WWW: Beyond the Basics

25. Methods for Web Bandwidth and Response Time Improvement

25.3. Alternative distribution

Another method to reduce bandwidth usage along the backbone is to use alternative distribution means. These include using multicast document delivery and remote­site distribution methods.

25.3.1 Multicast document delivery

Multicast delivery of Web documents is an experimental concept in which an alternative network delivery, or transport, mechanism is used. Typically Web document delivery is accomplished through the use of point­to­point (unicast) network connections, effectively creating a virtual wire from the source to the destination. Multicast Web services are on the point­to­multipoint model. That is an origin server (document source) will send one copy to multiple recipients. Multicast allows one copy of the message to be sent to multiple destinations, whereas multiple copies are needed in a unicast service.

One idea behind multicast Web services is that documents will be broadcast on a round­robin schedule and that clients join the multicast group and listen for the document that they are interested in. The primarily benefit is that packet traffic is reduced since multiple copies are not sent. However, the proposed scheme has an increased latency problem due to the round­robin document broadcast. It also may result in wasted bandwidth if nobody wants the document. An alternative idea is that the document is multicast after a certain number of requests or elapsed time period. One general problem to any multicast scheme is that multicast services are not currently available to every user.

An experimental version of NCSA Mosaic that implements multicast document delivery is available at ftp://sig.enst.fr/pub/multicast/mMosaic-2_7b4m3_tar.gz.

25.3.2 Remote distribution mechanisms

The remote distribution concept is similar to a common file distribution model used in FTP. Under this scheme, there are multiple server sites that contain copies of the document. These sites are generally closer to the user, which reduces the load and bandwidth requirements for individual servers. Since the site is closer to the user, response time is improved.

The primary difference between this method and caching is that the user may know that there are multiple sites that contain the document. The user chooses the site closest to him or her in this model. In a cache hierarchy, the cache server will retrieve the document from the closest copy in the caching hierarchy.

[PREV][NEXT][UP][HOME][VT CS]

Copyright © 1996 David C. Lee, All Rights Reserved

David C. Lee <dlee@vt.edu>
Last modified: Mon Nov 18 15:22:36 EST 1996