packet classification algorithms: from theory to practice
DESCRIPTION
Packet Classification Algorithms: From Theory to Practice. Author : Yaxuan Qi, Lianghong Xu, Baohua Yang, Yibo Xue, and Jun Li Publisher: IEEE INFOCOM 2009 Presenter: Fang-Chen Kuo Date: 2009/10/28. Related Work. RFC (SIGCOMM 1999) HSM (AINA 2005) HiCuts (MICRO 2000). RFC. - PowerPoint PPT PresentationTRANSCRIPT
1
Packet Classification AlgorithmsFrom Theory to Practice
Author Yaxuan Qi Lianghong Xu Baohua Yang Yibo Xue and Jun LiPublisher IEEE INFOCOM 2009Presenter Fang-Chen KuoDate 20091028
2
Related Work
RFC (SIGCOMM 1999) HSM (AINA 2005) HiCuts (MICRO 2000)
3
RFC
4
HSM (AINA 2005)
5
HiCuts(MICRO 2000)
6
HSM(In This Paper)
7
HiCuts(In This Paper)
8
Proposed HyperSplit
9
Strategies(Segment for Partition) Heuristic-1
Segment-Balanced decomposition Heuristic-2
Rule-Balanced decomposition Heuristic-3
Weighted Segment-Balanced decomposition
10
Strategies(Dimension for Partition) Heuristic-1 amp Heuristic-2
select the field with the largest (number of end-points) to apply space decomposition at each stage
Heuristic-3 select the field with minimum 10507511050751105075110507511050751
11
Performance(Metric)
With PC Memory Access Memory Requirement Pre-Processing Time
With Cavium OCTEON3860 network processor Throughput
12
Performance(Memory Access) (Bucket Size= 1)
13
Performance(Memory Access) (Bucket Size= 8)
14
Performance(Memory Requirement)
15
Performance(Pre-Processing Time)
16
Performance(Average Throughput)(Number of Cores)
17
Performance(Average Throughput)(Packet Size)
2
Related Work
RFC (SIGCOMM 1999) HSM (AINA 2005) HiCuts (MICRO 2000)
3
RFC
4
HSM (AINA 2005)
5
HiCuts(MICRO 2000)
6
HSM(In This Paper)
7
HiCuts(In This Paper)
8
Proposed HyperSplit
9
Strategies(Segment for Partition) Heuristic-1
Segment-Balanced decomposition Heuristic-2
Rule-Balanced decomposition Heuristic-3
Weighted Segment-Balanced decomposition
10
Strategies(Dimension for Partition) Heuristic-1 amp Heuristic-2
select the field with the largest (number of end-points) to apply space decomposition at each stage
Heuristic-3 select the field with minimum 10507511050751105075110507511050751
11
Performance(Metric)
With PC Memory Access Memory Requirement Pre-Processing Time
With Cavium OCTEON3860 network processor Throughput
12
Performance(Memory Access) (Bucket Size= 1)
13
Performance(Memory Access) (Bucket Size= 8)
14
Performance(Memory Requirement)
15
Performance(Pre-Processing Time)
16
Performance(Average Throughput)(Number of Cores)
17
Performance(Average Throughput)(Packet Size)
3
RFC
4
HSM (AINA 2005)
5
HiCuts(MICRO 2000)
6
HSM(In This Paper)
7
HiCuts(In This Paper)
8
Proposed HyperSplit
9
Strategies(Segment for Partition) Heuristic-1
Segment-Balanced decomposition Heuristic-2
Rule-Balanced decomposition Heuristic-3
Weighted Segment-Balanced decomposition
10
Strategies(Dimension for Partition) Heuristic-1 amp Heuristic-2
select the field with the largest (number of end-points) to apply space decomposition at each stage
Heuristic-3 select the field with minimum 10507511050751105075110507511050751
11
Performance(Metric)
With PC Memory Access Memory Requirement Pre-Processing Time
With Cavium OCTEON3860 network processor Throughput
12
Performance(Memory Access) (Bucket Size= 1)
13
Performance(Memory Access) (Bucket Size= 8)
14
Performance(Memory Requirement)
15
Performance(Pre-Processing Time)
16
Performance(Average Throughput)(Number of Cores)
17
Performance(Average Throughput)(Packet Size)
4
HSM (AINA 2005)
5
HiCuts(MICRO 2000)
6
HSM(In This Paper)
7
HiCuts(In This Paper)
