piece parts themselves is often far greater than the benefit gained.

Now customers can benefit from IBM's extensive experience with clustered UNIX® computers to help minimize this complexity and risk. IBM has designed the IBM ® Cluster 1350 to address each of these challenges. Using advanced Intel® and AMD processor-based server nodes, proven cluster management software and optional high-speed interconnects, the Cluster 1350 brings together the best of IBM and third-party technology. As a result, the installation of a Linux cluster can be greatly accelerated and its management and support simplified.

The Cluster 1350 is designed to be an ideal solution for industrial, financial services, life sciences, governmental and educational organizations requiring excellent price/performance for handling high performance computing (HPC) workloads. It is also an excellent choice for applications that require horizontal scaling capabilities such as Web serving and collaboration.

A comprehensive solution
The Cluster 1350 is a flexible, integrated offering which includes all of the hardware necessary to create a comprehensive cluster system tailored to specific customer needs. When combined with the customer's choice of Linux operating system and IBM Cluster Systems Management (CSM) for Linux software, the Cluster 1350 provides an easy to configure platform to help customers rapidly deploy cluster applications.

By reducing time and resources required for configuring, assembling, integrating, testing and tuning a Linux cluster, the Cluster 1350 can help speed time-to-production. More servers can be added at any time to handle increasing workloads and consolidate more servers or new applications.

Customers may acquire a supported version of the Linux operating system from IBM, a Linux distributor, or an authorized IBM Business Partner or reseller. In addition, customers may choose to install Linux and CSM themselves or to contract with IBM or an IBM Business Partner to perform the service.

IBM installation services may also provide planning, installation and configuration of additional cluster software including IBM General Parallel File System (GPFS) for Linux. For even higher levels of support, the optional Support Line for Linux Clusters is staffed by experts who understand the entire cluster environment, not just the individual components. These Support Line services cover all cluster hardware components, the Linux operating system, CSM for Linux and GPFS for Linux software.

To further simplify the deployment effort, IBM can provide project management support to help coordinate all aspects of delivery and installation, including hardware and software setup services. Attractive financing and leasing terms are also available.

High-performance cluster management
IBM offers CSM for Linux, advanced cluster management software that allows a cluster of supported Intel and AMD processor-based servers running Linux to be managed from a single point-of-control. This capability simplifies the management of the cluster and easily scales with the cluster to improve the efficiency of the system administrator.

CSM provides a single point-of-control that allows the administrator to monitor both hardware and software events. In addition, CSM will trigger automated recovery actions when appropriate. These event monitoring and automated recovery features enhance the efficiency of the administrator and the reliability of the cluster by enabling rapid problem detection and resolution.

CSM for Linux allows IBM System-x (xseries)®, IBM BladeCenterT and IBM eServer 325 Intel and AMD processor-based servers running Linux to be managed from a single point-of-control. CSM is also available for the AIX 5LT operating system and allows non-switched clusters of IBM eServer System-p5 (pseries)® nodes running AIX 5L or Linux to run in the same cluster with Intel and AMD processor-based server nodes. System-x (xseries) and BladeCenter nodes can also be monitored and controlled using IBM's award-winning management software for servers (PC Magazine Editor's Choice October 2002), IBM Director 4.1.

CSM contains several components designed to help simplify the management and increase the reliability and availability of Linux clusters:

• Distributed management server:
  provides a persistent repository of information about each node in the cluster, and maintains the status of each node.
   
• Event response resource manager:
  provides the ability to run a rich set of pre-defined commands or scripts in response to user-defined events. Cluster resources that can be monitored include nodes, adapters, file systems and processes.
   
• Remote hardware control:
  utilizes the integrated systems management processor in Cluster 1350 nodes. This feature enables the administrator to remotely reset or power the node on or off.
   
• Configuration file management:
  provides a repository for files that are common between the nodes. CSM synchronizes changes to configuration files across the cluster.
   
• Distributed shell:
  allows commands or scripts to be run remotely on all nodes in the cluster with options for combining output from multiple servers. Distributed command execution manager is an optional graphical user interface that integrates with the distributed shell, allowing easier management of nodes and node groups.

CSM provides node grouping, which is a convenient way to apply different rules to subsets of servers in the cluster. This capability is especially important when the cluster is deployed to consolidate multiple applications.

Administrative commands can be applied to an individual node, the entire cluster as a unit, or to groups of nodes as defined by the system administrator.

By providing a single point-of-control with robust cluster management capabilities, CSM can dramatically simplify total systems management and help increase cluster availability.

Advanced server technology
The Cluster 1350 is built from Intel and AMD processor-based servers. The Intel servers are System-x (xseries) systems including the x335 and x345 and the BladeCenter offering with HS20 blades. They are based upon IBM Enterprise X-ArchitectureT which combines key availability features of IBM eServer zSeries® servers with the advanced scalability features of System-p5 (pseries) systems. These servers offer affordable enterprise power, scalability and control that customers have come to expect from IBM.

