Performance of HTTP Application in Mobile Ad Hoc Networks Asifuddin Mohammad

Performance of HTTP Application in Mobile Ad Hoc

Networks

Asifuddin Mohammad

Performance of HTTP Application in Mobile Ad Hoc Networks

Introduction Related work Paper 1 Paper 2

Introduction

Objective Understanding the effect of factors like

mobility, congestion etc., on HTTP response time in Mobile Ad hoc Networks(MANETs)

Introduction

Preview Characterize the resilience of Client-Server

application in MANETs using Abdul's State-Space framework

Introduction

Major issues with TCP in wireless environment Mobility Congestion

Many changes Proposed

Related work

Paper 1 Performance benchmarking of wireless Web

servers Paper 2

On Routing Web and Multimedia Traffic in Mobile Ad Hoc Networks

Paper 1

Does an experimental analysis of HTTP transaction

Focus on HTTP transaction rate and end to end throughput achievable in Ad hoc networks

Studies the impact of factors such as Web object size, number of clients, and persistent HTTP connections

Paper 1

Portable short-lived network wireless classroom Sporting events Disaster recovery

Focus on wireless web servers

Paper 1

Web protocols: IP, TCP and HTTP IP

Connectionless network layer Provides global addressing and routing

TCP Connection-oriented transport layer Provides end to end data delivery

HTTP Request-response application layer Transfer web documents between web servers

and web clients

Paper 1

Paper 1

Each Notebook is running RedHat Linux OS

Cisco Aironet 350 Series Adapter configured to operate in ad-hoc mode

DCF,11Mbps,MTU=1500 bytes

Paper 1

This paper doesn't consider node mobility, multiphop and ad hoc routing issues

Httperf Used to generate client request Web workload generation tool developed at HP

laboratories for web performance measurements Apache webserver

HTTP complaint 70% deployed

Network Analyzer SnifferPro 4.6 Provides real time capture of all the observed

traffic

Paper 1

Experiment 1: Request rate Experiment 2: Multiple Clients Experiment 3: Persistent HTTP connections Experiment 4: Transfer Size

Paper 1

Experiment 1: Request rate Single client, sever and Sniffer Laptops Web object size is 1KB 10 request per second Non-persistent connections Each HTTP 1.0 transaction generates 10 TCP

packets

Paper 1

Paper 1

Paper 1

Paper 1

Paper 1

Experiment 2: Multiple Clients Two or more Clients More than 30% throughput then single client

Fairness Problem in Wireless network Load Imbalance Heterogeneous transmission rates Difference in wireless channel quality Contention patterns in the wireless channel

access Packet losses at a point of congestion shared by

upstream and downstream flows Relative phasing between clients causes

fairness issues

Paper 1

Paper 1

Paper 1

Experiment 3: Persistent HTTP connections TCP connection rate is 10 request per second Transfer size is 1KB Persistent HTTP

Results 350% increase

Paper 1

Paper 1

Paper 1

Experiment 4: Transfer Size Single Client,10 request per sec Varying Transfer size of

8KB,32KB,48KB ,64KB

Paper 1

Paper 1

Paper 1

Paper 1

Conclusion Wireless web servers can provide 1KB of HTTP

transaction rates of 110 connection for non-persistent connection and 320 connections for persistent

Bottleneck at server and client interface Unfairness Network trashing Persistent HTTP solution for some extent