Horizontal scaling with Network Dispatcher sample topology

Horizontal scaling refers to setting up multiple application server cluster members on two or more physical machines within a single WAS cell. In this variation on the basic horizontal scaling topology, Network Dispatcher distributes client requests to multiple HTTP servers. This topology requires WAS Network Deployment.

Network Dispatcher is part of the Load Balancer feature of the IBM WAS Edge Components product. The Edge components are included in IBM WAS Network Deployment.

Load balancing distribute HTTP requests among application servers that run on multiple physical machines. Network Dispatcher uses server availability, capability, workload, and other user-defined criteria to perform intelligent load balancing and determine which server should receive each client request. Network Dispatcher receives client HTTP requests and transparently distributes them among the available Web servers in the cluster. Network Dispatcher enables multiple TCP servers to be dynamically linked in a single entity that behaves as a single logical server.

Note: Network Dispatcher and other load balancing products are sometimes referred to as dispatchers or IP sprayers.

This figure shows an example of a horizontal scaling topology that uses Network Dispatcher:

This sample topology inclues these features:

• The use of two iSeries servers and LPAR provides process and data isolation, as well as hardware redundancy.
• Both Machine C and Machine D host a cluster member of each application server cluster.
• Each application server machine hosts its own Web server.
• Machine B acts as the backup Network Dispatcher node. If the primary Network Dispatcher node (Machine A) fails, the backup node becomes active.
• A single Network Deployment manager communicates with a node agent on each machine to manage the application servers.

Use Network Dispatcher with firewalls

You can also use Network Dispatcher with demilitarized zone (DMZ) topologies. For example, using Network Dispatcher can simplify the creation of a DMZ topology where one firewall protects the Web server from the public Web site and a second firewall uses proxy services to protect back-end systems from the Web server in the DMZ. A basic DMZ configuration with Network Dispatcher includes these features:

• An outer firewall that allows only HTTP and HTTPS traffic. This firewall separates the Internet from the Network Dispatcher.
• A DMZ that contains the Network Dispatcher node and the Web servers.
• An inner firewall that separates the Web servers from the application servers and back-end resources. The inner firewall is configured to handle non-HTTP protocols, such as IIOP and JDBC.

Advantages and disadvantages of horizontal scaling with Network Dispatcher

In addition to the advantages and disadvantages of horizontal scaling, this topology provides these benefits:

• Improves scalability by allowing you to add application servers as needed to adapt to changes in workload.
• Improves the performance of servers by distributing client requests among a group of servers.
• It increases the number of connected users.
• Eliminates the Web server as a single point of failure.
• Can be used in combination with WAS workload management to eliminate the application server as a single point of failure.
• Improves throughput by enabling multiple servers and CPUs to handle the client workload.

Configure a topology with Network Dispatcher

To configure the machines containing Web servers and application servers, see the instructions for the topology you plan to implement.

The dispatcher communicates with the Web server, which in turn communicates with application servers. This type of topology involves configuring communications between the dispatcher and the Web server. To configure a dispatcher to communicate with your Web servers, see the documentation for the load balancing product you plan to use.

Dispatchers typically includes utilities to send requests to the application server. These utilities, sometimes called advisors, allow the dispatcher to test the wellness of the entire path from client to Web server and from Web server to application server. Many commercially-available dispatchers include a set of preconfigued advisors. You can also write a customized advisor to test specific circumstances related to application server wellness.