|Oracle E-Business Suite Concepts|
Part Number E12841-04
Oracle E-Business Suite provides numerous options for building and tailoring an installation to meet specific business and technical requirements. At a simple level, this includes the capability to utilize varying numbers of machines to distribute the various Oracle E-Business Suite technology layers according to desired expenditure and required performance. There are also more specific factors, such as physical site organization and expected growth rate. Other aspects are the possible need to provide extra capability to cope with fluctuations in demand, and the possible need for resilience in the event of problems affecting some of the hardware components. A final (and very important) consideration in planning a system is the need for the appropriate level of security to be put in place.
Many of these subjects are described in other chapters of this book. This section concentrates on how to make the decisions needed to balance the load on various components or layers, which is of particular importance for two reasons:
Load balancing can involve the entire infrastructure of an Oracle E-Business Suite installation, with a change in one area potentially having significant effects elsewhere.
Making informed decisions about load balancing can often enable a higher level of performance to be obtained without expenditure on additional hardware.
Load balancing areas include:
Domain Name Server (DNS)
Web (HTTP) Services
Concurrent Processing Layer
The emphasis here is on describing load balancing strategies and their key features, to allow an informed decision to be made regarding the applicability and usefulness of a particular area of load balancing in achieving the desired technical and business requirements.
Load balancing is the means by which network traffic directed to a particular Web site is divided between one or more machines in a cluster of servers. Typically, the servers will all be running the same application, and use of load balancing presents the outside world with the appearance of a single server rather than a cluster. The relevant hardware device, often called a server load balancer, receives the relevant network traffic for the site, and distributes it to the various servers in the cluster based on the load balancing methodology that it supports. By sending requests to different nodes within the server cluster, system performance is optimized, scalability is simplified, and application availability (a key requirement of Web-based applications) is greatly enhanced.
Additional features of load balancing include monitoring server availability and context-based load distribution.
Monitoring server availability maintains a watch on the server pool, and periodically checking that all machines are responding to the traffic; if one is not (perhaps because of a network problem) it is taken out of the pool of servers, so that traffic can be routed to the servers that remain available.
Context-based load distribution is required for applications such as Oracle E-Business Suite that need to maintain session persistent-connections, whereby cookies are created when the session is initially established. To support this requirement, the load balancer reads and updates the network packet header information sent with each request made by the client, and routes the request to the node in the cluster with which the corresponding session was originally established and is subsequently being maintained.
Load balancers are not installed as part of Oracle E-Business Suite. However, Oracle E-Business Suite can be configured for use with them as required.
The main categories of load balancer are:
Session Persistent Load Balancers - After a client's HTTP connection is established with a particular server, subsequent HTTP requests from that client are directed to the same server, for the duration of the session. This persistency is also referred to as stickiness.
Non-Session Persistent Load Balancers - These load balancers use a round-robin strategy for balancing incoming HTTP requests, and do not maintain session persistent client connections. After a client's initial HTTP connection is directed to a given server, subsequent HTTP requests from that client will be not necessarily be directed to the same server.
Secure Sockets Layer (SSL) Accelerators - Secure Sockets Layer (SSL) accelerators can be used to reduce the SSL traffic and workload of the Web servers.
Usually, an SSL accelerator is the target for HTTPS browser requests, and thus the target for all client communication. It is responsible for converting HTTPS SSL requests to non-SSL HTTP requests, directing the subsequent request to the HTTP server (running in non-SSL mode). Before sending the response back to the client browser, the SSL accelerator converts the non-SSL requests back to SSL requests, in a reverse of the initial process.
Different types of load balancing can be employed with Oracle E-Business Suite. The most suitable choice will depend on the specific needs and existing infrastructure of the site where deployment is to take place.
Domain Name Server (DNS) Round Robin Load Balancing
This load balancing strategy distributes end-user requests across multiple server nodes, based on dynamic assignments of IP addresses to a given hostname and associated domain name. Utilizing a feature known as DNS round robin allows the association of more than one IP address with a given hostname, thereby facilitating distribution of the traffic across the list of IP addresses associated with that hostname.
Figure 10-1 shows an example of a configuration that uses DNS layer load balancing.
Figure 10-1 DNS Layer Load Balancing
Although DNS round robin load-balancing is a simple and effective way to distribute traffic among several servers, it has several limitations. These include unpredictable load distribution across servers (depending on user activities) and DNS caching issues with browsers.
HTTP Layer Hardware Load Balancing
This type of load balancing can be used if you have a hardware load balancer that accepts HTTP communication and forwards it to a group of server nodes, sometimes referred to as a farm.
Figure 10-2 shows an example of a configuration where an HTTP load balancer distributes the load across a farm consisting of two Web server nodes.
Figure 10-2 HTTP Layer Load Balancing
Hardware-based HTTP load balancers must be configured for persistent session connections for all traffic through all ports for Oracle E-Business Suite Release 12 environments. This is required as various modules shipped with Oracle E-Business Suite need to maintain session state. If this is not set, users may experience transaction state loss errors while using E-Business Suite.
OC4J Load Balancing
OC4J load balancing (clustering) utilizes Oracle Application Server 10g's built-in load-balancing capability at the JVM level to balance OC4J requests among multiple application tier nodes, thereby providing fault tolerance. A prerequisite for this strategy is a hardware load balancer that implements single or multiple entry points.
Note: For additional information on load balancing configurations, see My Oracle Support Knowledge Document 380489.1, Using Load-Balancers with Oracle E-Business Suite Release 12.
Copyright © 2000, 2010, Oracle and/or its affiliates. All rights reserved.