8
Proposed HyperSplit
9
Strategies(Segment for Partition) Heuristic-1
Segment-Balanced decomposition Heuristic-2
Rule-Balanced decomposition Heuristic-3
Weighted Segment-Balanced decomposition
10
Strategies(Dimension for Partition) Heuristic-1 amp Heuristic-2
select the field with the largest (number of end-points) to apply space decomposition at each stage
Heuristic-3 select the field with minimum 10507511050751105075110507511050751
11
Performance(Metric)
With PC Memory Access Memory Requirement Pre-Processing Time
With Cavium OCTEON3860 network processor Throughput
12
Performance(Memory Access) (Bucket Size= 1)
13
Performance(Memory Access) (Bucket Size= 8)
14
Performance(Memory Requirement)
15
Performance(Pre-Processing Time)
16
Performance(Average Throughput)(Number of Cores)
17
Performance(Average Throughput)(Packet Size)
5
HiCuts(MICRO 2000)
6
HSM(In This Paper)
7
HiCuts(In This Paper)
8
Proposed HyperSplit
9
Strategies(Segment for Partition) Heuristic-1
Segment-Balanced decomposition Heuristic-2
Rule-Balanced decomposition Heuristic-3
Weighted Segment-Balanced decomposition
10
Strategies(Dimension for Partition) Heuristic-1 amp Heuristic-2
select the field with the largest (number of end-points) to apply space decomposition at each stage
Heuristic-3 select the field with minimum 10507511050751105075110507511050751
11
Performance(Metric)
With PC Memory Access Memory Requirement Pre-Processing Time
With Cavium OCTEON3860 network processor Throughput
12
Performance(Memory Access) (Bucket Size= 1)
13
Performance(Memory Access) (Bucket Size= 8)
14
Performance(Memory Requirement)
15
Performance(Pre-Processing Time)
16
Performance(Average Throughput)(Number of Cores)
17
Performance(Average Throughput)(Packet Size)
6
HSM(In This Paper)
7
HiCuts(In This Paper)
8
Proposed HyperSplit
9
Strategies(Segment for Partition) Heuristic-1
Segment-Balanced decomposition Heuristic-2
Rule-Balanced decomposition Heuristic-3
Weighted Segment-Balanced decomposition
10
Strategies(Dimension for Partition) Heuristic-1 amp Heuristic-2
select the field with the largest (number of end-points) to apply space decomposition at each stage
Heuristic-3 select the field with minimum 10507511050751105075110507511050751
11
Performance(Metric)
With PC Memory Access Memory Requirement Pre-Processing Time
With Cavium OCTEON3860 network processor Throughput
12
Performance(Memory Access) (Bucket Size= 1)
13
Performance(Memory Access) (Bucket Size= 8)
14
Performance(Memory Requirement)
15
Performance(Pre-Processing Time)
16
Performance(Average Throughput)(Number of Cores)
17
Performance(Average Throughput)(Packet Size)
7
HiCuts(In This Paper)
8
Proposed HyperSplit
9
Strategies(Segment for Partition) Heuristic-1
Segment-Balanced decomposition Heuristic-2
Rule-Balanced decomposition Heuristic-3
Weighted Segment-Balanced decomposition
10
Strategies(Dimension for Partition) Heuristic-1 amp Heuristic-2
select the field with the largest (number of end-points) to apply space decomposition at each stage
Heuristic-3 select the field with minimum 10507511050751105075110507511050751
11
Performance(Metric)
With PC Memory Access Memory Requirement Pre-Processing Time
With Cavium OCTEON3860 network processor Throughput
12
Performance(Memory Access) (Bucket Size= 1)
13
Performance(Memory Access) (Bucket Size= 8)
14
Performance(Memory Requirement)
15
Performance(Pre-Processing Time)
16
Performance(Average Throughput)(Number of Cores)
17
Performance(Average Throughput)(Packet Size)
8
Proposed HyperSplit
9
Strategies(Segment for Partition) Heuristic-1
