Douglis / Davison Web Content Caching and Distribution

Proceedings of the 8th International Workshop
1. Auflage 2007
ISBN: 978-1-4020-2258-6
Verlag: Springer Netherlands
Format: PDF
Kopierschutz: 1 - PDF Watermark

E-Book, Englisch, 363 Seiten

Web caching and content delivery technologies provide the infrastructure on which systems are built for the scalable distribution of information. This proceedings of the eighth annual workshop, captures a cross-section of the latest issues and techniques of interest to network architects and researchers in large-scale content delivery. Topics covered include the distribution of streaming multimedia, edge caching and computation, multicast, delivery of dynamic content, enterprise content delivery, streaming proxies and servers, content transcoding, replication and caching strategies, peer-to-peer content delivery, and Web prefetching.

Web Content Caching and Distribution encompasses all areas relating to the intersection of storage and networking for Internet content services. The book is divided into eight parts: mobility, applications, architectures, multimedia, customization, peer-to-peer, performance and measurement, and delta encoding.

Autoren/Hrsg.

Douglis, Fred

Davison, Brian D.

Inhaltsverzeichnis

1;Contents;6
2;A Message from the Workshop Chairs;10
3;Credits;12
4;Contributing Authors;14
5;Mobility-aware server selection for mobile streaming multimedia content distribution networks;18
5.1;1. Introduction and Overview;18
5.2;2. Mobility Based Server Selection;19
5.2.1;2.1 Layout of servers in content distribution network;20
5.2.2;2.2 Considerations in mobility based server selection;21
5.2.3;2.3 Measurement of mobility rate and server residence time estimation;21
5.2.4;2.4 Server load and QoS information collection;22
5.2.5;2.5 Server selection algorithm;23
5.3;3. Simulation Setup;25
5.3.1;3.1 Mobility simulation;26
5.3.2;3.2 CDN layout;27
5.3.3;3.3 Simulation scenarios;28
5.4;4. Results;30
5.5;5. Related Work;33
5.6;6. Conclusions;33
5.7;References;34
6;Performance of PEPs in cellular wireless networks;36
6.1;1. Introduction;36
6.2;2. RelatedWork;37
6.3;3. Overview of Cellular Networks;38
6.4;4. Wireless PEP;39
6.5;5. Latency Components;40
6.6;6. Transport/Session Optimizations;41
6.6.1;6.1 TCP Tuning;42
6.6.2;6.2 TCP Connection Sharing;45
6.6.3;6.3 Number of TCP connections;46
6.6.4;6.4 Temporal Block Flow Release;48
6.6.5;6.5 Session-level overheads: DNS;49
6.7;7. Application Level Optimizations;50
6.7.1;7.1 Compression Results;50
6.7.2;7.2 Acceleration Results;51
6.7.3;7.3 Impact of Pipelining;52
6.8;8. Comparison;53
6.9;9. Conclusions;54
6.10;References;55
7;Edge caching for directory based Web applications: Algorithms and performance;56
7.1;1. Introduction;56
7.2;2. Notations;58
7.3;3. LDAP Caching Framework;59
7.4;4. LDAP Query Containment;60
7.4.1;4.1 Query containment problem;60
7.4.2;4.2 General filter containment;60
7.4.3;4.3 Template based filter containment;61
7.4.4;4.4 Query containment algorithm;62
7.5;5. LDAP Caching Algorithms;63
7.6;6. Directory Server Extensions;65
7.7;7. Application Offload and Prefetching;66
7.7.1;7.1 Directory application modeling: Example;67
7.7.2;7.2 Prefetching;68
7.8;8. Performance of Caching Algorithms;68
7.9;9. Conclusions;71
7.10;References;72
8;Computing on the edge: A platform for replicating Internet applications;74
8.1;1. Introduction;74
8.2;2. Issues;75
8.3;3. Architecture Overview;76
8.4;4. Application Distribution Framework;77
8.4.1;4.1 The metafile;77
8.4.2;4.2 Replica creation;78
8.4.3;4.3 Replica deletion;79
8.4.4;4.4 Consistency maintenance;79
8.5;5. Algorithms;80
8.5.1;5.1 Content placement algorithm;81
8.5.2;5.2 Request distribution algorithm;83
8.6;6. Performance;85
8.6.1;6.1 Request distribution;85
