ShanHsiang Shen Aditya Akella Outline Background and movtivation Challenge How DECOR works Implementation Evaluation results Why resource coordination Optimize network traffic Resources is limited ID: 257760
Download Presentation The PPT/PDF document "DECOR: A Distributed Coordinated Resourc..." is the property of its rightful owner. Permission is granted to download and print the materials on this web site for personal, non-commercial use only, and to display it on your personal computer provided you do not modify the materials and that you retain all copyright notices contained in the materials. By downloading content from our website, you accept the terms of this agreement.
Slide1
DECOR: A Distributed Coordinated Resource Monitoring System
Shan-Hsiang ShenAditya AkellaSlide2
Outline
Background and movtivationChallengeHow DECOR worksImplementationEvaluation resultsSlide3
Why resource coordination?
Optimize network traffic
Resources is limited
in different nodes
Resource coordinationSlide4
Why resource coordination?
Redundancy elimination: EndRE [Aggarwal’10]Node capacity (decoding): 5 Unit
4
Encode
decode
decode
Traffic: 10 Unit
Traffic: 10 Unit
Can only remove half of total redundancy
Still have resource available along the paths
Motivation 1:
Need coordination to use resources more efficientlySlide5
Centralized Resource Coordination
Encode
Decode
5
Traffic: 10 Unit
Traffic: 10 Unit
SmartRE
[
Anand’08
]
Node capacity (decoding): 5 Unit
Decode: 5 Unit
Decode: 5 Unit
Decode: 5 Unit
Decode: 5 Unit
A central control become bottleneck and a single point
of failure
Motivation 2:
A distributed frameworkSlide6
Challenge
Lack of a central controller to gather global view and make decision6
Local Optimization
Global OptimizationSlide7
DECOR Overview
7
Goal: assign job to each node to increase global benefit without using out resourcesSlide8
DECOR Overview
8
Optimize local benefit
Rearrange resource assignmentSlide9
Control Packets
Two kinds of control packets sent periodically:HELLO packets: collect accepted resources and traffic features along a path.ACK packets: deliver job assignment to each node.
HELLO
ACK
Ingress node
e
gress node
Calculate optimized job assignments
Accepted resources
9
Traffic feature
Resource arrangementSlide10
Resource Arrangement
Interior nodes assign resources to different paths.Calculate for each path.Assign more resources to the path that can contribute more benefit.
benefit
Unit resource
Ingress node
Interior
node
E
gress
nodeSlide11
Job assignment optimization
Optimize job assignment by linear programmingMaximize: total benefit along the path.Subject to: resources needed cannot surpass accepted resourcesVariable to be determined: the fraction of traffic each node needs to process
11
Ingress node
Interior
node
E
gress
nodeSlide12
Implementation
We implement distributed SmartRE (SmartRE + DÉCOR) in Click software router as modules.The Click modules is running in a desktop with Intel® Core™ 2 Quad CPU Q6700 and 3GB of memory.
12Slide13
The implementation issue of distributed SmartRE
An ingress node synchronizes cache with interior nodes by using buckets.
Ingress node
Interior node 1
Interior node 2
e
gress node
Bucket 1
interior node 1
Bucket 2
interior node 2
Bucket 3
egress node
13Slide14
Recovery from route failure
Stop using invalid buckets.
Ingress node
Interior node 1
Interior node 2
e
gress node
Bucket 1
interior node 1
Bucket 2
interior node 2
Bucket 3
egress node
Interior node 3
Bucket 1
interior node 1
Bucket 2
interior node 2
Bucket 3
egress node
Bucket 4
interior node 3
14Slide15
Evaluation setup
Compare distributed SmartRE with the following approaches.Hop-by-hop RE.Centralized SmartRE
.
Edge-based RE.
Ideal case.
Use a real trace with 2GB traffic.
15Slide16
Evaluation result
16Slide17
Convergence time evaluation
Topology (AS#)
PoPs
#Test Flows
#Iterations
NTT (2914)
70
34
6
Level 3 (3356)63
30
5
Sprint
(1239)
52
26
4
GEANT
22
10
3
Internet211
43
17Slide18
Other Applications
DECOR can apply to other path-based applicationsCSAMP [Sekar’08]Coordinate resources to sample trafficDECOR can provide a distributed solutionSlide19
Conclusion
Resource coordination to use resource more efficientlyDistributed solution to avoid bottleneck and single point of failure problemsDECOR can apply to multiple applicationsThe performance can be as good as centralized solution in SmartRE caseSlide20
Thank you
Question?Slide21
Backup slidesSlide22
CSAMP setup
NTT
Level3
Sprint
GEANT
Internet2
Flows(x
10
6
)514637
16
8Slide23
CSAMP setup
DECOR-based CSAMP: apply DECOR to coordinate resourcesFlow sampling: each node picks up one packet per 100 packets in each flowPacket sampling: each node picks up one packet per 100 packets of all trafficEdge packet sampling: edge nodes pick up one packet per 50 packets of all trafficMax sampling: each node samples as much as traffic as possibleSlide24
Evaluation resultsSlide25
Job assignment optimization
Bp,r is the maxima benefit path p
can provide in
node r
Distributed
SmartRE
:
B
p,r =
distancep,r × match
p,q,r
×
matchlen
p,q,r
Constraint:
Distributed
SmartRE
:Distributed SmartRE
: Maximize
25Slide26
Quota distribution
Limited resource quota in each node.Paths go through the node request resource quota with the node.The node arrange its resource according to the benefit each path can provide.
26Slide27
Job assignment
Multiple iterations are needed to converge.
Total: 5
Total: 5
HELLO
Total: 0
Total: 0
HELLO
5
HELLO
5
5
ACK
Assumption:
Both of paths need 5 unit resources in total
Path 2 can create more benefit for the network
Path 1
Path 2
27Slide28
Job assignment
Multiple iterations are needed to converge.
Total: 5
HELLO
Total: 0
Total: 0
HELLO
5
HELLO
5
ACK
Assumption:
Both of paths need 5 unit resources in total
Path 2 can create more benefit for the network
Path 1
Path 2
28Slide29
Job assignment
Multiple iterations are needed to converge.
HELLO
Total: 0
Total: 5
ACK
Assumption:
Both of paths need 5 unit resources in total
Path 2 can create more benefit for the network
Path 1
Path 2
29Slide30
Job assignment
Multiple iterations are needed to converge.
Total: 5
HELLO
Total: 0
Total: 0
HELLO
5
ACK
HELLO
5
5
Total: 5
Assumption:
Both of paths need 5 unit resources in total
Path 2 can create more benefit for the network
Path 1
Path 2
30Slide31
Job assignment
Multiple iterations are needed to converge.
Total: 5
HELLO
Total: 0
Total: 0
HELLO
5
ACK
Assumption:
Both of paths need 5 unit resources in total
Path 2 can create more benefit for the network
Path 1
Path 2
31Slide32
Redundancy-Aware R
outingIngress nodes direct traffic to get more benefit from inter-path redundancy.
Node 1
Node 2
Node 3
Route A
Route B
Route C
Route D