NEW  Understanding Publisher clustering

In high-demand environments, multiple Publisher computers can be configured in a cluster to improve rendering performance and to simplify the configuration of Publisher computers. Publisher clustering does not appreciably improve performance for publishing jobs that do not perform document rendering.

Note    Each Publisher computer in the cluster will claim an appropriate license as described in Understanding the licensing.

You should consider Publisher clustering when:

• More than one of the rendering modules will be used for many documents. Each of the rendering modules can be configured on a separate Publisher computer dedicated to that rendering module.
• Significant numbers of large documents will be rendered with different rendering modules. Each of the rendering modules can be configured on a separate Publisher computer dedicated to that rendering module.
• Only modest hardware resources are available for the Publisher computer but multiple computers are available. Multiple Publisher computers can be configured with the same rendering modules and groups of users can be assigned different publishing jobs that specify which Publisher server to process their jobs.
• Large document renditions must be available as soon as possible after submission for rendering. Each of the rendering modules can be configured on a separate Publisher computer dedicated to that rendering module. Or multiple Publisher computers can be configured with the same rendering modules and groups of users can be assigned different publishing jobs that specify which Publisher server to process their jobs.

For example, if Publisher job A specifies the AutoCAD rendering module and job B specifies the Inventor rendering module and only one Publisher computer is configured, rendering many large files can take considerable time and place a high demand on hardware resources as AutoCAD and Autodesk Inventor are repeatedly and alternately started to perform rendering.

Publisher supports clustering by allowing each Publisher computer in the cluster to run specific publishing jobs. All Publisher computers in the cluster share the same publishing database and Publisher Queue. One Publisher computer in the cluster, called the master, hosts the Publisher web service that delegates publishing jobs to the other computers, called subordinates, for rendering. The web services on the subordinates are not used unless near real-time publishing is configured as described in NEW Configuring near real-time publishing. The master Publisher computer can also act as a subordinate of itself and also perform rendering.

For example, all client computer job submissions and queue management functions are processed by the Publisher master computer. Jobs that specify Publisher job A are executed on the Publisher computer where job A is defined. Jobs that specify Publisher job B are executed on the Publisher computer where job B is defined, and so on for additional job definitions. Each Publisher computer executes its jobs asynchronously from the other Publisher computers, thereby allowing parallel rendering. Each Publisher job can specify the same or a different rendering module.

Note Publisher does not perform load balancing, also known as pooling, between the computers in the cluster. This means that the Publisher master computer cannot delegate publishing jobs to the next available subordinate computer. All job submissions for the same job definition will be executed on the computer where the job is defined, regardless of how many other computers are available. For this reason, if you want to distribute jobs between multiple Publisher computers, the jobs must be divided among separate source systems or divided among Publisher users (instruct groups of users to select different jobs in Meridian Enterprise).