8.6.2;6.2 Content placement;86
8.6.3;6.3 Redeployment threshold;89
8.7;7. RelatedWork;91
8.8;8. Conclusions;93
8.9;Acknowledgments;93
8.10;References;93
9;Scalable consistency maintenance for edge query caches: Exploiting templates in Web applications;96
9.1;1. Introduction;96
9.2;2. Semantic caching over theWeb;97
9.2.1;2.1 DBProxy overview;97
9.2.2;2.2 Common Local Store;98
9.3;3. Consistency Management;99
9.3.1;3.1 Update propagation approaches;100
9.4;4. Basic filtering;100
9.5;5. Template-based filtering;102
9.5.1;5.1 Template-based filtering: Single cache case;102
9.5.2;5.2 Template-based filtering: Multiple caches;105
9.6;6. Related work;106
9.7;7. Conclusions;106
9.8;References;107
10;Proxy+: Simple proxy augmentation for dynamic content processing;108
10.1;1. Introduction;108
10.2;2. Related Work;109
10.3;3. Summary of Previous Result;110
10.4;4. Proxy+ Architecture;111
10.4.1;4.1 ASP.NET output caching;112
10.4.2;4.2 Cache key generation;113
10.4.3;4.3 Tag generation and fragment caching;114
10.4.4;4.4 Cache keys notification and page composition;117
10.4.5;4.5 Summary of the protocol;119
10.5;5. Application Modifications;120
10.6;6. Experimental Results;122
10.7;7. Security aspect;124
10.8;8. Conclusions;124
10.9;References;125
11;Multicast cloud with integrated multicast and unicast content distribution routing;126
11.1;1. Introduction;126
11.2;3. Multicast Cloud;129
11.3;4. Integrated Channel Routing;130
11.4;5. Application-layer Traffic Control;132
11.5;6. Related Work;133
11.6;7. Conclusion;134
11.7;8. References;135
12;A large enterprise content distribution network: Design, implementation and operation;136
12.1;1. Introduction;136
12.2;2. Service Perspective and Network Infrastructure;137
12.3;3. CDN Architecture;138
12.3.1;3.1 Redirection System;140
12.4;4. CDN Operation and Lessons Learned;142
12.5;5. Related work;144
12.6;6. Conclusion and Future work;144
12.7;References;145
13;Architectural choices for video-on-demand systems;146
13.1;1. Introduction;146
13.2;2. Contribution and RelatedWork;147
13.3;3. The distribution network;148
13.4;4. Evaluation of Architectural Choices;148
13.4.1;4.1 Analytical model;148
13.4.2;4.2 Set-top box at the client side for pre.x storage;149
13.4.3;4.3 Use of satellite for suf.x transmission;152
13.5;5. Conclusion and FutureWork;153
13.6;References;155
14;Dynamic cache reconfiguration strategies for a cluster-based streaming proxy;156
14.1;1. Introduction;156
14.2;2. Architecture of cluster-based streaming proxy;158
14.3;3. Optimal Cache Placement;159
14.3.1;3.1 Optimal Cache Placement: Problem Formulation;159
14.3.2;3.2 Cache Placement Heuristics;160
14.4;4. Dynamic Cache Recon.guration;161
14.4.1;4.1 MWPM cache recon.guration;162
14.4.2;4.2 Swapping-based cache recon.guration;162
14.5;5. Performance Evaluation;164
14.5.1;5.1 Simulation setting;165
14.5.2;5.2 Evaluation of cache placement algorithms;165
14.5.3;5.3 Evaluation of cache reconfiguration algorithms;168
14.6;6. RelatedWork;172
14.7;7. Conclusions and future work;173
14.8;References;173
15;Stream engine: A new kernel interface for high-performance Internet;176
15.1;1. Introduction;176
15.2;2. Characteristics of Internet Streaming Protocols;177
15.3;3. Implementation Experience;178
15.3.1;3.1 Initial prototype;178
15.3.2;3.2 Identifying performance bottlenecks;179
15.4;4. The Stream Engine Interface;181
15.4.1;4.1 Definitions of a stream engine;181
15.4.2;4.2 Performance improvements from stream engine;183
15.5;5. Implications for TCP Offloading Hardware;183
15.6;6. RelatedWork;185
15.7;7. Conclusion and FutureWork;186
15.8;References;186
16;Streaming flow analyses for prefetching in segment-based proxy caching to improve delivery quality;188
16.1;1. Background and Motivation;188
