WARF TECHNOLOGIES

Information is transmitted through networks as data packets, which include information to help routers determine which action to perform on each packet. To implement services with the capability to distinguish and isolate the traffic of data packets in different flows, a packet classifier uses a set of predefined rules. Header fields for each packet include values such as source and destination addresses, protocol fields and port numbers, which if matched to a rule result in the corresponding action associated with the rule.

Packet classification with multiple header fields is a processor intensive operation, and needs to be performed at high speeds to keep up with the demand for Internet throughput.  “Smart rule caches” are used to balance the tradeoff between the memory usage and classification speed. A system and method is needed that provides efficient packet classification while requiring as little additional hardware as possible.

UW–Madison researchers have developed a computer-implemented system similar to a smart rule cache to classify received packets. The system uses a comparison of a hardware cache of evolving rules and a software cache of the original rule set. This allows for faster cache updating, searching algorithms and conflict resolution, thereby achieving significantly better packet classification performance than current rule cache computation algorithms.

• Implementation in software, hardware or a software/hardware combination for improved packet classification speed and capacity
• Possible use in any packet-based network including Internet Protocol (IP) networks, cellular networks and Asynchronous Transfer Mode (ATM) networks

• Improves network throughput by increasing packet classification speed
• Increases router performance by decreasing time required to re-compute cached rules

For More Information About the Inventors

• Suman Banerjee

Tech Fields

• Information Technology : Networking & telecommunications