Plan your Semaphore installation

Architecture planning and partitioning for Semaphore

When designing architecture for Semaphore, one of the most important considerations is resource usage of the Semaphore components. The demand components have on resources largely determine how the components are partitioned and the required specifications of their servers. It's worth noting that resource demand can be influenced by a number of factors, such as the number of documents processed and how often documents are created or modified. These factors should be considered when planning your architecture.

The following is the resource usage by component for a typical implementation:

• Semaphore Studio: Large amounts of RAM but low CPU (on average)

• Classification and Language Service (CLS): RAM and CPU activity is generally very high. As a result of the high resource demand, we recommend that you isolate this component as much as possible from other components.

• Semantic Enhancement Service (SES): Only used as requested by users so only uses resources as needed, however, performance is critical as this is, generally, the interface by which users other than Information Scientists directly access the Semaphore information.

Note: For performance reasons, on installation servers, anti-virus software configured for on-access virus scanning should be configured to consider how temporary and data files are stored by the Semaphore software. See Excluding Semaphore server data from virus scanning for further details.

Semaphore components can be installed on separate servers (See  Network firewall configuration for firewall considerations). By installing on multiple servers, the architecture can be designed in such a way to ensure that component execution does not interfere with other components installed on the same machine. For example, with three servers, an implementation for an average size implementation would take the following structure:

• Server A: Hosts Semaphore Studio, including Knowledge Model Management with the Publisher

• Server B: Hosts Semantic Enhancement Server

• Server C: Hosts Classification Server

This design ensures that the high activity of Classification Server does not interfere with the serving of Semantic Enhancement Server information. Maintaining the performance of the Semantic Enhancement Server is critical because its information is delivered to the end user.

• A few things to note when designing the architecture for your Semaphore implementation: For implementations that handle large numbers of documents, deploying the Classification Server on multiple servers is highly recommended.

• Although using load balancers is possible for a number of Semaphore components, they should not be used for Semaphore Studio or any of its sub-components, as the connection by clients to this application must be persisted.

• Progress Semaphore can assist in the design of any architecture required for a particular implementation. Contact Progress for more information.

Windows services created and used by Semaphore

During installation, the following services are created on the server:

Starting and stopping services

You can start or stop Semaphore services:

1. From the Windows start button, open the Control Panel for Windows.

2. Select Windows Tools.

3. In the Windows Tools window, select Services.

4. In the Services dialog opens, navigate to the Semaphore services. Note:

   • The names for semaphore services all start with Semaphore.

   • Ensure that the Startup Type for all the services is set to Automatic. This is the default.

5. Start/stop any services by selecting the services and using the controls on the ribbon menu.

Network firewall configuration

Any network configuration and firewall software should allow communication between the relevant external systems (for example, a CMS you may be using) and the Semaphore installation server. The following are the default TCP/IP ports used by the Semaphore components (note that all port numbers are configurable):

* Ports marked as Inter-Product, allow communication between Semaphore components; ports marked External provide services to external-facing applications, such as any product integrating with Semaphore; and ports marked Local are for sub-components within a single installed Semaphore component and on the same physical machine, so network/firewall configuration is not required.

Note: If Semaphore components are installed on multiple servers, then ensure that the relevant External or Inter-Product TCP/IP ports are open.

External triplestore products supported

Knowledge Model Management has built-in support that uses a Jena TDB triplestore, but it also supports the use of the following external triplestore products

• Marklogic 8.04 (or later)

MarkLogic Server Configuration

If using a MarkLogic Server to store model information, then the following is recommended for optimal performance:

• Use the latest version of MarkLogic available

• Ensure using fastest disk possible (e.g. local SSD) for database storage

• Use host machines with at least 4 cores

Install the product according to the MarkLogic documentation and tune your database for performance. After installation, ensure the following if wishing to use a high availability cluster configuration:

• Ensure the Triggers, Security and Schema databases (forests) are replicated

• Configure your model databases to use forests on all nodes and ensure that the forests are replicated across all nodes

Next steps

Install Semaphore