16.2;2. Prefetching Methods for Segment-based Proxy Caching;190
16.2.1;2.1 Look-ahead window based prefetching method;191
16.2.2;2.2 Active prefetching method;193
16.3;3. Segment-based Proxy Caching Strategies with Least Proxy Jitter;196
16.3.1;3.1 Minimum number of segments cached for proxy jitter free;196
16.3.2;3.2 Trade-off between low proxy jitter and high byte hit ratio;197
16.4;4. Performance Evaluation;197
16.4.1;4.1 Workload summary;198
16.4.2;4.2 Exponential segmentation strategy;199
16.4.3;4.3 Uniform segmentation strategy;200
16.5;5. Conclusion;202
16.6;References;202
17;Subscription-enhanced content delivery;204
17.1;1. Introduction;204
17.2;2. Subscription-based content delivery;206
17.2.1;2.1 Access-based caching: the baseline approach;207
17.2.2;2.2 Single cache and single replacement method;207
17.2.3;2.3 Dual-caches approaches;209
17.3;3. Simulator and workload;210
17.3.1;3.1 Simulator and workload for news delivery;210
17.3.2;3.2 Generating subscription information;211
17.3.3;3.3 Tagging accesses as notification- driven or not;212
17.4;4. Experimental results;212
17.4.1;4.1 Metric and experimental setup;212
17.4.2;4.2 Comparing RSG2 and HUG;213
17.4.3;4.3 Behavior of DC- TA;214
17.4.4;4.4 Comparing SG1, HUG and DC;216
17.5;5. Related work;218
17.6;6. Conclusion;220
17.7;References;220
18;Cooperative architectures and algorithms for discovery and transcoding of multi-version content;222
18.1;1. Introduction;222
18.2;2. Related work;224
18.3;3. Main features of the intermediate infrastructure;225
18.4;4. Hierarchical topologies;226
18.5;5. Flat topologies;227
18.5.1;5.1 Cooperative discovery;227
18.5.2;5.2 Cooperative transcoding algorithms;228
18.6;6. Workload model;230
18.7;7. Experimental results;231
18.7.1;7.1 Comparison of the architectures;231
18.7.2;7.2 Cooperative transcoding algorithms;235
18.8;8. Conclusions;237
18.9;Acknowledgements;237
18.10;References;237
19;User specific request redirection in a content delivery network;240
19.1;1. Introduction;240
19.2;2. User Specific Request Redirection;241
19.2.1;2.1 Service differentiation based on user specific information;241
19.2.2;2.2 IP address based redirection;242
19.3;3. Implementing user specific request redirection;243
19.4;4. Request Defection;244
19.4.1;4.1 Step 1: A priori transformation of embedded URLs;244
19.4.2;4.2 Step 2: On-the-fly HTTP response header modification;245
19.4.3;4.3 Wire-speed request de.ection;246
19.4.4;4.4 Resolving server IP addresses;247
19.5;5. Conclusions;248
19.6;References;249
20;Friendships that last: Peer lifespan and its role in P2P protocols;250
20.1;1. Introduction;250
20.2;2. Background;251
20.2.1;2.1 Related work;252
20.3;3. Peer Lifespan Distribution;253
20.3.1;3.1 Collecting observed peers’ lifespans;253
20.3.2;3.2 Peer lifespan distribution;254
20.4;4. Peer Lifespan and P2P Protocols;256
20.4.1;4.1 Lifespan-based friend selection;256
20.4.2;4.2 Lifespan-based friend selection and recommendation;257
20.4.3;4.3 Taking available connections into consideration;257
20.5;5. Evaluation;257
20.5.1;5.1 Experimental setup;258
20.5.2;5.2 Comparison;260
20.6;6. Conclusions and FutureWork;261
20.7;Acknowledgments;262
20.8;References;262
21;A fine-grained peer sharing technique for delivering large media files over the Internet;264
21.1;1. Introduction;264
21.2;2. Fine-Grained Peer Sharing in CDNs;265
21.2.1;2.1 Peer sharing in CDNs;265
21.2.2;2.2 Fine-grained peer sharing;266
21.3;3. Communication Protocol and Scheduling Algorithm;267
21.3.1;3.1 Communication protocol;268
21.3.2;3.2 Scheduling algorithm;269
21.4;4. Performance Evaluation;270
21.5;5. RelatedWork;273
21.6;6. Concluding Remarks;273
21.7;References;273
22;Proxy-cache aware object bundling for Web access acceleration;274
22.1;1. Introduction;274
