Designing of SDN-Assisted Bandwidth and Latency Aware Route Allocation

04/13/2023 SWoPP 2014 1

Designing of SDN-Assisted Bandwidth and Latency Aware Route Allocation

Pongsakorn U-chupala,Kohei Ichikawa, Putchong Uthayopas,

Susumu Date, Hirotake Abe

04/13/2023 SWoPP 2014 2

Agenda

1. Introduction2. Approach3. Design4. Use Case5. Evaluation Plan6. Conclusion

04/13/2023 SWoPP 2014 3https://secure.flickr.com/photos/twicepix/4333178624

04/13/2023 SWoPP 2014 5

Spanning Tree Protocol

Unused!

04/13/2023 SWoPP 2014 6

Applications

Remote Desktop

Web ServerVoIP

Video Streaming

GameSocial Network

Secure Shell

04/13/2023 SWoPP 2014 7

Network Properties

Latency

Bandwidth

Distance

04/13/2023 SWoPP 2014 9

“Requirements”

Remote Desktop

Game

Applications NetworkProperties

LatencySecure Shell

BandwidthVideo Streaming

04/13/2023 SWoPP 2014 10

Path Diversity

Bandwidth

Latency

04/13/2023 SWoPP 2014 11

Objective

Align applications’ diverse requirements with different properties of each path in

the network and route accordingly

04/13/2023 SWoPP 2014 12

Prerequisites

• Deep packet inspection• Per-application network flow management

04/13/2023 SWoPP 2014 13

Agenda

1. Introduction2. Approach3. Design4. Use Case5. Evaluation Plan6. Conclusion

04/13/2023 SWoPP 2014 14

LatencyBandwidth

04/13/2023 SWoPP 2014 15

Categorizing Network Application

Bandwidth-Oriented• File Transfer• Computation of Big Data• Video/Audio Streaming

Latency-Oriented• VoIP• Game• Remote Desktop• Secure Shell

04/13/2023 SWoPP 2014 16

Link Dynamicity

• Cable type and condition• Congestion• Traffic engineering policies

Bandwidth?

Latency?

04/13/2023 SWoPP 2014 17

Path Dynamicity

Bandwidth?

Latency?

04/13/2023 SWoPP 2014 18

Direct Network Measurement

• Because of the dynamicity, calculating available resources by usage accounting is not practical

• Direct measurement yields more accurate result

04/13/2023 SWoPP 2014 19

Bandwidth and Latency Aware Routing

Monitor bydirect measurement

04/13/2023 SWoPP 2014 20

Agenda

1. Introduction2. Approach3. Design4. Use Case5. Evaluation Plan6. Conclusion

04/13/2023 24SWoPP 2014

Architecture1. OpenFlow Network2. Bandwidth and Latency (BW/LAT) Monitor3. BW/LAT Aware OpenFlow Controller4. BW/LAT Controller Supported Application

04/13/2023 SWoPP 2014 25

1. OpenFlow Network

• OpenFlow allows us to control route specifically for each application

• Centralize programmable controller allows us to aggregate information into a single location and reroute any connection as needed

OpenFlow Switch

OpenFlow Controller

04/13/2023 SWoPP 2014 26

2. Bandwidth and Latency Monitor• Overlord provides near-real-time available bandwidth

and current latency information of each link• Monitored information is forwarded to OpenFlow

controller through REST API

04/13/2023 SWoPP 2014 28

3. BW/LAT Supported Application• Each application register its preference to

BW/LAT controller through REST API• Preference information includes– Path identifier: source IP:port, destination IP:port– Preference: DEFAULT, MAX BW, MIN LAT

Path Identifier(src_ip, src_port, dst_ip, dst_port)

Preference(DEFAULT / MAX_BW / MIN_LAT)

(10.0.0.1, 1234, 10.0.0.2, 80) DEFAULT(10.0.0.1, *, 10.0.0.2, 80) MAX_BW(10.0.0.2, 80, 10.0.0.1, *) MAX_BW(10.0.0.1, *, 10.0.0.2, *) MIN_LAT(*, *, *, *) DEFAULT

04/13/2023 SWoPP 2014 30

4. BW/LAT Aware Controller

• Central to the design of BW/LAT network• Take– Available bandwidth of each link– Current latency of each link– Application preferences

• Pre-calculate possible routes for each pair of switches

• Then allocate route for each application accordingly

04/13/2023 SWoPP 2014 31

Route Pre-calculation

Minimum latency path and minimum hop count path is calculated with Dijkstra AlgorithmMaximum bandwidth path is calculated with a slightly modified version of Dijkstra Algorithm

04/13/2023 SWoPP 2014 32

04/13/2023 SWoPP 2014 33

Complexity Analysis

n-Times Dijkstra

Floyd-Warshall

[1] Ichikawa, K. and Abe, H.: A network performance-aware rout- ing for multisite virtual clusters, 19th IEEE International Confer- ence on Networks (ICON), Ieee, pp. 1–5 (online), DOI: 10.1109/I- CON.2013.6781935 (2013).

Similar calculation was done using Floyd-Warshall algorithm in Ichikawa et al. work [1]. However, complexity analysis shows that using n-Times Dijkstra is both simpler and more efficient.

04/13/2023 SWoPP 2014 34

OpenFlow Controller Framework

Simplicity

Documentation

Python

04/13/2023 SWoPP 2014 35

Route Allocation

BW/LAT Controller

Path Preference TableBW

LAT

HOP

(src, *, dst, *) | BW

04/13/2023 SWoPP 2014 36

Agenda

1. Introduction2. Approach3. Design4. Use Case5. Evaluation Plan6. Conclusion

04/13/2023 SWoPP 2014 37

Traditional Routing

04/13/2023 SWoPP 2014 38

BW/LAT Aware Routing

No congestion!Better performance!

04/13/2023 SWoPP 2014 41

Agenda

1. Introduction2. Approach3. Design4. Use Case5. Evaluation Plan6. Conclusion

04/13/2023 SWoPP 2014 42

Evaluation Plan

1. Emulation2. Experiment using Virtual Environment3. Real World Experiment with PRAGMA-ENT

Testbed

04/13/2023 SWoPP 2014 43

Emulation

• n-Switches mesh topology

• Randomly introduce latency and congestion

• Compare traditional routing and BW/LAT aware routing

• Measure average bandwidth and latency

04/13/2023 SWoPP 2014 44

Experiment using Virtual Environment

• Repeat selected cases from “Emulation” phase using virtual machines

• Measure real application performance– HTTP transfer speed representing bandwidth-

oriented application– Telnet communication response time representing

latency-oriented application

04/13/2023 SWoPP 2014 45

Real World Experiment withPRAGMA-ENT Testbed

• PRAGMA-ENT is a global-scale OpenFlow testbed that is still being developed [2]

• We wish to deploy BW/LAT aware routing on a production network and collect real usage performance statistics

[2] Scientific Expeditions - PRAGMA, , available from <http://www.pragma-grid.net/expeditions.php> (accessed 15/06/14).

04/13/2023 SWoPP 2014 46

Agenda

1. Introduction2. Approach3. Design4. Use Case5. Evaluation Plan6. Conclusion

04/13/2023 SWoPP 2014 47

Conclusion

• Aligning application requirements with network properties using OpenFlow

• Bandwidth-oriented and latency-oriented application

• Path property dynamicity• Bandwidth and latency aware routing• Design of BW/LAT network• Use case which our proposed solution excel• 3-phase evaluation strategy

04/13/2023 SWoPP 2014 49

THANK YOU FOR YOUR ATTENTIONQ&A