Video Staging: A Proxy-Server-Based Approach to End-to-End Video Delivery over Wide-Area Networks

Zhi-Li Zhang, Yuewei Wang,

David H. C. Du, Dongli Su

IEEE/ACM TRANSACTIONS ON NETWORKING,

AUGUEST 2000

Outline Introduction Video Staging: A Single Video Case

– without Smoothing– with Smoothing– Empirical Evaluation

Video Staging: Multiple Video Case– Staging Hot Video Only(SHVO)– Largest Bandwidth Reduction Ratio First(LBRRF)– Empirical Evaluation

Conclusion

Introduction

Video Staging– Reduce the bandwidth requirement in the Backbo

ne WAN – Prefetch a predetermined amount of video data a

nd store them a priori at proxy servers– Only part of a video stream is retrieved directly fr

om the central video server across the backbone WAN

Video Staging(cont.)

– Trading the disk bandwidth of a proxy server for the backbone WAN bandwidth

– The video data are staged at the proxy server on a fairly long period of time instead of caching in and purged out dynamically

Video Staging:A Single Video Case

Video Staging Without Smoothing

i : the index of a video

Ni: total number of frames

F : frame period

sij : the size of jth frame j = 1,…, Ni

Pi : the peak rate of video i

Pi = (max1<=j<=Ni sij) /F

Ci : cut-off rate

0 <= Ci <= Pi*F = max1<=j<=Ni sij

Video Staging Without Smoothing

The upper part consists of a sequence of partial frames with size

sij ,u = ( si

j - Ci )+ j = 1,…, Ni

where x+ = max{x,0}

The lower part consists of a sequence of partial frames with size

sij ,l = si

j - ( sij - Ci )+ j = 1,…, Ni

Video Staging Without Smoothing

The smaller Ci is,the more video data is staged at a proxy server.

As Ci decreases, the lower part of the video becomes less bursty, and eventually approaches to an essentially CBR stream.

The backbone WAN bandwidth requirement from Pi to Ti = Ci / F

The upper part of the video consume

Di = (max1<=j<=Ni sij,u) / F disk bandwidth

Bandwidth reduction ratio Ri = (Pi - Ti) / Di

Video Staging With Smoothing

Cut-Off After Smoothing (CAS)si

~j : referred to as smoothed frames size

Pi : the peak rate of smoothed stream

Ci : cut-off rate

0 <= Ci <= Pi*F = max1<=j<=Ni si~j

Ti = Ci / F amount of backbone WAN bandwidth is reserved

Di = (max1<=j<=Ni si

~j,u) / F amount of disk bandwidth is required in the worst case

Video Staging With Smoothing

Cut-Off Before Smoothing– Smoothing on the Lower Part(SOLP)

Ti = Pi = (max1<=j<=Ni si~j,l) / F

– Smoothing on the Upper Part(SOUP)• Reduce the disk bandwidth required

– Smoothing on the Upper and Lower Parts(SOLP)

Bl = B * (Ci / Pi * F)

Bu = B * (1- (Ci / Pi * F))

Proxy server disk resource requirements for a

single stream

Resource requirements for a single stream

Ratio of backbone WAN bandwidth reduction to proxy server disk bandwidth

Video Staging: Multiple Video Case

Disk bandwidth Total amount of storage space The number of expected concurrent accesses Total reduction in the backbone WAN bandwidth

is maximized

Disk bandwidth constraint

Disk storage constraint

Staging Hot Video Only (SHVO) Regard the user access pattern as the most important fact

or in determining which video to stage at the proxy server. The videos are ordered with respect to their relative “pop

ularity” The number of k hottest videos that can be staged at the p

roxy server is determined

Either a video is entirely staged at the proxy server or not at all

Largest Bandwidth Reduction Ratio First (LBRRF) Use the backbone WAN bandwidth reduction ratio Ri i

n determining which video and what percentage of it to be staged

The burstier video and the video with larger number of concurrent accesses is more likely to have a higher Ri

CPi percentage of video i is staged at the proxy server The algorithm starts with CPi = 0 for all videos and iter

ates by incrementing CPi by P amount at each step.

Impact of proxy server disk resources:clients have no smoothing buffer

Impact of proxy server disk resources:clients have 64KB smoothing buffer

Conclusion

A major objective of our approach is to reduce the backbone WAN bandwidth requirement.

We have designed various video staging methods and evaluated their effectiveness in trading the disk bandwidth of a proxy server for the backbone WAN bandwidth.

The proposed proxy-server-based video staging technique provides a cost-effective and scalable solution to the problem of the end-to-end video delivery over WANs.