Research Progress Presented By: Mohamed Ramez Atassi Supervisor: - - PowerPoint PPT Presentation

Presented By: Mohamed Ramez Atassi Supervisor:

• Dr. Ahmed Sameh

Research Progress

How to use a group of untrusted volunteers to

increase the capacity of the web server of a dynamic web site

Volunteers contribute by providing disk space and

network bandwidth

Internet forums application (Message Boards) is

used as a Case Study

1- Increase server capacity and performance 2- Scalability 3- Load Balancing 4- Data Integrity 5- Low Latency 6- Security 7- Up-to-date data 8- Adaptive 9- Application-Specific Requirements

The web server creates and maintains a DHT Volunteers join the DHT Popular web files are periodically inserted into DHT using

content hashing as key

Links to immutable files like images are inserted directly on

web pages using dht:// links

Example: dht://dht.mysite.com/key=A2SD74EF5F389 These dht:// links are resolved by a browser’s extension Web pages are accessed using a script on server Script directs requests to either DHT or files on server Redirection to DHT is achieved by returning an html file with

an embedded iframe with DHT link as source or by JavaScript

1 - Approach Modifications 2 – Client Extension Alternatives 3 – DHT Peer program implementation

Two approaches were suggested in first proposal: Complete

page storage and storage of objects as html fragments

Complete page storage causes wasted storage, hence wasted

bandwidth to synchronize replicas

Second approach causes more server load and higher failure

rates

New approach, combination of both, is suggested Store complete web page as one fragment, and add other

fragments that modifies the first fragment – using JavaScript – as needed

An extension of the client’s browser is used to do the

following tasks:

• 1- dht:// links are used to retrieve a file from the DHT network. The

extension must first contact the server to get an address of a random node IP/Port

• 2- The node address should be stored in a cookie. The same address

should be used till a timout or a failure occur. Then, a new random node address would be retreived

• 3- The retrieved file must be hashed. The resulting value is compared to

file key to ensure data integrity

An attractive objective to add to our objectives would be

“Client-side transparency”

Question: Can we fulfill the required tasks using JavaScript

without using a browser extension

Many DHT implementations are available: Chord, Tapestry,

Pastry, CAN, Kademilia, Bamboo, …

Chord/Dhash has been chosen Default implementation versus Other ports Choosing Parameters:

• Simple replication vs Erasure codes
• Number of replicas
• Replicas maintenance algorithm
• Number of virtual nodes per real node

Modify the implementation to act as a mini web server