|M.Sc Student||Yankilevich Yevgeny|
|Subject||Fast and Efficient Soft Errors Detection and Correction|
in CAM and TCAM
|Department||Department of Electrical Engineering||Supervisors||Professor Ran Ginosar|
|Dr. Leonid Yavits|
|Full Thesis text|
The inherent parallel search ability of Content-Addressable Memory (CAM) and Ternary CAM (TCAM) enables using them in network devices for packet forwarding and classification. On the other hand, the simultaneous access to the entire content of the CAM and TCAM arrays during lookup operations makes it impossible to apply efficient data integrity algorithms such as used for RAMs. Several techniques for soft error detection and correction for CAM and TCAM have been proposed. However, they either require a change of the CAM architecture or are very time consuming, exposing the device to a high rate of soft errors.
We propose new algorithms for soft error detection and correction in CAM and TCAM. Our algorithms do not require any change to the CAM or TCAM architecture. They are faster, as they do not depend on the number of entries in the array, but rather on their width. They are more area efficient, as they require very little additional storage, while enabling detection and correction of any number of soft errors in the CAM and TCAM. An example implementation on a typical TCAM is described, demonstrating these benefits.
The algorithm was tested and implemented in industry Network Processor Chips of Intel and has U.S. Patent Application filed for it.