Packet classification using a bloom filter in a leaf-pushing area-based quad-trie

Abstract

Packet classification is one of the most essential functions that Internet routers should perform at wire-speed for every incoming packet. An area-based quad-trie (AQT) for packet classification has an issue in search performance since many rule nodes can be encountered in a search procedure. A leaf-pushing AQT improves the search performance of the AQT by making a single rule node exist in each search path. This paper proposes a new algorithm to improve the search performance of the leaf-pushing AQT further. The proposed algorithm builds a leaf-pushing AQT using a Bloom filter and a hash table stored in on-chip memories. The level of a rule node and a pointer to a rule database are identified by sequentially querying the Bloom filter and by accessing the hash table, respectively.

Cuckoo filter based IP packet filtering using M-tree

“…BSOL-RC [17] builds a decision trie with very high level so that it needs a separated decision trie for each field in the filter and a new trie level when ever the fields ranges are overlapped. DBAMCP [20] performs a sequential search to its integrated bloom filters and hash tables, also it need many hash tables that affects its search time. It can be noticed that using CF gives the proposed method an advantage over the DBAMCP [20], because it uses bloom filter, which is known to be slower than CF.…”

“…In their algorithm, they tried to minimize the number of replicated filters by performing a replication rule with the BSOL algorithm. Lim and Byun [20] build a packet classifier by integrating a bloom filter into an area based quad tree (BF-AQT). The internal nodes with rules have been pushed into leaves to get better performance.…”

“…processor, with a CPU @ 1.60 GHz 4 and 8 GB DDR3 RAM. A set of related modern method including (BSOL-RC [19], decomposition-based approach for scalable many-field packet classification on multi-core processors (DBAMCP) [20], and BF-AQT [21]) have been chosen to compare the proposed method results.…”

GenMatcher

“…If the keys are biased, a grouping with more trie nodes might perform better when the trie data structure is unbalanced. As needed, we can employ tree balance techniques [6,11,18,23,41,43] to rebalance the data structure. Thus, when the number of groupings is fixed, fewer trie nodes always perform better.…”

“…GenMatcher is able to process a larger key. For the larger key, we can utilize the single instruction multiple data (SIMD) techniques [21][22][23] to take advantage of wider instructions available in modern processors. To further improve search performance, we can extend this work to support a parallel [29,38] arbitrary matching architecture to process multiple tries in parallel.…”