Server capacity planning is the process of monitoring and alerting on server resources, such as CPU, memory, disk, and network. Being able to forecast when your server will run out of resources will help optimize hardware investments.
Unlike customer-service applications where the load is defined by the company's activities, the marketplace for e-commerce is basically beyond the control of the company. This brings new challenges to workload characterization and forecasting. It also requires "what if" performance models to help operations plan for different scenarios with very different traffic volumes.
Well, in general, the site's IT infrastructure wasn't ready to support the increased traffic or the delivery of larger files such as images and sound files. The IT managers should have asked themselves questions such as: Is the bandwidth of the link connecting our site to the ISP large enough to support the traffic increase while maintaining acceptable service levels? Are the various servers -- HTTP, authentication, and database -- that support the e-commerce site properly sized to handle the new load? What percentage of connections will be rejected when the load increases? Where will the bottlenecks be during the holiday season? Should we have more computing power, more bandwidth, higher I/O bandwidth, or a different system configuration? These are typical capacity-planning questions. In this article we provide you with a roadmap to performing capacity planning for Web environments.
For example, you may establish as a goal for your Web site that the average response time for static pages should not exceed 0.5 second and that database searches should not take more than 1 second, on the average, for a load of 6 HTTP requests per second. You could also establish that if the load doubles to 12 HTTP requests per second, performance should not degrade by more than 30 percent. You may also want to establish that a maximum of 1 percent of the connections may be refused at peak hours.
Unlike customer-service applications where the load is defined by the company's activities, the marketplace for e-commerce is basically beyond the control of the company. This brings new challenges to workload characterization and forecasting. It also requires "what if" performance models to help operations plan for different scenarios with very different traffic volumes.
Well, in general, the site's IT infrastructure wasn't ready to support the increased traffic or the delivery of larger files such as images and sound files. The IT managers should have asked themselves questions such as: Is the bandwidth of the link connecting our site to the ISP large enough to support the traffic increase while maintaining acceptable service levels? Are the various servers -- HTTP, authentication, and database -- that support the e-commerce site properly sized to handle the new load? What percentage of connections will be rejected when the load increases? Where will the bottlenecks be during the holiday season? Should we have more computing power, more bandwidth, higher I/O bandwidth, or a different system configuration? These are typical capacity-planning questions. In this article we provide you with a roadmap to performing capacity planning for Web environments.
For example, you may establish as a goal for your Web site that the average response time for static pages should not exceed 0.5 second and that database searches should not take more than 1 second, on the average, for a load of 6 HTTP requests per second. You could also establish that if the load doubles to 12 HTTP requests per second, performance should not degrade by more than 30 percent. You may also want to establish that a maximum of 1 percent of the connections may be refused at peak hours.
No comments:
Post a Comment