DESIGN AND IMPLEMENTATION OF CLASS BASED QUEUING IN PROXY SERVER

Abstract

Growth of Internet as a fundamental network has resulted in very slow Internet access. This situation is further exacerbated by a small number of users who use a disproportionate share of the available bandwidth for file downloads, etc., starving other users of bandwidth. Therefore some means of controlling and managing the bandwidth available to users is necessary. We can manage bandwidth by classifying traffic in to various categories and allocating each class a portion of total available bandwidth, thus we can restrict high bandwidth consuming applications like movie, real audio, MP3 download to a limit, as specified by network administrator. Most organization provides web access to end users through cache/proxy servers. As all web accesses are channeled through the cache server, using the same server for bandwidth management is expedient. Squid is a popular proxy and web cache server, which contain a bandwidth management feature called "delay pools". However the existing implementation of delay pools did not provide satisfactory performance, as bandwidth allocations are static and does not adept to varying usage patterns. Our solution adjusts the bandwidth allocated to each class to provide an optimum level of access to all users. During off-peak hours, each class of application can use essentially the full available bandwidth, while during peak hours it's bandwidth is limited to min value as specified by network administrator. Thus light interactive users get better performance at the expense of heavy and non-interactive users