Binary search on levels (BSOL) is a decision-tree algorithm for packet classification with superior speed performance. However, the most decision-tree-based algorithms, like BSOL, may suffer from a memory explosion problem caused by filter replications. In this work, we improve the storage performance of BSOL by employing a scheme, replication control. Our scheme dynamically generates multiple decision trees to eliminate filter replications in BSOL. The experimental results show that the new scheme achieves better performance than the existing decision-tree-based algorithms.

Packet classification is an enabling technology for various network services. Fast single-match packet classification can be achieved by using ternary content addressable memory (TCAM) because of the superior speed performance. TCAM has some drawbacks including incapability to store arbitrary ranges, confined TCAM capacity and limited choices of entry lengths. Moreover, TCAM only reports the first matching entry to impose a limitation on supporting multi-match packet classification, which requires all matching rules. The existing algorithms deal with the issues of TCAM-based multi-match packet classification by burdening TCAM with extra entries and/or accesses. In this work, we offload the overhead of TCAM to static random access memory (SRAM) to achieve efficient multi-match packet classification. Our scheme synthesizes TCAM compatible entries by using binary decision trees and employs SRAM for further comparisons. Each synthesized entry can be stored in one TCAM entry to significantly reduce TCAM consumption and fulfill low power consumption. The experimental results show that our scheme can lower the demand of TCAM to improve both search latency and energy efficiency. The scalability of TCAM-based multi-match packet classification can thus be improved drastically.

Ternary content addressable memory (TCAM) plays an important role in modern routers due to its capability of performing fast IP address lookup. However, it is expensive, space limited, and a major source of power consumption in a router. In addition, because TCAM only reports the first matching entry, updating TCAM entries would be slow due to necessary entry reordering. In this paper, we present a trie-based algorithm, longest suffix split, to reduce the number of TCAM entries for IP address lookup. The algorithm divides route prefixes into two portions, subprefix and suffix. The route prefixes with the same subprefix and similar suffix portions can then be represented by one TCAM entry and one SRAM entry. Each SRAM entry stores one of two succinct data structures, depending on the threshold number of similar suffixes. The experimental results show that our scheme can reduce 50% to 95% TCAM entries for the original routing tables. Our scheme also supports incremental updates. Because the drawbacks of TCAM are related to the number of required entries, our scheme significantly improves the feasibility of TCAM-based IP address lookup. While network virtualization may store multiple forwarding information bases in a router, the number of supported virtual routers can be increased by our scheme.

OpenFlow discovery protocol (OFDP) is the standard protocol for discovering network topologies for OpenFlow. Each controller generates and receives OFDP messages for each switch to yield a network topology. However, for a network with a large number of switches, the controller may suffer from high overhead to perform OFDP even though the network topology remains unchanged. In this paper, we improve the efficiency of network topology discovery by implementing a delegated function in OpenFlow switches. Our mechanism can reduce controller’s overhead for topology discovery to shorten the time of discovering a network topology. It also provides with the interoperability with switches which only perform OFDP. The experimental results demonstrate that the proposed scheme can improve the time for discovering a new network topology by decreasing the number of messages. Our scheme also eliminates unnecessary Packet-Out and Packet-In messages when the network topology remains unchanged. As a result, the controller overhead, including CPU usage and traffic load, is greatly reduced to shorten the response latency of controllers. Hence, the proposed scheme improves the feasibility of achieving scalable OpenFlow networking.