Friday, May 23, 2014

Cisco Patent for Traffic Classification in Wi-Fi

A patent application by Cisco Technology (a subsidiary of Cisco Systems for patents and trademarks) for classification of traffic for application-aware policies was published.

Inventors are Chandrasekaran BalachanderNalamalapu Kasi, and Nayak Seema;


In one embodiment, a method includes performing stateful application classification on packets received at a controller and transmitting classification information to an access point. The classification information includes flow information and stateless rules for applying policies. The access point is configured to use the classification information to perform stateless application classification and apply policies to packets received from a mobile device. An apparatus and logic are also disclosed herein.

Network policies such as QoS policies are typically applied at either an access point or a controller in a wireless network. Each implementation has drawbacks for application aware policies. For example, since stateful application classification is based on multiple packets within a flow, classification at the access point results in throughput and roaming issues. Classification performed at the controller does not allow for prioritization of traffic from wireless clients across the wired network.

  • In one embodiment, the stateful classifier 18 is a classification engine configured for NBAR (Network Based Application Recognition) or other technology used to classify applications. The classifier 18 is operable to recognize a wide variety of applications, including Web-based and client/server applications. The applications may include, for example, Skype, YouTube, Netflix, WebEx, Google Voice, BitTorrent, Citrix, virtual desktop, PCoIP, or any other application. The classification engine may be configured, for example, to identify generic protocols and perform heuristic analysis for encrypted protocols. The classifiers 18, 22 are configured to perform deep packet inspection (DPI), which provides the ability to look into the packet past basic header information so that the contents of a particular packet can be determined. 
  • In one embodiment, the classification information 26 transmitted from the controller 12 to the AP 14 includes tuple information for a flow (e.g., source IP address, destination IP address, source port, destination port, and protocol), application identifier (ID), and stateless DPI information. Stateless DPI information includes classification and sub-classification information (e.g., fixed or variable offset with a pattern or regular expression) and rules for applying policies on the sub-classified packets. The policies may include, for example, drop packet, mark a DSCP (Differentiated Services Code Point) value in the packet, or rate limit the traffic.

1 comment: