Download presentation
Presentation is loading. Please wait.
Published byLewis Sharp Modified over 8 years ago
1
Distributed Server Scheduler Eyal Serero Alex Fishgate Supervisor : Vitaly Suchin
2
The goal: The goal: »improve the server response time with a cost effective solution »Quality of service will depend on client priority The solution: The solution: »Distribute the load over more then one server
3
Existing solutions Round robin DNS Round robin DNS Single point of failure Single point of failure Cache problem Cache problem Static Load balancing Static Load balancing DNS server Web server
4
Our solution A distributed dispatcher A distributed dispatcher Dynamic load balancing Dynamic load balancing Support for client with different priorities Support for client with different priorities
5
A distributed dispatcher No single point of failure No single point of failure No cache problem No cache problem Web server The internet
6
Basic concepts Based on the jigsaw – the W3C’s web server Based on the jigsaw – the W3C’s web server A java object representation of the site content. A java object representation of the site content. Load Table Load Table Server for internal communication Server for internal communication Client priority table Client priority table
7
Internal messages Load information Load information Redirected client information Redirected client information Resource time calculation Resource time calculation Web server
8
Priorities Client priority will be statically configured (before the server is started). Client priority will be statically configured (before the server is started). For each request, the server will search the priority in a special table (there will be a default value) For each request, the server will search the priority in a special table (there will be a default value) The server will determined the thread priority According to the client priority The server will determined the thread priority According to the client priority
9
Client redirection process Web server A Web server B 1.Client send a request to server A 2.Server A find out that server B is best 3.Server A send a redirection message to the client 4.Server B will process the request 5.Server B will calculate the new load information for the resource and send it to all the other servers
10
fin
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.