Architectural Overview

Semaphore Overview

Semaphore is the name for a suite of products distributed by Progress Software. This suite is tailored for specific implementation requirements but generally provides:

• Storage/management of taxonomical/ontological information
• Automatic classification services of content
• Enhancement of search facilities based upon taxonomical/ontological information

Semaphore can be integrated into, effectively, any CMS solution and search solutions. Progress Software itself provides specific integrations for SharePoint and Solr.

Conceptual Architecture

This diagram shows the conceptual architecture of a Semaphore installation in terms of how it interacts with external systems.

Typical Physical Deployment

This diagram shows a typical installation of the major Semaphore components across a multiserver installation. The default ports for access are shown.

Note: the full list of ports that are required are listed here (Linux) or here (Windows).

Technical Architecture

Most components can be installed on separate physical servers for load-balancing/partitioning purposes.

Please see the Architectural Overview (Windows) or Architectural Overview (Linux) for details on the individual components.

Service Access Requirements

In the above diagram the “Semaphore Server” is logical rather than necessarily physical construct in that the components can be installed on different machines, as required. There are, however, specific limitations on some of this configuration.

Service Component Access Requirements

The following are the specific Semaphore server component architectural limitations:

1. Semaphore Studio - Must be accessible by all users wishing to be able to update the model or access other studio tools such as “Document Analyzer”. The system on which it is installed must have network access to all Semaphore services it accesses (such as Classification Server and Semantic Enhancement Server).
2. Classification Server - Listens on a configured TCP/IP port (by default 5058) this must be accessible across the network by any integration wishing to use the automatic classification services it provides (and must also be accessible by the Publisher application).
3. Semantic Enhancement Server - Listens on a configured TCP/IP ports (by default 8983 and 9983). 8983 is available to service calls from clients. 9983 is used by the Publisher to create new indexes and so should be visible to the machine on which the Publisher or Publisher Plugin is run.

Text Miner

Text miner is a Windows client application that may be required for some Information Scientists in some instances. This can be installed on a Windows machine. It must be able to see the Classification Service running on port 5058 of the Classification Service machine. Note that from Semaphore 5.4 there are text-mining features being included directly in the KMM and Classification products.

Backing up of Semaphore Resources

Knowledge Model Management

It is vitally important that the models created within Semaphore Knowledge Management Tool are backed up in case of system failure. If using an external triplestore, the backup and restore functionality of that triplestore can be used. Otherwise, we recommend using the backup/restore functionality from within KMM. This can be automated using the KMM-Client tool - see here for more details.

Semaphore Studio

The Studio configuration should be backed up. This will only be necessary occasionally as the data will change infrequently. This is done by backup up the Studio/data directory.

Classification and Language Server

The data for the Classification and Language Server is all created whenever a model is published to that server. If a short amount of downtime is allowed in the case of a system failure, then the easiest way to recover is to re-publish the models.

However, if a more rapid recovery is required, then the rulebase directory of the Classification Server install should be backed up and restored before the failover CS instance is restarted. If multiple Classification Server instances are being used, then the can either share their rulebases directory, or in the case of a failure of one of them, then any instance can be copied to recreate a new instance.

If the Classification Server configuration is edited in a particular instance, then this should be backed up in case of system failure.

Semantic Enhancement Server

As with the Classification Server, the best way to restore indexes in the case of a system failure is to republish any generated models.

However, the data itself can be backed up by backing up the Indexes directory of the installation (typically C:\ProgramData\Smartlogic\SES\Indexes on Windows and /var/opt/semaphore/SES/data/Indexes on Linux) as well as the configuration directory (C:\Program Files\Smartlogic\SES\solr\server\solr\zoo_data on Windows and /var/opt/semaphore/SES/data/Zookeeper on Linux).