Thursday was my second day at OOW 2004. Yesterday, we presented our first session, everyhting went fine. Today we are presenting to more sessions: "Project MegaGrid: Performance Management in Large Clusters" and "MegaGrid: Capacity Planning for Large Commodity Clusters". Both of the presentations were a success. If you plan on migrating from a single instance to a grid/cluster environment especially the capacity planning presenation will provide you with a methodology to achieve that task.
Abstract: "Project MegaGrid: Performance Management in Large Clusters"
In database clusters consisting of a large number of Oracle instances on small machines that provide the resources for one or more applications, the collection and analysis of performance data are challenging. With functions such as dynamic workload distribution and dynamic Oracle Database 10g Real Application Clusters (RAC) kernel adjustments to optimize certain workloads, as well as the trend of using nodes with different capacity, this task becomes more complex. This session presents techniques and methods leveraging the specific features for RAC built into the Automatic Workload Repository (AWR) and other functions of the Server Manageability framework for RAC. The approach is practical and uses examples from real-world applications. The presentation also attempts to integrate SQL advisories, execution plan, and data access optimizations into the method. It provides a view of future requirements for complete and easy performance management in large database clusters. The presentation can be downloaded from here
Abstract: "MegaGrid: Capacity Planning for Large Commodity Clusters"
Customers are evaluating the deployment of low-cost, high-performance servers and networked storage to build an enterprise grid computing environment in which resources can be dynamically added and removed to meet processing demands. The primary resources in an enterprise grid computing environment consist of CPUs, servers for database and application processing, networks and network devices such as routers, switches or IP virtualization devices, disks, and networked storage arrays. The expected performance or service-level objectives of the business should determine the server hardware, network topology, networked storage systems, OS, and application configuration requirements and drive the initial configuration for deployment. The planning also should involve consideration of the future growth of the services hosted in the environment. Using a workload model designed for clusters of Oracle Database 10g Real Application Cluster instances, this session reviews design characteristics and how to predict response times and throughput with Oracle statistics and suggests the optimal way to balance load in an existing or simulated system and how to translate service-level requirements and workload characteristics into required hardware resources. The presentation can be downloaded from here
No comments:
Post a Comment