The Cluster 1350 also offers customers an even broader choice of leading server technologies by including AMD OpteronT server technology. The IBM eServer 325 (e325) server features Opteron processors incorporating an integrated memory controller and Hypertransport technology to help minimize bottlenecks related to memory access and I/O bandwidth. The e325 is ideal for customers that require leading 32-bit performance combined with a simplified, affordable migration path to future 64-bit computing environments.

The x335 rack-dense servers continue to provide leading-edge Intel XeonT processor-based compute performance and manageability. For customers who value the combination of performance, density, integration and investment protection, the BladeCenter HS20 provides ideal compute node capabilities by combining Xeon processor technology with a revolutionary architecture that delivers double the compute density available in a traditional 1U server. And, the x345 delivers outstanding management server and cluster I/O performance features to meet the range of customers' cluster management and storage requirements.

Tested, flexible configuration
Standard configurations of the Cluster 1350 include a management node and up to 511 cluster nodes, including up to 32 optional storage nodes that provide shared file storage. A special order process is available for customers requiring larger or other non-standard configurations.

Each Cluster 1350 also includes a management Ethernet VLAN for highly secure internode communications, a cluster Ethernet VLAN for application internode communication and a terminal server network, which provides remote console capability. The cluster comes standard with one 10/100Mbps Ethernet switch for the management VLAN and a choice of 10/100Mbps Ethernet switch, Gigabit Ethernet switch or MyrinetT-2000 switch for the cluster VLAN.

Compute nodes (x335, x345, HS20 blades or e325) can be configured with either single or dual processors, with up to 8GB of memory (up to 12GB on e325). Compute nodes can be configured with either one or two disk drives (maximum storage capacity per node depends on choice of node type).

The management node has dual Intel Xeon processors, from 512MB to 8GB of memory, up to 880GB of hot-swappable disk storage and the appropriate adapter cards for cluster management.

As previously noted, additional disk storage is available through the use of optional storage nodes, which allow additional file system storage to be configured.

For example, an x345 storage node can be configured with single or dual Xeon processors, from 512MB to 8GB of memory and up to 880GB of hot-swappable disk storage.

For increased capacity, these nodes can be configured to support externally attached storage subsystems. For high availability, they can be configured to provide redundant paths to all data. The standard configuration can support up to 32 storage nodes.

A minimum of one Keyboard/Video/Mouse (KVM) switch is required with the system. Remote console support is provided through the terminal server.

Expanding possibilities
The Cluster 1350 offers a number of optional components to meet specific computing requirements, including a choice of interconnect technologies.

In addition to standard 10/100Mbps Ethernet or Gigabit Ethernet, customers can select Myrinet-2000-scalable interconnect technology from Myricom, Inc. Myrinet is a cost-effective, high-performance packet communication and switching technology that has been widely used in Linux operating system-based clusters. It is particularly well suited for high-performance or high-availability clustering.

Companies can also take advantage of GPFS for Linux. GPFS is a high- performance, scalable, shared-disk file system that provides fast data access from all nodes in a Linux cluster and NFS export capabilities outside the cluster. Parallel applications running across multiple nodes of the cluster as well as serial applications running on a single node can readily access shared files using standard UNIX file system interfaces. Furthermore, GPFS can be configured for failover from both disk and server malfunctions.

Fibre Array Storage Technology (FAStT) provides highly reliable data storage for business-critical applications that require high-speed transfer and large amounts of data. Optional cluster components include IBM FAStT600, FAStT700 and FAStT900 Storage Servers with the FAStT EXP400 and EXP700 Storage Expansion Units.

Summary
Creating a computing infrastructure is an exercise in balancing price and performance to deliver the appropriate solution for any given task.

For some high-performance workloads, the most appropriate solution is clustering. By harnessing the power of many servers in parallel, it's possible to solve computationally intense problems with an excellent price/performance ratio. Clustering can also be an excellent approach for consolidating multiple workloads, which can provide enhanced manageability and high availability.

The advent of Linux has made it possible to build powerful clustered systems using affordable, Intel and AMD processor-based hardware. It also allows organizations to tap into the growing skill base and contributions of the Open Source community.

The IBM eServer Cluster 1350 is a comprehensive solution that can greatly simplify and speed deployment of a Linux cluster. IBM provides a single point-of-contact for the entire cluster, not just individual components, thus reducing risk for deployment of a Linux cluster solution.

For any organization looking for the economic advantages of deploying a Linux cluster, but concerned about the time and technical resources necessary, the Cluster 1350 is the right choice.