DYNAMIC LOAD BALANCING ON WEBSERVER SYSTEMS by Valeria
DYNAMIC LOAD BALANCING ON WEB-SERVER SYSTEMS by Valeria Cardellini Michele Colajanni Philip S. Yu
Contents ± ± ± ± INTRODUCTION CLIENT-BASED APPROACH DNS-BASED APPROACH DISPATCHER-BASED APPROACH SERVER-BASED APPROACH COMPARING THE APPROACH CONCLUSIONS
INTRODUCTION ± ± disproportionate increase in client requests to popular Web sites solution ® mirroring ® replicate information across a mirrored-server system ® not user-transparent ® distributed Web-server system ® Information can be distributed among server nodes: ® ® ± 1 st : content tree replication – LAN, WAN 2 nd : information sharing – LAN Successful load-balancing ® transparent users ® appear as a single host to outside world ± four distributed Web-server architectures ® Client-based ® DNS-based , dispatcher-based , server-based
CLIENT-BASED APPROCH ± Web Clients ® Web client selects a node of the cluster and submits the request to the selected node ® ® Netscape Navigator – random select Limited practical applicability and is not scalable ® Smart Clients ® Migrates server functionality to the client through a Java applet ® Increase network traffic and network delay ± Client-Side Proxies ® Web Cluster standpoint, proxy servers are similar to clients ± not universally applicable
DNS-BASED APPROACH ± ± ± The cluster DNS translates URL to the IP address User transparent ex) www. yahoo. com www. cnn. com
DNS-BASED APPROACH ± (con’t) Drawbacks ® The DNS a limited control on the request reaching the Web cluster ® ± Between the client and the web server DNS, many intermediate name servers can cache the logical name to IP address mapping to reduce network traffic and every web browser typically caches some address resolution scheduling algorithm that the cluster DNS uses to balance the Web-server node’s load ® constant TTL algorims ® System-stateless ® Server-state-based ® Client-state-based ® adaptive TTL algorims ± DNS-Based Architecture Comparison
DISPATCHER-BASED APPROACH ± ± Centralize request scheduling and completely control client-request routing Request routing among server is transparent-unlike DNS-based ® ± ± ± DNS deals address at the URL level, the dispatcher has a single, virtual IP address(IP-SVA) Dispatcher uniquely identifies each server in the system through a private address Dispatcher typically use simple algorithms to select the Web server difference by routing mechanism ® ® Packet Single-Rewriting Packet Double-Rewriting Pachet Forwarding (network dispatcher, One-ip address) HTTP redirection
Packet Single-Rewriting ± ± TCP router acts as an IP address dispatcher High System availability
Packet Double-Rewriting ± Two solution using this approach ® Magicrouter ® round-robin, random, incremental load algrithm ® Cisco System’s Local Director ® least number of active connections
Packet Forwarding ± Network Dispatcher ® Dispatcher forward packets to the selected server using its physical address without IP modification ® Level 1 : single-rewriting mechanisim ® Level 2 : LAN network dispatcher
Packet Forwarding (con’t) ± ONE-IP address ® uses the ifconfig alias option to configure a Web- server system with multiple machines ® implemented with two techniques Routing-based dispatching ® Broadcast-based dispatching ®
HTTP Redirection ± ± ± among the Web-server nodes through the HTTP’s redirection mechanism no IP address modification two techniques ® Server-state-based dispatching ® used by Distributed Server Groups architecture ® Location-based dispatching ® ± used by Cisco System’s Distributed. Director appliance Dispatcher-Based Architecture Comparison
SERVER-BASED APPROACH ± Use two level dispatching mechanism ® Integrating the DNS based approach with redirection techniques executed by Web server ® Solves most DNS scheduling problem ± Two Solution ® HTTP redirection ® Packet redirection
COMPARING THE APPROACHES ± Approach Trade-off Summary
COMPARING THE APPROACHES ± Performance Evaluation
CONCLUSIONS ± ± network bandwidth can constrain loadbalancing performance LAN-distributed Web-server cluster are thus a limited solution to increased client requests
- Slides: 16