Switch Architecture for Efficient Transfer of High - Volume Data in Distributed Computing Environment

to be dropped by the TCP layer of the end systems. The redundant packets in the retransmitted segments clog the network switches and thus exagerating the congestion and further packet loss. In distributed computing environment that involves high-volume data transfers, it is very important to minimize packet loss at intermediate switches/routers by maximizing the utilization of memory resources deployed at the switch. Switches deploying shared memory resources are known to incur lowest packet loss compared to other memory allocation schemes. However, the memory being inherently slow, it becomes difficult for a large number of high-speed links to share the common memory space deployed in a shared-memory switch. There have been several attempts made to overcome the memory-speed limit that prevent a shared memory switch from scaling to a larger size. In this paper, we discuss architectures that deploy parallel memory modules to enhance the overall capacity of the shared-memory based switches. We also present performance comparison in terms of memory-speed requirement & packet loss ratio.