Efficient Large Data Transfer Over Interprocess Communication Method

摘要

One of the fundamental issues to operating systems is how to allow multiple processes needing to share resources dynamically communicate each other by exchanging data and control messages, say Interprocess Communication (IPC), which is implemented by using pipes, signals, message queues, semaphores, shared memory, and sockets. The pipe mechanism for IPC offers a convenient way of coordinating processes exchanging data using a pipe; however the emplaced synchronization of the pipe operation would more often result in delaying and blocking the data exchange between the piped processes, especially when communicating with large data. In this paper the behavior of the pipes dealing with large data sets is thoroughly analyzed in order to learn of a most efficient exchange rate between a producer and consumer processes, which leads to dealing with complex data movement optimizations in multiple processors. Linux kernel 3.0 has been installed and modified, upon which the pipe programs are modified and implemented.