Mean Response Time Approximation for HTTP Transactions over Transport Protocols

This paper addresses mean response time that end-users experience when using the Internet. HTTP (Hyper Text Transfer Protocol) is a widely used transfer protocol to retrieve web objects in the Internet. Generally, HTTP uses TCP (Transmission Control Protocol) in a transport layer. But it is known that HTTP interacts with TCP inefficiently. As an example of such inefficiencies, HTTP does not require TCP to deliver the rigid order, which may cause head-of-line blocking. As another transport layer protocol, SCTP (Stream Control Transmission Protocol) has attractive features such as multi-streaming and multi-homing unlike TCP. Within an SCTP association, multi-streaming allows for independent delivery among streams, thus can avoid the head-of-line blocking. In addition, SCTP provides very large number of streams; therefore, it can transfer multiple objects more efficiently than the typical HTTP/1.1 over TCP which limits the number of pipelines. Mean response time is one of the main measures that end users using Internet concern. This paper presents the simple analytical model and algorithm to find the mean response time for HTTP over SCTP including the previous HTTP over TCP. Some computational experiences show that the proposed model and algorithm are well approximated to the real environment. Also, it is shown that mean response time for HTTP over SCTP can be less than that for HTTP over TCP.