cost-based cache replacement and server selection for multimedia proxy across wireless internet qian...
TRANSCRIPT
Cost-Based Cache Replacement and Server Selection for Multimedia Proxy Across Wireless
InternetQian ZhangZhe XiangWenwu ZhuLixin Gao
IEEE Transactions on Multimedia, August 2004
Outline
Introduction Multimedia Proxy Architecture Cost-Based Cache Replacement in the
Single-Server Environment Sever Selection in Multiple Server
Environment Simulation Conclusion
Introduction
Multimedia Proxy Architecture
Scalable Video Structure
Layer 0
Layer 1
Layer 2
Video Object
Key effectiveness of the proxy cacheReplacement Policy
Cache gain : Video object’s contribution of the proxy to achieve high performance
Cost-Based Cache Replacement
Performance metrics of traditional replacement algorithmHit ratio Byte hit ratio
Performance metrics of cost-based replacement algorithmVideo qualityNetwork cost savingStart-up delay
Cost-Based Cache Replacement
Multiple objectives for multimedia proxyDifferent portion of continuous media have
different characteristicsReduce the consumption of network resource
by reducing the throughput of Internet transmission between proxy and server
Cache a video prefix in order to reduce startup latency for streaming.
Different layer have different video-quality impact.
Video-Quality Revenue Rate
Rmse(v) [root-mean-square error] : represents the quality of video object.f (v) : request frequency of the video object
Redundant video object
Network-Saving Revenue RateObjective : reduce the consumption of network resource between proxy and serverThroughput Revenue Rate :
Size (v) : size of the video objectDist (RTT(v)) : distance between the server and proxy
Network-Utilization Revenue Rate :
μ(v, t) : network utilization for video v at time t
Startup Latency Revenue RateLatency between remote video server and proxy dominates end-to-end latency.
Prefix caching is an effective way to reduce startup latency for end clients.
Cost-Based Replacement PolicyTotal revenue rate of video object v
Total revenue rate of redundant video object vR
Cache Gain
Cost-Based Replacement Policy in Multiple Server
: integration network utilization for video v at time t
Video with fewer replicas gets a higher network-utilization revenue reate
(Single Server)(Multiple Server)
Server-Selection Algorithm
Multimedia proxy measures the potential revenue of each candidate video server for the request.
Server-Selection Algorithm
To perform load balance for consumed network resource
Selection Gain
To avoid the overload for the hot link while improving the utilization of the cold link
Simulation
Simulation
Simulation
Simulation
Simulation
Multimedia proxy over wireless Internet where the data cache and computational cache were introduced considering the different characteristics of wireless and Internet links.
For the single server case, novel cost-based replacement so as to improve all three aspects of performance, i.e., video quility.
For the multiservers case, new cost-based server-selection policy for multimedia proxy to improve the video quality.
Conclusion