Reducing US/NZ Web Page Latencies

M. Pearsona, A. McGregorb and J. Clearya aThe University of Waikato,
Private Bag 3107, Hamilton, New Zealand bNational Laboratory for Applied Network Research (NLANR),
San Diego Super Computer Center,
10100 Hopkins Drive, San Diego, CA 92186-0505, USA

Abstract:

Much of New Zealand's Web traffic crosses the Pacific Ocean on sub-oceanic fiber channels. This imposes approximately 60ms of latency in each direction. Because fetching a web page often requires the fetching of many HTML components and each component requires several round trip times this latency is multiplied many times in a typical web page retrieval. Caches and proxies are often used to reduce the effect of these delays. Providers of Internet service have several choices in how to deploy these components. Depending on the architecture chosen the traffic on the international circuit might be made up of a large number of independent connections or a smaller number of connections carrying aggregated traffic. The appropriate approach is not immediately apparent because there are opposing performance factors. In this paper we describe a discrete event simulation of the effect of carrying multiplexed HTTP connections over an simulated NZ/US circuit. We show that a high degree of multiplexing mitigates against TCP's bandwidth delay product limits but that carrying every HTTP connection in a seperate TCP causes a significant increase in delay. The simulation environment used includes a TCP/IP stack derived from a real TCP implementation and is driven by data collected from a large New Zealand web cache.

TCP/IP performance, HTTP, cache, proxy