Transcript PPT

On the Effect of Server Adaptation
for Web Content Delivery
Yin Zhang (AT&T)
Joint work with
Balachander Krishnamurthy (AT&T)
Craig Wills (WPI)
IMW ’02, Marseille, Nov. 2002
Motivation


Web sites have a strong
incentive to reduce
time-to-glass
Challenge

Internet

Natural solution –
server adaptation

foobar.com
11/8/2002
client connectivity is
heterogeneous
IMW 2002
client connectivity +
content characteristics +
client capability +
server load + …
 action to take
2
Study: What?

Basic question – What exactly is the
performance impact of server adaptation?




When and how much can server adaptation help?
Which action should the server take?
Lots of previous work … but typically
focusing on one individual action
This study –


Provides a unified framework for assessing the
impact of different server actions
Obtains useful insights through multi-site widearea measurements
11/8/2002
IMW 2002
3
Factors Considered

Client connectivity


Latency, bandwidth
Content characteristics


Criteria: total bytes, container bytes, #objects
3x3x3 = 27 buckets



derived from large proxy logs
further justified by examining popular Websites’ pages
Server actions

Altering the content


Altering the location of the content


compression, bundling
Altering protocol options


using a Content Distribution Network (CDN)
Altering manner of delivery


reducing number of images, reducing image size
using persistent connections
Combination of different actions
11/8/2002
IMW 2002
4
Experiment Methodology

350 uk.icir

uk.wpi
300
A multi-site study
Server: Apache
Round-Trip Time (ms)


250
modem.icir
au.wpi
Client: httperf

200
au.icir
modem.wpi
150

de.icir

isdn.icir
100
att.icir de.wpi
0
20
11/8/2002
40
100
80
60
Throughput (KB/sec)
att.wpi(462.6)
140
120
IMW 2002

US: att, modem, isdn
Intl: de, au, uk
Canonical content served
at each site

isdn.wpi
50
0

West coast: icir
East coast: wpi
covering the space of
buckets
Experiments repeated at
different times of day
5
Results

Compression of HTML is not universally useful


Persistent connections alone has limited benefit





Exception: bandwidth-constrained clients
Bundling gives significant improvement


Little improvement for all client/server combo
Pipelining gives significant improvement


It only works for bandwidth-constrained clients
Bundling alone is similar to pipelining
Compressed bundles help a lot under all conditions
CDN-served bundles – good idea for well-connected clients
Reducing image size by half has little benefit
Reducing the number of objects by half helps a lot
under most conditions
Baseline: 4 parallel HTTP/1.0 connections
11/8/2002
IMW 2002
6
Contribution and Further Work

Contribution

A unified framework for evaluating the impact
of server adaptation



Can be applied by individual Web site
Insights we gained can be useful for improving
client performance
Further work

Evaluation of the feasibility of online client
classification and server adaptation through
real implementation

11/8/2002
Our results are encouraging
IMW 2002
7
Acknowledgments

People who gave us accounts / logs
11/8/2002
IMW 2002
8