Multipath Routing - PowerPoint PPT Presentation

1 / 24
About This Presentation
Title:

Multipath Routing

Description:

appropriate ISP. Does smart routing improve performance? Does choice of ISPs matter? ... Enterprise Stand-in. Mar 1, 2004. Dhanashri Kelkar OGI School of ... – PowerPoint PPT presentation

Number of Views:326
Avg rating:3.0/5.0
Slides: 25
Provided by: kel86
Category:
Tags: does | for | isp | multipath | routing | stand | what

less

Transcript and Presenter's Notes

Title: Multipath Routing


1
Multi-path Routing
  • CSE 525 Course Presentation
  • Dhanashri Kelkar
  • Department of Computer Science and Engineering
  • OGI School of Science and Engineering

2
Multi-path Routing
  • A. Akella, B. Maggs, S. Seshan, A. Shaikh, R.
    Sitaraman, "A Measurement-Based Analysis of
    Multihoming", ACM SIGCOMM 2003.
  • D. Andersen, A. Snoeren, H. Balakrishnan,
    "Best-Path v. Multi-Path Overlay Routing", IMC
    2003.

3
Multihoming Advantages The Gist
  • A study of multihoming performance and
    reliability
  • Data collected from Akamai content distribution
    network
  • High-volume content providers
  • Enterprises that mainly receive data
  • Analysis
  • Improve performance and reliability
  • Choosing right set of providers important

4
Multihoming
  • Technique to achieve resilience to service
    interruptions
  • Customer network having more than one external
    link, either to single ISP or to different
    providers
  • Mainly used for reliability

5
K-Multihoming
  • Customer network multihomed to K (K2) service
    providers
  • Expect incremental performance

6
Multihoming Two Models
  • Enterprise perspective
  • Route data being downloaded through appropriate
    ISP
  • Web server perspective
  • Route data being provided through appropriate
    ISP
  • Does smart routing improve performance?
  • Does choice of ISPs matter?

7
Data Collection Enterprise Perspective2-Multiho
ming
  • Data set A1
  • 27 monitoring nodes
  • Two nodes per city connected to different ISP
  • Every 6 min. nodes download objects from Akamai
    customers
  • Log turnaround time for request

8
Data Collection Enterprise PerspectiveK-Multiho
ming (Kgt2)
  • Data set H1
  • Multiple Akamai servers per city
  • Each server connected to different ISP
  • Servers download from customers periodically
  • Log avg turnaround time each
    hour

9
Performance 2-multihoming
  • Use best provider for each download instead of
    single provider for all downloads
  • Performance metric
  • Measures how much each ISP loses compared to
    multihoming solution (1)

10
Performance K-Multihoming
  • Performance metric
  • particular K-multihoming solution
  • Best multihoming obtained if we choose best of
    all ISPs

11
Enterprise 2-Multihoming Results
  • 2-multihoming shows performance benefits but to
    varying degrees

12
Enterprise K-Multihoming Performance
  • Each line represents different city
  • No significant improvement after 4 or 5
  • Knowing best ISP in advance is important

13
Data Collection Web Server Perspective
14
Web Server Perspective Contd
  • Data set A2
  • In 5 metro areas, pick servers attached to
    distinct upstream ISPs
  • Every 6 min. each server downloads 50 KB object
    from other Akamai servers
  • Turnaround time for request

15
Web Server K-Multihoming
  • Use Akamai servers to emulate multihomed data
    centers and their active clients
  • Metric for comparison same as with enterprises
  • Not much benefit beyond K4

16
Reliability
  • Data set containing traceroute measurements from
    nodes of keynote systems to Akamai servers
  • 50 geographically diverse keynote nodes, 2 per
    city
  • 20 Akamai servers per city (top 20 ISP)
  • Information about IP-level connectivity
  • Robustness to IP-level failures

17
Reliability Metrics
  • Fraction of total path diversity captured by
    solution
  • Higher value shows better performance
  • Degree of overlap in paths
  • Lower value shows better performance

18
Reliability Analysis
  • For both metrics, significant difference in
    optimal, average, and worse solution
  • Difference about 80
  • Choosing ISPs very crucial

19
Conclusion
  • Multihoming helps, at least 20 improvement on
    average
  • But not much beyond 4 providers
  • Careful choice necessary
  • Cannot just pick top individual performers
  • Poor choice can affect performance significantly

20
Best-path vs. Multi-path Routing
  • Analysis of performance of reactive and mesh
    routing
  • Reactive routing measure path quality using
    probes and send on best path
  • Mesh routing send redundant duplicates

21
Design
  • Probe-based reactive overlay routing
  • Periodic probes for availability, latency, loss
    rate
  • Best path performance
  • Redundant multi-path routing
  • Sends redundant data to multiple paths
  • Path independence

22
Routing Methods
  • Direct Single packet, direct path
  • Direct direct 2 packets, direct, no spacing
  • DD 10ms 2 packets, direct, 10ms spacing
  • DD 20ms 2 packets, direct, 20ms spacing
  • Lat Reactive routing, min latency
  • Loss Reactive routing, min loss
  • Direct Rand 2 pkts, Redundant routing
  • Lat Loss 2 pkts, Redundant multi-path

23
Duplication Reduces Loss Rate
  • Type Loss
  • direct 0.42
  • direct direct 0.30
  • dd 10ms 0.27
  • dd 20ms 0.27
  • Lat 0.43
  • Loss 0.33
  • Direct Rand 0.26
  • Lat Loss 0.23

24
Measurement Summary
  • Redundant beats reactive for low loss
  • Reactive finds specific good paths
  • Latency improvements
  • Low loss paths
Write a Comment
User Comments (0)
About PowerShow.com