WebLogic Clustering - Failover, and Load Balancing

1
WebLogic Clustering - Failover, and Load Balancing

• Bryan Ferrel and Ramarao Desaraju
• CS 522
• Computer Communications
• December 4, 2002

2
Introduction

• What is Clustering?
• Close cooperation of two or more replicated
  servers to ensure fast, continuous service to
  users.
• Clustering must provide the following
• No bottlenecks to scaling
• No single point of failure
• Transparency to application developers
• Single-system image to administrators

3
Clustering

• Benefits of Clustering
• High-Availability and Scalability
• Key Capabilities of Clustering
• Failover
• New object can take over for the failed object
• Load Balancing
• Even distribution of jobs
• Challenges
• Replication without increasing load
• Propagating changes to servers participating in
  the cluster
• Complex setup

4
WebLogic Clustering

• What is WebLogic?
• J2EE Application Server that is a platform for
  developing distributed applications
• A hosting environment for EJBs
• Provides several services such as JMS, JNDI,
  JDBC, Transaction control, etc
• WebLogic Clustering
• Consists of multiple copies of the WL Server
  working togetherso that there are alternative
  objects to do the same job.

5
WebLogic Clustering

• Types of objects that can be clustered
• Servlets and JSPs
• EJBs
• RMI objects
• JMS destinations
• JDBC Connections

6

• Replica-aware stubs created at compile time
• Contains load-balancing algorithm

7
Typical Clustered Multi-Tier Architecture
8
Load Balancing

• Algorithms used
• Round-Robin
• Cycles through instances in order
• Weight-Based
• Allocation of requests based upon relative
  weights
• Random

9
What we did

• Set up a WL server in a clustered environment
• 2 Nodes and 1 Admin. server
• Deploy an EJB
• Stateful Session Bean that is replicated on the
  cluster
• Simulate the action of multiple clients
• Remote Java client that creates multiple threads
  and sends off requests in quick succession
• Server logs client requests
• Used for gathering statistics

10
Test Setup
11
Results

• Most effective algorithm was Round-Robin
• Weight-based is useful when there are differences
  in the cluster hardware
• Degenerates to Round-Robin when using identical
  weights in the cluster.
• Random was close to Round-Robin
• Scalability
• 1 server results in cacheful exceptions
• Adding another node to the cluster removed this
  problem with identical clients

12
(No Transcript)
13
References

• Girdley, Woollen, Emerson, J2EE Applications and
  BEA WebLogic Server, Prentice Hall, Upper Saddle
  River, NJ, 2002
• On Line References
• http//e-docs.bea.com/wls/docs70/cluster/
• http//www.getgamma.com/wlscert/wls6clustering.htm
  l
• http//www.distributedobjex.com/resources/techbrie
  fs_BEA.htm
• http//java.sun.com/j2ee