The Technology
TCAM devices are increasingly used for performing high-speed packet classification, which is an essential component of many networking applications, such as routing, monitoring and security. For packet classification, routers use a classification database, which contains rules (sometimes called filters). Each rule specifies a certain pattern, which is typically based on certain packet header fields, such as the source/destination addresses, source/destination port numbers and protocol type, and an action to apply to the packets that match the pattern. The patterns are stored as entries in the TCAM. Memory chips suffer from error events (often called soft errors), which may be caused by low-energy alpha particles, neutron hits, or cosmic rays. In a TCAM error event, a TCAM symbol can change its value to any symbol in {“0”, “1”, “*”}. The problem of soft errors in conventional random access memory (RAM) is typically handled by appending check symbols to the data according to some error-detecting or error-correcting code (ECC), and then applying an ECC check to each memory word upon access. This sort of solution is impractical in TCAM, however, since all entries are accessed in parallel. Soft errors in the TCAM device can cause various types of matching errors: An error in a TCAM entry may result either in the entry rejecting a search key, even though it should have returned a match (a false miss), or in an entry matching a search key that it should have rejected (a false hit). Matching a search key with the wrong entry can have adverse effects on the packet classification process, since different entries have different actions associated with them in the rule database. As an example, such an error may cause malicious packets that should have been dropped for security reasons to be erroneously forwarded, or vice versa. The technology is a method for error detection includes storing in an associative memory multiple data entries, each data entry including a data item together with one or more check symbols computed with respect to the data item. A predetermined sequence of search keys is applied to the memory, thereby causing the memory to generate, in parallel, match results with respect to the data entries. The match results are processed in order to identify an error in at least one of the data entries.
Advantages
An order of magnitude reduction in the number of lookups required for the error location process, thereby increasing performance and reliability.
Applications and Opportunities
Networking applications such as routing, monitoring and security.
Business Development Contacts
