• Step 2 - Change Management: Clusters change over time as new nodes or architectures are deployed, operating systems are updated, and interconnects are added. An enterprise Linux cluster management solution enables SAs to manage heterogeneous clusters as a cohesive entity through a single point of management, not just as a collection of individual machines, allowing quick updates and modifications of clusters through a simple user interface. For example, when adding a node to a cluster, the management tool should handle all configuration changes both on the new node and on any other node running cluster services that need to be updated (e.g., NIS, job schedulers, cluster file systems).

• Step 3 - Health and Resource Monitoring: The management solution should provide a quick and comprehensive view on the availability, performance, and utilization of all resources. For larger environments, it is crucial to have aggregated and filtered monitoring capabilities in order to provide actionable information on the cluster as a whole rather than per-node data. Alarming capabilities are important to alert SAs to problems in order to reduce downtime. The most advanced management products use fault prediction algorithms to proactively highlight potential issues before a job starts, instead of after a cluster crashes.

• Step 4 - Fault Handling and Root Cause Analysis: To handle faults effectively and identify root causes, the management solution must provide information on all cluster resources – server hardware and OS, I/O, and interconnect. The SA can then start from a high-level view of problems in the environment and then drill down into detailed data points for relevant resources. To further enable handling large and complex environments, automated corrective actions should be enabled in response to triggered alarms. With faster root cause analysis and automated response, downtime can be minimized as clusters quickly return to full operation.

• Step 5 - Resource and Performance Optimization: Proactive management is essential in optimizing performance and achieving maximum return on investment. Advanced cluster management solutions enable SAs to understand the relationship between the applications running and the hardware that is used, to tune the environment for maximum performance. SAs can then isolate hardware bottlenecks as well as application-related issues such as data set size or communication overload situations.

• Step 6 - Cluster Expansion and Capacity Planning: Advanced management solutions include a resource usage database that captures historical information about the use of a cluster. With this data, IT Managers can understand how infrastructure is being used, and whether it is deployed with a proper balance between processor power, memory, interconnect, and storage. Information from the historical database is also used for planning purposes, providing insight into how best to increase cluster capacity and performance.

The first step toward a wide range of benefits
When clusters are managed across their lifecycle, IT organizations can take the first steps toward using shared Linux resources in meeting stringent service level agreements, as well as achieving:

• Lower management costs and complexity by standardizing on a unified management system and streamlining the administration of clusters;

• Increased utilization, performance and output by having tools to analyze results, improve service levels, and minimize downtime, thereby meeting enterprise computing needs;

• Unification and centralization of cluster management activities, thereby effectively handling resources across technological or geographical boundaries;

• Flexible alignment of clusters with changing business objectives to drive results; and,

• Preparation to take advantage of strategic technology initiatives such as utility and grid computing.

Clearly, enterprise-class management holds the key to achieving the true benefits of Linux-based cluster computing: cost-effective, high performance compute resources that can be easily shared and repurposed to meet dynamic business requirements. The good news is that these management tools exist today and are market-proven in enterprise environments.