22.2;2. Existing Methods to Retrieve Embedded Objects;276
22.3;3. Traffic Characteristics of Embedded Objects;277
22.3.1;3.1 Number of embedded objects per page;277
22.3.2;3.2 Content of embedded objects;278
22.3.3;3.3 Sizes of embedded objects;278
22.3.4;3.4 Retrieval latency of embedded objects;279
22.4;4. Proxy-Cache Aware PC-Bundle Transfer of HTTP Messages;280
22.4.1;4.1 Basic mechanism;280
22.4.2;4.2 Assumptions;281
22.4.3;4.3 Detailed Specifications of Modifications to HTTP for;282
22.5;5. Experimental Results on PC-Bundle Mechanism;284
22.5.1;5.1 Environment of simulation;284
22.5.2;5.2 Results with respect to cache size;285
22.5.3;5.3 Results with respect to cache replacement policy;287
22.5.4;5.4 Results with respect to degree of parallelism;288
22.6;6. Conclusions;289
22.7;References;290
23;A case for dynamic selection of replication and caching strategies;292
23.1;1. Introduction;292
23.2;2. Evaluation methodology;293
23.2.1;2.1 Simulation model;293
23.2.2;2.2 Adaptation mechanisms;294
23.3;3. The Need for Dynamic Adaptation;295
23.4;4. Strategy Selection Heuristics;296
23.4.1;4.1 Performance evaluation metrics;296
23.4.2;4.2 Selection heuristics;297
23.5;5. Related work;298
23.6;6. Conclusions and FutureWork;299
23.7;References;299
24;Link prefetching in Mozilla: A server-driven approach;300
24.1;1. Introduction;300
24.2;2. Prefetching Directives;301
24.3;3. Determining When To Prefetch;303
24.4;4. Browser Imposed Restrictions;304
24.5;5. Identifying Prefetch Requests;305
24.6;6. Practical Limitations;305
24.7;7. Applications and Results;306
24.8;8. Conclusion;306
24.9;Acknowledgments;307
24.10;References;307
25;A generalized model for characterizing content modi.cation dynamics of Web objects;310
25.1;1. Introduction;310
25.2;2. Related Work;311
25.3;3. Modeling Content Modification Dynamics of Web Objects;312
25.3.1;3.1 Quantity matrix for content modification dynamics;312
25.3.2;3.2 Calculation of quantity matrix;313
25.4;4. Verification of Quantity Matrix for Content Modification Dynamics of Objects;316
25.5;5. Conclusions;317
25.6;References;317
26;Server-friendly delta compression for efficient Web access;320
26.1;1. Introduction;320
26.1.1;1.1 Delta compression for Web access;321
26.1.2;1.2 Discussion of known approaches;322
26.1.3;1.3 Our approach and contributions;323
26.2;2. Delta Compression Schemes for Site Visits;324
26.2.1;2.1 Reference file selection policies;324
26.2.2;2.2 Experimental setup;325
26.2.3;2.3 Experimental evaluation of policies;326
26.2.4;2.4 An efficient and nearly optimal protocol;330
26.2.5;2.5 Impact of duplicates;332
26.2.6;2.6 Summary of observations;333
26.3;3. Utility of rsync for Web Access;333
26.3.1;3.1 Experiments for related pages;334
26.3.2;3.2 Experiments for versions of a page;335
26.3.3;3.3 Discussion and a generalized approach;336
26.4;4. Concluding Remarks;338
26.5;Acknowledgements;338
26.6;References;338
27;Evaluation of ESI and class-based delta encoding;340
27.1;1. Introduction;340
27.2;2. ESI;341
27.3;3. A Page Content Model;342
27.4;4. The System Model;344
27.5;5. Simulation;345
27.6;6. ESI Simulation Results;346
27.6.1;6.1 Simulation Results - Online Bookstore’s Book Pages;346
27.6.2;6.2 Simulation Results - Other Resources;349
27.7;7. Class-Based Delta Encoding;351
27.7.1;7.1 System Model;351
27.8;8. Delta Encoding Simulation;352
27.8.1;8.1 Simulation Results - Online Bookstore;352
27.8.2;8.2 Simulation Results - My Page;353
27.8.3;8.3 Same-Base reply for Online Bookstore;353
27.9;9. Comparison of ESI and Class-Based DE;355
27.9.1;9.1 My Page Resource;355
27.9.2;9.2 Online Bookstore Resource;357
27.10;10. Additional Considerations;358
27.11;Acknowledgments;359
27.12;References;359
28;Author Index;362
29;More eBooks at www.ciando.com;0

