Redback Networks Inc.

Telefonaktiebolaget Lm Ericsson (publ)

Thomas Ng

Paul Terry

Renwei Li

Henry Lee

Desmond Yan

In one aspect, the present invention reduces average power consumption in a distributed processing system by concentrating an overall processing load to the minimum number of processing units required to maintain a defined level of processing redundancy. When the required number of active processing units is fewer than all available processing units, the inactive processing units may be held in a reduced-power condition. The present invention thereby maintains the defined level of processing redundancy for reallocating jobs responsive to the failure of one of the active processing units, while reducing power consumption and simplifying jobs allocation and re-allocation when expanding or shrinking the active set of processing units responsive to changing processing load. As a non-limiting example, the distributed processing system is implemented within a telecommunications network router or other apparatus having a configured set of processing cards, such as control-plane processing cards.

Method and apparatus for managing processing resources in a distributed processing system

Providing different levels of quality of service for different data flows being transported over a data link requires a very fast way to classify individual packets. Providing meaningful classification generally requires classifying data packets in several dimensions. A classification method represents a rule base as a hierarchy of linked tables and sequentially matches each parameter value in a packet signature against tables in the hierarchy. The method supports longest prefix matching and avoids time consuming backtracking by adding rules to the rule base.

Methods and apparatus for packet classification with multi-level data structure

Providing different levels of quality of service for different data flows being transported over a data link requires a very fast way to classify individual packets. Providing meaningful classification generally requires classifying data packets in several dimensions. A classification method processes multiple parameter values for a packet in parallel to obtain answer sets indicating which rules are matched by each parameter value. Answer indexes identify logical blocks in each answer set which contain TRUE values. The method performs an AND operation on the answer indexes to identify those blocks in the answer sets which could contain bits corresponding to matched rules. An AND operation is performed on the identified blocks to identify the particular rule matched by the packet. The classification method allows extremely fast wire speed packet classification.

Growing community of inventors

Richmond, CA, United States of America

Thomas Ng