Adaptive Content Delivery for Scalable Web Servers

1
Adaptive Content Delivery for Scalable Web
Servers

• Authors Rahul Pradhan and Mark Claypool
• Presented by David Finkel
• Computer Science Department
• Worcester Polytechnic Institute
• Worcester, MA, USA

2
Outline

• Introduction
• Managing Load
• Approach
• - Methodology
• - Design
• Experiments
• Client Side Experiments
• Future Work Conclusions

3
Introduction

• Enormous Growth of Web
• Exponential in number if users, pages, and sites
• New Web uses require different resources
• Static Disk, Network
• Dynamic CPU, Network
• Media streaming CPU, Disk, Network

4
Web Server Overload

• Loaded server rejects connections, denying
  service
• Companies lose revenue
• Loaded server increases response times
• Slower pages viewed as less interesting

5
Current Approaches to Managing Load

• Over provisioning
• Beef up single server
• Can still become loaded during flash crowds
• Load balancing
• Server farm or CDN
• Individual servers may still become over-loaded
• Content adaptation
• Reduce resources needed upon heavy load

6
Content Adaptation Examples

• Using thumbnails instead of full, inline images
• Reducing the number of local links
• Reducing the number of embedded objects
• Changing the quality of images
• Current approaches manual!
• Our approach is to adapt content automatically

7
Our Approach

• Two versions of Web pages
• High quality to serve under normal load
• Low quality to serve under high load
• Monitor the server load
• A separate light weight standalone process
• Upon heavy load, server switches to low quality
  transparently
• Requires no modification to
• Server
• Browser
• http protocol

8
Adaptive Content Delivery System Architecture
Base System
Our System
Requests
Adaptation Module
Web Server
Load Monitor
Disk
Response
Content Switching
CPU, Disk, Network
9
Load Monitor

• Continuously monitors the utilization of the
  server and the observed response time
• Developed utilities to measure utilization
• CPU, Network, Disk
• Observed Response Time
• Used Linux /proc file system, but techniques
  general enough for any OS

10
Adaptation Module

• Input of load values from the load monitor
• Decides low or high load
• Low and high thresholds
• Threshold values determined by prior measurement
• Scripts to induce load on server
• httperf to generate requests
• measured response time (using httperf)

11
Response Time vs CPU Util
Thresholds 60 and 75
12
Content Selector

• Transparently switches content depending on the
  decision made by the adaptation module.
• We use symbolic links to make the same file
  point
• to different qualities of content.
• indexhigh.html
• index.html
• indexlow.html

13
Experiments

• Server
• P-III, 500,128 MB RAM, IDE, 10 Mpbs, Linux
  2.2.14, Apache 1.3.12
• Workloads
• Static Workload
• Dynamic Workload
• Multimedia Workload
• Metrics
• Throughput (Responses/sec)
• Average Response Time
• Percentage of Errors
• Frame Rate (for Multimedia Clients)

14
Response Time (ms) vs Requests/sec
15
Percent Errors vs Requests/sec
16
Frame rate vs Number of MM Clients
17
Overhead For the Adaptive Content Delivery System
18
Client Side Experiments

• Experiments on real servers to determine
• the impact of file size on response time.
• Used a modified httperf for our
• measurements to generate requests
• Measured the response time along with
• the connection set up time and transfer time.

19
Conclusions

• Server load critical
• We present a mechanism to
• Quantify server load
• Adapat transparently to client
• Improves server performance
• supports 25 more static requests
• supports twice as many Multimedia clients
• supports 15 more CGI requests

20
Future Work

• Adapting to heterogeneous client environment
• Clients may have different bandwidths
• Adding QoS features to the Web server
• Range of content quality at server
• Maximize QoS for user

21
Adaptive Content Delivery for Scalable Web
Servers

• Authors Rahul Pradhan and Mark Claypool
• Presented by David Finkel
• Computer Science Department
• Worcester Polytechnic Institute
• Worcester, MA, USA

22
Extra slides past here .
23
JPEG Quality JPEG Quality Factor vs
Percentage Savings in File Size
24
MPEG Quality MPEG Q Scale Factor vs Percentage
File Size Savings
25
Response Time vs Number of CGI Requests
26
Percentage Of CGI Requests Rejected vs Number of
CGI Requests