reducing bandwidth requirement for delivering video over wide

Reducing Bandwidth Requirement for Delivering Video Over Wide Area

Networks With Proxy Server

Wei-hsiu Ma

David H.C.Du

IEEE TRANSACTIONS ON MULTIMEDIA,VOL,4,NO,4,DECEMBER 2002

Outline

• Introduction

• Background

• Frame Staging Selection Algorithms• Analytical

• Simulation Results

• Proxy Stream Consideration

• Conclusions

Introduction

• The proxy server stores portion of a video such that the average rate of the transport from the central server is reduced for this video.

• A proxy server can be considered as an intermediate node along the server-client path.

Introduction (cont.)

Background

• Video Transmission Constraints

B: Client buffer size

L(t): Consumption schedule or buffer underflow curve

U(t): buffer overflow curve

R(t): cumulative data amount sent by the video server

Background (cont.)

• A video consist N frames• Frame size fi for the ith frame

• F is the set of frame sizes• L(t) = • •

Background (cont.)

• CBR Transmission Without Proxy

Background (cont.)

Frame Staging Selection Algorithms

• Assume– The synchronization at the client.– The transmission on WAN is CBR with

minimum delay and without data loss.

– The LAN bandwidth is fast enough to handle the local delivery in VBR.

– The video data stored in the proxy is prefetched.

Frame Staging Selection Algorithms (cont.)

Frame Staging Selection Algorithms (cont.)

• Just-In-Time Algorithm– To decide which frames are staged at the

proxy.

• Minimum Difference Algorithm– To select a frame before frame I with the

smallest possible size to be staged in the proxy.

• Chunk Algorithm

Chunk Algorithm

Analytical

• Lower bounds of client buffer

Analytical (cont.)

• Lower bounds of client buffer is

Analytical (cont.)

• Lower bounds of the proxy storage is

Sp : The amount of data stored in the proxy

Svideo : The size of video

Sc : The amount of data sent by the central server

Simulation Results

Simulation Results (cont.)

Simulation Results (cont.)

Proxy Stream Consideration

Buffer complementary effect

• B’ is the necessary buffer size:

bufc(t): The buffer occupancy functions for central streams

bufp(t): The buffer occupancy functions for proxy streams

• Good complementary effect will make B’ close to max{Bc,Bp} instead of Bc + Bp.

Client Buffer Partition

Client Buffer Partition

• Incremental Equal-Division Buffer Assignment (IEBA) approach to approximate the best partition for Bc and Bp.

• The basic idea is to assign Bc and Bp equally if possible and gradually increase both values at the same pace until B’ is closest to B.

Client Buffer Partition (cont.)

•IEBA:

Client Buffer Partition (cont.)

Conclusions

• Chunk algorithm can reduce the number of chunks with client buffer and proxy storage requirement close to the lower bounds of these resource.

• After using chunk algorithm, the staged data are delivered in VBR rate and buffer complementary effect is found so that buffer can be better utilized for both central and proxy streams.