Session based high bandwidth throughput testing. - PowerPoint PPT Presentation

010101010101101010101110110100010010110111101101100101011011000100101011010101101100011001101001001001110100101001010010101001001001 Session based high bandwidth throughput testing. 010101010101101010101110110100010010110111101101100101011011000100101011010101101100011001101001001001110100101001010010101001001001 RP2 System and Network Engineering System and Network Engineering Student: Bram ter Borch

Research Question 2 010101010101101010101110110100010010110111101101100101011011000100101011010101101100011001101001001001110100101001010010101001001001 010101010101101010101110110100010010110111101101100101011011000100101011010101101100011001101001001001110100101001010010101001001001 Student: Bram ter Borch

UDP vs TCP 3 010101010101101010101110110100010010110111101101100101011011000100101011010101101100011001101001001001110100101001010010101001001001 UDP TCP connectionless connection oriented 8 byte header 20 byte header no ordering ordering lightweight heavyweight reliable congestion control 010101010101101010101110110100010010110111101101100101011011000100101011010101101100011001101001001001110100101001010010101001001001 Student: Bram ter Borch

Research Question 4 What is needed to perform high bandwidth T h e C h a n n e l session based throughput tests and how to go beyond pure network infrastructure testing? 010101010101101010101110110100010010110111101101010101001001010010100100010010010100101010100101001010100101010010101001010100101010 100101010101010101001010100101001010101010101011100101011011000100101011010101101100011001101001001001110100101001010010101001001001 The term ”high bandwidth” references to at least 40Gb/s. The term ”session based” references to TCP traffic. 010101010101101010101110110100010010110111101101100101011011000100101011010101101100011001101001001001110100101001010010101001001001 Student: Bram ter Borch

S Test environment 5 Vlan100: 10.10.10.0/24 010101010101101010101110110100010010110111101101100101011011000100101011010101101100011001101001001001110100101001010010101001001001 40Gb/s 40Gb/s 40Gb/s 100Gb/s 100Gb/s A C 4x10Gb/s B D E 1Gb/s 1Gb/s 1Gb/s 1Gb/s 100Gb/s Vlan 66: 10.60.66.0/24 Internet 010101010101101010101110110100010010110111101101100101011011000100101011010101101100011001101001001001110100101001010010101001001001 MGMT Student: Bram ter Borch M

Tools 6 010101010101101010101110110100010010110111101101100101011011000100101011010101101100011001101001001001110100101001010010101001001001 MoonGen pktgen-kernel pktgen-DPDK BoNeSi 010101010101101010101110110100010010110111101101100101011011000100101011010101101100011001101001001001110100101001010010101001001001 Student: Bram ter Borch

First important result 7 010101010101101010101110110100010010110111101101100101011011000100101011010101101100011001101001001001110100101001010010101001001001 DPDK is the way to go 010101010101101010101110110100010010110111101101100101011011000100101011010101101100011001101001001001110100101001010010101001001001 Student: Bram ter Borch

Use cases 8 010101010101101010101110110100010010110111101101100101011011000100101011010101101100011001101001001001110100101001010010101001001001 NR Use case DUT Explanation Bandwidth The goal is to see if the client is capable of filling up the link UC1 Client generation and to reach the maximum amount of pps Generate the maximum amount of bidirectional data to make UC2 Throughput Switch/Router sure the intermediate hardware is able to forward 
 at line rate Session per UC3 Client/Server Get the hardware limitations of the end hosts. second Server and intermediate The clients will try to overload the server with requests at UC4 Application devices application level 010101010101101010101110110100010010110111101101100101011011000100101011010101101100011001101001001001110100101001010010101001001001 Student: Bram ter Borch

Experiment UC1 bandwidth 9 010101010101101010101110110100010010110111101101100101011011000100101011010101101100011001101001001001110100101001010010101001001001 010101010101101010101110110100010010110111101101100101011011000100101011010101101100011001101001001001110100101001010010101001001001 Student: Bram ter Borch

Experiment UC1 PPS 10 010101010101101010101110110100010010110111101101100101011011000100101011010101101100011001101001001001110100101001010010101001001001 010101010101101010101110110100010010110111101101100101011011000100101011010101101100011001101001001001110100101001010010101001001001 Student: Bram ter Borch

Experiment UC3 HTTP link usage 11 010101010101101010101110110100010010110111101101100101011011000100101011010101101100011001101001001001110100101001010010101001001001 010101010101101010101110110100010010110111101101100101011011000100101011010101101100011001101001001001110100101001010010101001001001 Student: Bram ter Borch

Experiment UC3 sessions 12 010101010101101010101110110100010010110111101101100101011011000100101011010101101100011001101001001001110100101001010010101001001001 010101010101101010101110110100010010110111101101100101011011000100101011010101101100011001101001001001110100101001010010101001001001 Student: Bram ter Borch

SURFnet Real world test 13 1 s 0 b / 0 G G 0 b /s 0 1 VCP 5x10Gb/s 010101010101101010101110110100010010110111101101100101011011000100101011010101101100011001101001001001110100101001010010101001001001 Router Router1A Router1B 4x10Gb/s 4x10Gb/s 4x10Gb/s AE113 AE112 Active - passive A P Firewall1 Firewall2 sync QFX10k2 40Gb /s AE0 AE0 4 40Gb /s 0 b /s 40Gb /s G b /s G 0 4 2x40Gb/s 2x40Gb/s 40Gb /s CLIENT DC1-1A DC1-1B DC2-1A DC2-1B 40Gb /s 40Gb /s 40Gb /s 40Gb /s 010101010101101010101110110100010010110111101101100101011011000100101011010101101100011001101001001001110100101001010010101001001001 Student: Bram ter Borch SERVER A

Real world result - bandwidth 14 010101010101101010101110110100010010110111101101100101011011000100101011010101101100011001101001001001110100101001010010101001001001 NGINX RAW TCP HTTP 010101010101101010101110110100010010110111101101100101011011000100101011010101101100011001101001001001110100101001010010101001001001 Student: Bram ter Borch

Conclusion 15 010101010101101010101110110100010010110111101101100101011011000100101011010101101100011001101001001001110100101001010010101001001001 DPDK is the way to go for high bandwidth session based throughput testing The use cases are valid The weakest link in a path can be found Monitoring is very important 010101010101101010101110110100010010110111101101100101011011000100101011010101101100011001101001001001110100101001010010101001001001 Student: Bram ter Borch

Future work 16 010101010101101010101110110100010010110111101101100101011011000100101011010101101100011001101001001001110100101001010010101001001001 DPDK tests have to be run using 100Gb/s interfaces IPv6 support has to be added to WARP Other layer 7 protocols have to be added to WARP 010101010101101010101110110100010010110111101101100101011011000100101011010101101100011001101001001001110100101001010010101001001001 Student: Bram ter Borch

010101010101101010101110110100010010110111101101100101011011000100101011010101101100011001101001001001110100101001010010101001001001 Questions Thank you 010101010101101010101110110100010010110111101101100101011011000100101011010101101100011001101001001001110100101001010010101001001001 Student: Bram ter Borch