Persistent HTTP

Persistent HTTP
An Introduction

Prof. David Bernstein
James Madison University

Computer Science Department

bernstdh@jmu.edu

Motivation

A Typical Application of HTTP

A Client-Side "Fix"

GET the HTML file using one connection.
Parse the HTML file and use "parallel" connections for the subsequent GET requests.

Problems with this Approach

Multiple TCP Connections:
- Each of which requires a 3-way handshake
Short TCP Connections:
- Stuck in slow start
- Small congestion window
- Bad round-trip-time estimates

Persistence

The Idea:
- Use one TCP connection for multiple GET requests
The Process:
- Server leaves the connection open after sending a response
- Client uses an existing connection if there is one

Persistent HTTP

An HTML File with One IMG Element

Persistence and Pipelining

Without Pipelining:
- Client issues a new request after the previous response has been received
With Pipelining:
- Client sends request immediately
- Server must handle responses in the order of the requests