Packet classification is an essential function of QoS, MPLS and several network services. There are numerous various investigations have addressed this problem. This thesis attempts to improve the original bitmap intersection algorithm, which has memory explosion problem for large rule table. This study introduces the notion of bit compression to significantly decrease the storage requirement, creating what we called the CBV. Bit compression is based on the fact that ‘1’ bits are sparse enabling redundant ‘0’ bits to be removed. By region segmentation, the bit compression algorithm segments the range of dimension into CRs and then associates each CR with an index list. Merging rule sets reducing the number of CRs further. For rule table with wildcared rules, the bit compression propose a novel idea, “Don’t Care Vector” to save plenty storage space. The experiments for measuring maximum overlap led us to believe that plenty of redundant ‘0’ bits exist, such that removing ‘0’
bits can significantly improve memory storage.
Compared with bitmap intersection, the storage complexity is reduced from O( ) of bitmap intersection to O(dN·logN). In our experiment, our bit compression scheme only needs less than 380 Kbytes to store the 2-dimensional rule table with 10K rules, while bitmap intersection needs 48 Mbytes. Furthermore, comparing with memory access speed, our algorithm accesses average 96% less bits than bitmap intersection. Additionally, by exploiting the memory hierarchy to store the DCV, our bit compression scheme requires much less memory access time than bitmap intersection. Even though extra processing time for decompression is required for bit compression. The bit compression scheme still outperforms bitmap intersection scheme on the classification speed.
dN2
Future work on our study is to prove the practicability of our algorithm. We attempt to implement our algorithm on the hardware based platform, such as Intel IXP2400 or FPGA (Field Programmable Gate Array) to verify the practicability of bit compression algorithm.
A good packet classification engine also needs to quickly update the algorithm data structure to accommodate the changes in the rule table. In terms of fast update, bitmap requires pre-processing entire rule table each time when the rule changes.
However, since our bit compression algorithm segments dimension range into CRs, it is enough to update the bit vectors in CRs which overlapped with changed rule only.
In future work, we will investigate exhaustively the update policy of our algorithm.
Finally, we will extend our algorithm to support different field of the IP header. In the thesis, we consider the rule table only with layer-3 destination and source address fields. Bit compression algorithm performs well with these two layer-3 address fields.
However, does bit compression scheme adapt to other IP header fields? For example, bit compression scheme does not suit for layer-4 protocol type field. Because the protocol type field is restricted to a small set of exact values, such as TCP, UDP, ICMP, etc., the projection of rules in protocol type field are few, which means the probability of rule overlapping is high. Therefore, the maximum overlap in protocol type field would be large. It leads to performance degradation using bit compression scheme. In the future work, we will measure the maximum overlap of layer-4 source port and destination port from real-life rule database. By measure the maximum overlap, we can investigate the adaptation of the bit compression algorithm to layer-4 source port and destination port fields.
References
[1] M.H. Overmars and A.F. van der Stappen, “Range searching and point location among fat objects,” Journal of Algorithms, vol.21, no.3, pages 629-656, November 1996.
[2] V. Srinivasan, S. Suri, G. Varghese, and M. Waldvogel, “Fast and Scalable Layer four Switching,” Proceedings of ACM Sigcomm, pages 203-14, September 1998.
[3] V. Srinivasan, S. Suri, and G. Varghese, “Packet Classification using Tuple Space Search,” Proceedings of ACM Sigcomm, pages 135-46, September 1999.
[4] P. Gupta and N. McKeown, “Packet Classification on Multiple Fields,” Proc.
Sigcomm, Computer Communication Review, pages 147-160, September 1999.
[5] P. Gupta and N. McKeown, “Packet Classification using Hierarchical Intelligent Cuttings,” IEEE Micro, vol. 20, no. 1, pages 34-41, January/February 2000.
[6] F. Baboescu, S. Singh, G. Varghese, “Packet classification for core routers: is there an alternative to CAMs,” Proceedings of Infocom, vol.1 pages 53-63, March 2003.
[7] S. Singh, F. Baboescu, G. Varghese, and J. Wang, “Packet classification using multidimensional cutting,” Proceedings of SIGCOMM, pages 213-224, August 2003.
[8] T.V. Lakshman and D. Stiliadis, “High-Speed Policy-based Packet Forwarding Using Efficient Multi-dimensional Range Matching,” Proceedings of ACM Sigcomm, pages 191-202, September 1998.
[9] F. Baboescu, G. Varghese, “Scalable Packet Classification,” Proceedings of ACM Sigcomm, pages 199-210, August 2001.
[10] G. Zhang, H.J. Chao, J. Joung, “Fast packet classification using field-level trie,”
Global Telecommunications Conference, vol. 6, Pages 3201-3205 December 2003.
[11] D. Shah and P. Gupta, “Fast Incremental updates on Ternary-CAMs for routing
lookups and packet classification,” Proceedings of IEEE Hot Interconnects, page 145-153, August 2000.
[12] P. Tsuchiya, "A Search Algorithm for Table Entries with Non-contiguous Wildcarding," unpublished report, Bellcore, 1992.
[13] M. M. Buddhikot, S. Suri, M.Waldvogel, “Space Decomposition Techniques for Fast Layer-4 Switching,” Proceedings of conference on Protocols for High Speed Networks, page 25-41, August 1999.
[14] X. Sun and Y.Q. Zhao, “Packet classification using independent sets,”
Proceedings of ISCC, vol. 1, page 83-90, June/July 2003.
[15] W.T. Chen, S.B. Shin and J.L. Chiang. “A two-stage packet classification algorithm,” Proceedings of Conference on AINA, page 762-767, March, 2003.
[16] P. Gupta and N. McKeown, “Algorithms for packet classification,” IEEE Network, vol. 15, no. 2, page 24–32, 2001.
[17] M. A. Ruiz-Sanchex, E. W. Biersack, and W. Dabbous, “Survey and Taxonomy of IP Address Lookup Algorithms,” IEEE Network, vol. 15, issue 2, page 8-23, March/April 2001.
[18] M. Waldvogel, G. Varghesez, J. Turnerz and B. Plattnery, “Scalable High Speed IP Routing Lookups,” Proceedings of ACM Sigcomm, vol. 27 , issue 4, page 25-36, October 1997
[19] H. Liu, “Efficient Mapping of Range Rule table into Ternary-CAM,”
Proceedings of IEEE Hot Interconnects, page 95- 100, August 2002.
[20] “IXP1200 Network Processor: Datasheet”, Part No. 278298-010 December 2001.
[21] “IXP1200 Network Processor: Development Tools Use’s Guide”, Part No.
278302-010, March 2002.
[22] Mae-West routing database, “The Internet Performance measurement and analysis (IPMA) project”, http://www.merit.edu/ipma/routing_table, 2002.