Segment-Balanced decomposition Heuristic-2
Rule-Balanced decomposition Heuristic-3
Weighted Segment-Balanced decomposition
10
Strategies(Dimension for Partition) Heuristic-1 amp Heuristic-2
select the field with the largest (number of end-points) to apply space decomposition at each stage
Heuristic-3 select the field with minimum 10507511050751105075110507511050751
11
Performance(Metric)
With PC Memory Access Memory Requirement Pre-Processing Time
With Cavium OCTEON3860 network processor Throughput
12
Performance(Memory Access) (Bucket Size= 1)
13
Performance(Memory Access) (Bucket Size= 8)
14
Performance(Memory Requirement)
15
Performance(Pre-Processing Time)
16
Performance(Average Throughput)(Number of Cores)
17
Performance(Average Throughput)(Packet Size)
9
Strategies(Segment for Partition) Heuristic-1
Segment-Balanced decomposition Heuristic-2
Rule-Balanced decomposition Heuristic-3
Weighted Segment-Balanced decomposition
10
Strategies(Dimension for Partition) Heuristic-1 amp Heuristic-2
select the field with the largest (number of end-points) to apply space decomposition at each stage
Heuristic-3 select the field with minimum 10507511050751105075110507511050751
11
Performance(Metric)
With PC Memory Access Memory Requirement Pre-Processing Time
With Cavium OCTEON3860 network processor Throughput
12
Performance(Memory Access) (Bucket Size= 1)
13
Performance(Memory Access) (Bucket Size= 8)
14
Performance(Memory Requirement)
15
Performance(Pre-Processing Time)
16
Performance(Average Throughput)(Number of Cores)
17
Performance(Average Throughput)(Packet Size)
10
Strategies(Dimension for Partition) Heuristic-1 amp Heuristic-2
select the field with the largest (number of end-points) to apply space decomposition at each stage
Heuristic-3 select the field with minimum 10507511050751105075110507511050751
11
Performance(Metric)
With PC Memory Access Memory Requirement Pre-Processing Time
With Cavium OCTEON3860 network processor Throughput
12
Performance(Memory Access) (Bucket Size= 1)
13
Performance(Memory Access) (Bucket Size= 8)
14
Performance(Memory Requirement)
15
Performance(Pre-Processing Time)
16
Performance(Average Throughput)(Number of Cores)
17
Performance(Average Throughput)(Packet Size)
11
Performance(Metric)
With PC Memory Access Memory Requirement Pre-Processing Time
With Cavium OCTEON3860 network processor Throughput
12
Performance(Memory Access) (Bucket Size= 1)
13
Performance(Memory Access) (Bucket Size= 8)
14
Performance(Memory Requirement)
15
Performance(Pre-Processing Time)
16
Performance(Average Throughput)(Number of Cores)
17
Performance(Average Throughput)(Packet Size)
12
Performance(Memory Access) (Bucket Size= 1)
13
Performance(Memory Access) (Bucket Size= 8)
14
Performance(Memory Requirement)
15
Performance(Pre-Processing Time)
16
Performance(Average Throughput)(Number of Cores)
17
Performance(Average Throughput)(Packet Size)
13
Performance(Memory Access) (Bucket Size= 8)
14
Performance(Memory Requirement)
15
Performance(Pre-Processing Time)
16
Performance(Average Throughput)(Number of Cores)
17
Performance(Average Throughput)(Packet Size)
14
Performance(Memory Requirement)
15
Performance(Pre-Processing Time)
16
Performance(Average Throughput)(Number of Cores)
17
Performance(Average Throughput)(Packet Size)
15
Performance(Pre-Processing Time)
16
Performance(Average Throughput)(Number of Cores)
17
Performance(Average Throughput)(Packet Size)
16
Performance(Average Throughput)(Number of Cores)
17
Performance(Average Throughput)(Packet Size)
17
Performance(Average Throughput)(Packet Size)