6756

6298

كارشناسي ارشد

معماري كامپيوتر

اصفهان: دانشگاه صنعتي اصفهان، دانشكده برق و كامپيوتر

1390

سيزده، 100ص.: مصور، جدول، نمودار

ص.ع. به فارسي و انگليسي

حسين سعيدي

مسعودرضا هاشمي

22/3/91

نغمه سادات مويديان، علي فانيان

1396/10/06

كتابنامه

برق و كامپيوتر

مهندسي برق و كامپيوتر

ID6298

به فارسي و انگليسي: قابل رويت در نسخه ديجيتالي

Review and Improvement of the Network Packet Classification Algorithms Saeideh Alinezhad Chamazcoti s alinezhad@ec iut ac ir Date of Submission 2011 09 17 Department of Electrical and Computer Engineering Isfahan University of Technology Isfahan 84156 83111 Iran Degree M Sc Language Farsi Supervisor Hossein Saidi hsaidi@cc iut ac irAbstract At first a survey on packet classification problem and its available solutions is done in this thesis Packetclassification which has been widely deployed on the Internet is the core mechanism that enables routersto perform many networking services such as firewall packet filtering virtual private networks VPNs network address translation NAT quality of service QoS load balancing traffic accounting monitoring Integrated Sevices Int serv Differentiated services Diff serv and etc In packet classification we try tofind the best matching rule for an input packet among the rules of rule table In general five fields of eachpacket source IP Address destination IP Address Protocol source port and destination port are comparedwith the stored rules to find the best match The matching rule should match the packet in all fields There are two major groups of packet classification solutions hardware based methods and software based methods Software solution suffers from low speed so most of the recent work focused on thehardware solutions One of the most popular hardware solutions for classification is based on TCAM It provides high speed search capability in one clock cycle An introduction of TCAM its benefits anddrawbacks specifically in the packet classification problems are given and some methods used to solvethose problems are reviewed as well One of the major problems of TCAM is its high power consumptionwhich limits its performance drastically Another issue that is considered in recent articles about TCAM isproviding multi match packet classification Both of these issues are considered in this thesis We proposed a novel idea for low power packet classification with multi match output using TCAM Inthis approach we present a solution for the high power consumption problem in TCAM In this method thepacket classification process is done in two phases In the first phase which is in done during the offlinemode classification rules of the rule table are partitioned into specific groups partitions Then in thesecond phase which is done in online mode the decision about each received packet is made At first itdetermines the group that the packet belongs to and then the search operation continues within that specificgroup The group selection is done via the packet information By partitioning the rules storing eachpartition in a separate TCAM and searching in only one TCAM the power consumption will decreasewhich results to a power efficient packet classification Additionally as a part of the solution for multimatch packet classification our design finds all matches in one clock at the same time regardless of totalnumber of filters We have implemented our approaches using FPGA and compared the results with other availablesolutions The presented results show improvement in power consumption without significant hardwarecost in comparison to other solutions Additionally different practical and software optimization techniqueshave been discussed in details Keywords Low Power TCAM Multi Match Packet Classification Packet Classification RuleGrouping TCAM

حسين سعيدي

مسعودرضا هاشمي

نغمه سادات مويديان، علي فانيان