Apple Mac OS X Server v10.3 - Manual (Page 47)

HTTP load balancing provides a way to distribute user load among a group of application servers.
The application servers can be standalone or configured as a cluster, in which case they are know as
nodes. Load balancing is better used with sticky sessions. This means that once the load balancer (a
web server) forwards a client request to a particular application server, it sends all further requests
from the client to the same application server.
Using load balancing across standalone application servers allows you to scale your deployment with
little increase in request-processing overhead. However, when an application server fails, other
application servers cannot pick up the failed-server’s load, which may provide an undesirable user
experience: Users may have to log in to the application again or may lose the contents of their
“shopping carts.”
Load balancing across clustered application servers (or nodes) provides session replication among
the nodes, so that when a node fails, another node can take over its duties with little or no user impact.
However, as you add nodes to the cluster, each request may take longer to process.
This chapter explains how to enable an application to be distributable among cluster nodes and walks
you through configuring HTTP load balancing for Sun’s Pet Store using three computers: One serving
only as the web server and load balancer, and the other two serving as application-server nodes.
Distributable Applications
Before deploying an application in a cluster of nodes using the deploy-cluster configuration, make
sure that the application is distributable.
To make an application distributable set Distributable to yes in the Web-App pane of the application’s
WebApp window. “Figure 6-1” shows the WebApp window of the
petstore.ear archive.
Distributable Applications 47
2005-08-11 | © 2003, 2005 Apple Computer, Inc. All Rights Reserved.
C H A P T E R 5
Balancing User Load and Replicating
Sessions