Leseproben

SERVER-FRIENDLY DELTA COMPRESSION FOR EFFICIENT WEB ACCESS (p. 303-304)

Anubhav Savant and Torsten Suel
CIS Department, Polytechnic University


Abstract

A number of researchers have studied delta compression techniques for improving the efficiency of web page accesses over slow communication links. Most of these schemes exploit the fact that updated web pages often change only very slightly, thus resulting in very small sizes for the transmitted deltas. However, these schemes are only applicable to a minority of page accesses, and require web or proxy servers to retain potentially many different outdated versions of pages for use as reference files in the encoding. Another approach, studied by Chan and Woo [4], encodes a page with respect to similar files located on the same web server that are already in the client’s browser cache. Based on the latter approach, we study different delta compression policies for web access. Our emphasis is on web and proxy server-friendly policies that do not require the maintenance of multiple older versions of a page, but only use reference files accessed by the client within the last few minutes. We compare several policies for identifying appropriate reference files and evaluate their performance on a set of traces. We show that there are very simple policies that achieve significant benefits over gzip compression on most web accesses, and that can be efficiently implemented at web or proxy servers. We also study the potential of file synchronization techniques such as rsync [28] for web access.


1. Introduction

Delta compression (delta encoding) is the process of encoding a target file with respect to one or several, usually similar, referenceles. This encoding, called a delta, describes the target file in terms of the reference files, and a recipient that receives the encoding and already knows the reference files can thus efficiently reconstruct the target. Delta compression has numerous applications in scenarios where there are several versions of a file or many similar files, such as software revision control systems, distribution of software updates, content distribution networks, or efficient storage of related files. Several tools for delta compression, such as bdiff, vcdiff [10, 13], Xdelta [14], and zdelta [25], are freely available. We refer to [23] for an overview of delta compression techniques and applications.


1.1 Delta compression for Web access

A number of authors have proposed the use of delta compression techniques to improve the ef.ciency of web access [1, 4, 7, 9, 16, 17, 21, 27, 29]. In particular, when web pages are updated, they typically do not change by much, and thus delta compression can be used to very succinctly encode the difference between a new version of a web page and an outdated version already in the client’s browser cache. Most proposals focus on encodings between different versions located at the same URL, which results in small sizes for the deltas but is restricted to pages that have been previously visited by the client. One exception is the work by Chan andWoo [4], which proposes to use as reference files other pages on the same site recently visited by the client, which tend to have a significant degree of similarity due to common layout features and HTML structure. In general, delta compression schemes for web access can be distinguished along the following axes.