1 / 25

A New Approach for the Construction of ALM Trees using Layered Coding

A New Approach for the Construction of ALM Trees using Layered Coding. Yohei Okada, Masato Oguro, Jiro Katto Sakae Okubo International Conference on Autonomic and Autonomous Systems and International Conference on Networking and Services (ICAS/ICNS 2005). Outline. Pre-knowledge

zuriel
Download Presentation

A New Approach for the Construction of ALM Trees using Layered Coding

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. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. A New Approach for the Construction of ALM Trees using Layered Coding Yohei Okada, Masato Oguro, Jiro Katto Sakae Okubo International Conference on Autonomic and Autonomous Systems and International Conference on Networking and Services (ICAS/ICNS 2005)

  2. Outline • Pre-knowledge • Issue of ALM with Layered Coding • Proposal System • Evaluation & Implementation • Conclusion

  3. Pre-knowledge • Multicast can save bandwidth • The same data don’t need to transmit redundant • IP layer Multicast /Application Layer Multicast • IP layer Multicast • Need router supporting on internet • Slow due to the complexity of route control • Issue of Multicast • Join/leave problem • Load balance problem

  4. Issue:Join/leave problem • How a New participant find a proper parent? • How to define “proper”?? • When a host leave , how to rejoin the node’s sub-tree? • recover overhead

  5. Issue:Layered Coding • Layered Coding • The more bandwidth you have, the better quality you get! • ALM with layered coding • Low position host can’t receive data more than its ancestor • Conflict with above! • Need to find a “good” parent.

  6. Proposal System • Formal definition of “degree” • Fi:the forwarding capacity of node i • R1:the forwarding base layer steaming rate • Di:degree of node i • Accumulative rate of each layer’s stream • shown by real number as the ratio to R1 • Out-degree • Di = Fi / R1 • In-degree • The wanted receive rate of host • The presentation is the same as out-degree

  7. 100 50 50 50 50 50 50 Proposal System Layered coding rate 100 accumulative rates 1:2:4 50 50 Ex: A host i‘s sending capacity is 300 (kbps) the degree of the host is 6 6 50 50 50 4 base and 1 2layer 1 3layer and 1 2layer 50

  8. Proposal System • “degree” parameter • Present the children capacity ability • Known in advance • New participant can look for proper parent depend on it • Refining when node join/leave • Constrain: • A host which is able to forward little data should not receive much. Otherwise, other hosts have less streams in return for satisfaction of the host. • out-degree ≥ in-degree (at each host)

  9. Proposal System • “Join” Principle • Participants find the proper parent • it should be able to decide the host at the same time when joining the tree. • Parent’s out-degree ≥ child’s in-degree • To satisfy each host’s in-degree, its parent must have out-degree which is bigger than in-degree of the child.

  10. Proposal System • “Join” Principle • Find not only a parent but its own child • a participant joins the tree not only as a leaf host but also as an intermediate host • Adopt “delay” as the second metric • To avoid the problem of bandwidth waste by connecting • RTT is used

  11. Proposal System • Process of constructing ALM tree • Step.1 Send a join-request massage to the source • New participant host N sends a join-request massage to the source with its in and out degree information. • Step.2 Search for candidates of the parent host • The host receiving the join-request becomes the parent candidate of new host when each of the following two conditions is satisfied at least. N’s in-degree ≤ its (remaining) out-degree N’s out-degree ≥ out-degree of its children hosts (host that has big out-degree should be in higher-position of the tree) • Otherwise, it forwards the request message to its children

  12. Proposal System • Process of constructing ALM tree • Step.3 Decide the parent • The candidates send response messages to host N. Then, host P which has minimum RTT is decided to be the parent host of N. • Step.4 Exchange link connections locally • If host N becomes a intermediate node , parent P needs to exchange connections in the local area which includes hosts from P to P’s grand-children based on the degree and N can join any host of them.

  13. Proposal System

  14. Proposal System Condition 1 Condition 2 Problem: The receive layer may less than output streaming layer A constrain should be added “N’s in-degree ≤ parent’s in-degree”

  15. Proposal System • “Leave” Principle • Node departures are of two kinds: • Graceful departure • The descendants can decide new parents that have unused out-degree beforehand and reconnect to them seamlessly. • Sudden failure • Due to a computer crashing or congestion of the network links, etc. • The descendants can’t prepare for the reconnection. • The descendant hosts needs to find a new parent • The simplest way is to rejoin the source • this takes too much recovery time and overhead.

  16. Proposal System • “Leave” Principle • All Hosts memorize their parent candidates • Found at Step 2 of tree construction • When it needs to rejoin the tree • reconnect directly • descendants only execute Step4 • shorten the recovery time and overhead. Problem: The record may not match condition now!!!

  17. Evaluation & Implementation • Evaluation environment • ns-2 simulator • http://netlab.cse.yzu.edu.tw/ns2/ • Layered coding in {125, 125, 250} (kbps) • Streaming protocol is UDP • The number of hosts of which degrees are assigned randomly is changed from 20 to 100 • Measure • average throughput of all hosts • average delay of each link • the number of control packets (overhead).

  18. Evaluation & Implementation • Evaluation under two system • RTT method • Round-Robin (RR) method

  19. Evaluation & Implementation • The importance of the in-degree restriction • When in-degree is not limited, the total throughput decreases as the # of nodes increases. • Because the hosts which have small out-degree are located in a higher-position of the tree.

  20. Evaluation & Implementation • All hosts have proper parents and receive at the rate which they desire by using the redefined degree parameter.

  21. Evaluation & Implementation • The proposal doesn’t necessarily sacrifice the delay and overhead.

  22. Evaluation & Implementation • Compared the two recover method • Memorize their parent candidates • Rejoin the source

  23. Evaluation & Implementation • Implement for proposal • observe the time lag caused by tree reconstruction. • Video stream is encoded by H.263+ and the layering is carried out in a temporal scalable manner, by simply splitting I-picture packets and P- picture packets. • 10 frames per second, and 2 I-pictures per second. The average latency until I-picture arrives is 0.25 sec (min is 0 sec, max is 0.5 sec).

  24. Evaluation & Implementation • the implementation are almost similar to those of the simulation.

  25. Conclusion • A new approach to construct ALM trees for layered video stream is described. By adopting newly defined “degree” and constructing trees according to those Parameters. • The method can be directly applied to the system using Layered-MDC (Multiple Description Coding) , which is very flexible for network congestion control and heterogeneity of receiving environment.

More Related