Optimizing the resource-updating period behavior of HTTP cache servers for better scalability of live HTTP streaming systems

The resource-updating period behavior of mainstream HTTP cache server implementations is suboptimal in that these cache servers send redundant requests to the upstream server if clients request a file that is already being updated from the upstream server. Although this is a corner case issue in previous web applications, its ramifications is significant in live HTTP streaming systems because it will generate periodical unnecessarily high peak load on the upstream HTTP server and thus prevent the system from achieving the expected capacity. We suggest an optimized resource-updating period behavior for cache servers to eliminate this problem. The suggested behavior is implemented on the popular Nginx cache server. And evaluation experiment demonstrated the feasibility and effectiveness of the proposed optimization.