dimensional analysis of scalability of distributed systems n.
Download
Skip this Video
Loading SlideShow in 5 Seconds..
Dimensional analysis of scalability of distributed systems PowerPoint Presentation
Download Presentation
Dimensional analysis of scalability of distributed systems

Loading in 2 Seconds...

play fullscreen
1 / 12

Dimensional analysis of scalability of distributed systems - PowerPoint PPT Presentation


  • 85 Views
  • Uploaded on

Dimensional analysis of scalability of distributed systems. Erik Aurell joint work with Sameh El-Ansary, Per Brand, Seif Haridi. Solution 1. Solution 2. Model Analytically. Simulate. What is this talk about?. Problem : We want to understand the behavior

loader
I am the owner, or an agent authorized to act on behalf of the owner, of the copyrighted work described.
capcha
Download Presentation

Dimensional analysis of scalability of distributed systems


An Image/Link below is provided (as is) to download presentation

Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author.While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server.


- - - - - - - - - - - - - - - - - - - - - - - - - - E N D - - - - - - - - - - - - - - - - - - - - - - - - - -
    Presentation Transcript
    1. Dimensional analysis of scalability of distributed systems Erik Aurell joint work with Sameh El-Ansary, Per Brand, Seif Haridi

    2. Solution 1 Solution 2 Model Analytically Simulate What is this talk about? Problem: We want to understand the behavior of self-properties of structured overlay networks Hard, Complex Prohibitive Dimensional analysis!!

    3. Sameh says 1: Self-organization • Without central coordination do: • A network graph with “good” properties: Logarithmic diameter & arity • Dynamic load balancing of storage among nodes • Chord , Pastry, Tapestry, P-grid, CAN, Viceroy, Koorde, Kademlia, Naor, Ulysses, DKS, Skipnet, HyperCup … • Studied deeply!!

    4. Sameh says 2: Self-Repair • How to maintain the graph? How to replicate data on the graph? • Current argument: “Just stabilize periodically and eventually everything will be OK, believe me!!”

    5. The research question Is it possible to find intensive variables that describe the characteristics of structured overlay networks that deploy self-organization and self-repair policies?

    6. Static properties of Chord

    7. Extra avg look-up length vs density

    8. A self-repair-related intensive variable •  is the typical time between perturbations to the network (leave/join),  is the time between corrective action (per node • =P/ is the dimensionless ratio perturbation to correction

    9. At given P, dependence on /

    10. Is  an intensive variable? (3/3)

    11. Average look-up length as fct of 

    12. Conclusion • [Done] Identified good candidates for intensive variables: • Self-organization-related: density • Self-repair-related: Perturbation/Stabilization • [To be done] • Find more intensive variables • Increase the accuracy through better statistics • Look for phase transitions • Start making use of the analysis in adaptive algorihms