Prediction-based Prefetching to Support VCR-like Operations in Gossip-based P2P VoD Systems

Tianyin Xu, Weiwei Wang, Baoliu Ye Wenzhong Li, Sanglu Lu, Yang Gao

Nanjing University

Dislab, NJU CS

Nanjing University

2

Outline

Background P2P VoD streaming; Gossip-based systems; VCR-like interactive behavior.

Motivation

Solutions System architecture; Prefetching model;

Data scheduling; VCR-like operation support.

Performance Evaluation Conclusions

Dislab, NJU CS

Nanjing University

Background (1) P2P media streaming

Everyone can be a content producer/provider. Cache-and-relay mechanism: peers actively cache media contents and

further relay them to other peers that are expecting them.

3

* P2P live streaming is very successful!- CoolStreaming

(INFOCOM’05),- PPLive, Joost

Dislab, NJU CS

Nanjing University

Background (2)

P2P VoD streaming is challenging! Provide free access to any segment in the video at anytime by VCR-like operations.

VCR-like (Video Cassette Recorder) operations random seek, pause, fast forward/backward (FF/FB) For VCR-like operations, “jump” process is the most important.

• Most VCR-like operations can be implemented by “jump”.– random seek & pause: 1 jump; FF/FB: series of jump;

4Dislab, NJU CS

Nanjing University

Motivation (1)

How to support the “jump”?

Optimizing the index overlay to realize fast segment relocation• Jump => locate-and-download process;• Necessary, but far more sufficient.

Prediction-based Prefetching• Expect a zero jump delay;• Proactively prefetch segments that are likely to be requested by future VCR-

like operations;• Rely on prediction accuracy.

5

Question: Is the prediction feasible?

Dislab, NJU CS

Nanjing University

User Access Patterns (1)

User rarely view the movie from the beginning to the end. The total playing time of a user is quite limited and tends to be short. Because some popular segments (called highlights) attract more user

requests than non-popular segments.

Brampton et al., NOSSDAV-2007 Zheng et al., P2PMMS-2005

6Dislab, NJU CS

Nanjing University

User Access Patterns (2)

Probability distribution of object and segment popularity Log-normal distribution Zipf distribution

Brampton et al., NOSSDAV-2007 Yu et al., EUROSYS-2006

7Dislab, NJU CS

Nanjing University

User Access Patterns (3)

Fast Forward is more frequent than Fast Backward. Short Jump is more frequent than Long Jump.

Cheng et al., IPTPS-2007

Brampton et al., NOSSDAV-2007

8Dislab, NJU CS

Nanjing University

Motivation (2)

Our Objective: Effective Prediction-based Prefetching Scheme

Effective prediction model• Based on user access patterns

Easy to be integrated in current P2P VoD systems

Practical data scheduling

9Dislab, NJU CS

Nanjing University

System Architecture (1)

Solution 1: Let the server do prediction for each user [1] Pro: Server has large volumes of user viewing logs Con: poor scalability

Solution 2: Let the client exchange user logs and do prediction [2] Pro: scalable Cons: 1. lack of large volumes of user logs 2. high computing cost & training time

[1] Huang et al, “A User-Aware Prefetching Mechanism for Video Streaming”, WWW-2003

[2] He et al, “VOVO: VCR-Oriented Video-On-Demand in Large-Scale Peer-to-Peer Networks”, TPDS-2009

10

Our solution:Server side: offline pattern mining => prediction model

Peer side: lightweight online prediction

Dislab, NJU CS

Nanjing University

System Architecture (2) Take full advantage of tracker

Tracker has large volume of user viewing logs; Every node have to contact the tracker to join the system

• initiate its neighbor & partner list

11Dislab, NJU CS

Nanjing University

Prediction Approach: Overview

Frequent Sequential Pattern Mining PerfixSpan[1] : Mining Sequential Patterns Efficiently by Prefix-

Projected Pattern Growth.

Splitting Video Segments into Abstract States Mapping User Logs to Abstract States Construct Contingency Table (CCT) Model Utilization

[1] Pei et al., “Mining Sequential Patterns by Pattern Growth: The PrefixSpan Approach”, TKDE-2004. 12Dislab, NJU CS

Nanjing University

Prediction Approach (1)

13

111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111

111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111

Frequent Sequential Patterns

Dislab, NJU CS

Nanjing University

Prediction Approach (2)

Sequential patterns found may be overlapped? e.g. <1,2,3,4,5,6,7> and <5,6,7,8,9,10,11,2>

Splitting Approach Filter out the sub-patterns

• e.g. <1,2,3,4>,<1,2,3,4,5>,<1,2,3,4,5,6>,<1,2,3,4,5,6,7>

Scan over the remaining sequential patterns• Cut them into intervals without overlapping

- e.g. <1,2,3,4,5,6,7> and <5,6,7,8,9,10,11,2>[1,7],[8,12]

Take intervals not exist in the mined sequential patterns as separate intervals

Split the contiguous intervals into appropriate granularity intervals(States)• - MIN, MAX

14Dislab, NJU CS

Nanjing University

Prediction Approach (3)

Map Raw User logs into State Transitions <s,s’> e.g. <1,2,3,4,5,6,7,8,9,10> map to [1,6][7,13]

Transition Table Construction Simple Frequency Counting

15Dislab, NJU CS

Nanjing University

Data Scheduling

Two stage scheduling strategy: Stage 1: fetch urgent segments into playback buffer

• Guarantee the continuity of normal playback• Urgent line mechanism [1]

Stage 2: prefetch based on prediction• Reduce jump latency• Utilize residual bandwidth

[1] Li et al., “ContinuStreaming: Achieving High Plackback Continuity of Gossip-based Peer-to-Peer Streaming”, IPDPS-2008.

16Dislab, NJU CS

Nanjing University

VCR-like Operation Support

The jump process caused by VCR-like operations:

• Case 1. The jump segment is already prefetched on the local peer => Just playback!!

• Case 2. The jump segment is cached on the partners’ buffer => download and playback!

• Case 3. Neither cached on the local peer nor cached by the partners => relocate, connect and download

17Dislab, NJU CS

Nanjing University

Simulation Settings

User Log Generation• Modify GISMO [1]

– Using log-normal distribution to let users trend to jump around hot scenes.

The simulation is built on top of a topology of 5000 peer nodes based on the transit-stub model generated by GT-ITM.

The streaming rate is S = 256 Kpbs, the download bandwidth is randomly distributed in [1.5S, 5S].

The default size of the playback buffer is 30Mbytes, i.e., each peer can cache 120 second recent stream (100 for playback, 20 for prefetching).

The arrival of peers follows the Poisson Process with λ = 5.

[1] GISMO: A Generator of Internet Streaming Media Objects and Workloads

18Dislab, NJU CS

Nanjing University

Performance Evaluation (1)

19Dislab, NJU CS

Nanjing University

Performance Evaluation (2)

2

Dislab, NJU CS

Nanjing University

Performance Evaluation (3)

3

Dislab, NJU CS

Nanjing University

Performance Evaluation

4

Dislab, NJU CS

Nanjing University

Conclusions

A practical architecture that can be used in almost all existing P2P VoD systems

A novel and simple prediction approach State abstraction plays an important role

A two stage data scheduling

23Dislab, NJU CS

Nanjing University

24Dislab, NJU CS

The End