Post a Comment
FROM EDN EUROPE: Fishing in the data streat as it flows by
By Graham Prophet -- EDN, 12/7/2000
Designing data-communications hardware to deal with high-data-rate streams of aggregated traffic is difficult enough; with growing demands for monitoring the content of those data streams on the fly, it is about to get a lot harder. You need to monitor the content of the data stream for a variety of reasons. One reason is security; you must ensure that the traffic your link is carrying has the right to be there. (Because bandwidth is a traded commodity, bandwidth theft becomes an issue that you must deal with.) Another reason is quality of service (QoS). On the fly, you need to examine the content of the many sessions that constitute an aggregated stream and make decisions about the QoS to which these sessions are entitled. Then you must ensure that the routing hardware handles them with appropriate priority. You would like to do this without disaggregating the data stream. Hi/fn Inc, a silicon and software company, offers this ability. The company deals in data compression, encryption, and this latest function, which it calls "flow classification." Flow classification is a more sophisticated approach than packet classification because it not only can identify and assign QoS attributes to individual packets, but also can track multiple threads and sessions in the data. According to Hi/fn's CEO Chris Kenber, the flow-classification software can identify 750 protocols (out of a possible 3500) that Internet traffic commonly uses; if that doesn't seem like a high number, Kenber adds that only about 50 are common. According to Kenber, looking at the content of frames, the context of one packet to another, and the context of one flow to another is the key to building networks that are both intelligent and secure. The engine that Hi/fn builds to carry out these functions constructs models of the multiple sessions it finds in the data stream Managing this task has taken some innovations in areas such as thread handling and garbage collection, which the company hints may find applications in other areas.
