DAH - Distributed Action Handler

Autonomy's Distributed Action Handler is a distribution server that allows you to distribute action commands (e.g. Query, Suggest, Summarize, Highlight etc.) to the ACI servers. This enables you to scale your system in a linear manner, increasing the speed with which actions are executed and saving processing time. Multiple copies of IDOL servers to which the Distributed Action Handler distributes actions furthermore ensure uninterrupted service in the instance that any of the IDOL servers should be interrupted due to hardware failure or network outage.

You can run the Distributed Action Handler in the following modes:

Distribution in Mirror mode

The IDOL servers that the Distributed Action Handler distributes ACI actions to are identical (that is all of the IDOL servers are exact copies of each other, each one is configured the same way and contains the same data).When running the Distributed Action Handler in Mirror mode it distributes actions using one of the following methods:

• Load Balancing
• Fail Over

Load Balancing

The Distributed Action Handler assigns each incoming action to just one of the IDOL severs that it is connected to (using a cumulative predictive algorithm that is designed to spread the action load efficiently). When this IDOL server™ responds with a result the Distributed Action Handler forwards it to the client software. If an IDOL server™ stops responding for any reason (for example, a hardware failure or network outage), the Distributed Action Handler marks it as "down" and only assigns incoming actions to the IDOL servers that are running, saving the time it would take to attempt to communicate with the failed IDOL server™.

Fail Over

The Distributed Action Handler forwards incoming actions to the first IDOL server™ that is listed in the Distributed Action Handler's configuration file. If this IDOL server™ stops responding for any reason (for example, a hardware failure or network outage), the Distributed Action Handler marks it as "down" and switches to the next IDOL server (the second one listed in the configurations section) and so on, thus seamlessly continuing to service client actions.

The Distributed Action Handler also periodically checks IDOL servers that are "down". If it finds that a IDOL server™ has come online again (for example, because the hardware has been rebooted or the network connection repaired), it adds it to the "active" list of IDOL servers. For Load Balancing, this means that the IDOL server™ becomes a valid choice for actions again. For Fail Over, this means that the first IDOL server™ takes over as the sole destination for actions (as it has precedence over the IDOL server™ that is currently serving results).

Distribution in Non-mirror mode

The non-mirror architecture is implemented when each Intelligent Data Operating Layer™ is configured differently and contains different data. Therefore, when running the Distributed Action Handler in this mode it will be necessary to set-up Virtual Databases.

Virtual Databases

The following types of IDOL server™ virtual databases can be set up in an Autonomy environment:

Combinator Virtual Databases

The Virtual Database forwards an action command to all the databases that it comprises. It collates and sorts the results before it returns them.

Distributor Virtual Databases

The Virtual Database forwards an action command to one of the databases it comprises. These databases must be identical (that is all of the databases are exact copies of each other and contain the same data). The way it forwards the action is determined by the distribution method.

When running the Distributed Action Handler in Non-mirror mode using Virtual Databases of the Distributor type (each of which is mapped to a set of identical databases that contain identical data), it distributes actions using one of the following methods:

Load Balancing

The distributing Virtual Databases assign each incoming action to just one of the IDOL servers™ databases that they map to (using a cumulative predictive algorithm that is designed to spread the action load efficiently). When this IDOL server™ database responds with a result, the Distributed Action Handler forwards it to the client software. If an IDOL server™ stops responding for any reason (for example, a hardware failure or network outage), the Distributed Action Handler marks it as "down" and only assigns incoming actions to the IDOL server™ databases that are running, saving the time it would take to attempt to communicate with the failed IDOL server™ database.

Fail Over

The distributing Virtual Databases forward incoming actions to the first IDOL server™ database that is listed. If this IDOL server™ database stops responding for any reason (for example, a hardware failure or network outage), the Distributed Action Handler marks it as "down" and switches to the next IDOL server™ database (the second one listed) and so on, thus seamlessly continuing to service client actions.

The Distributed Action Handler again periodically checks IDOL servers that are "down". If it finds that an IDOL server™ has come online again (for example, because the hardware has been rebooted or the network connection repaired), it adds it to the "active" list of IDOL servers. For Load Balancing, this means that the IDOL server's database becomes a valid choice for actions again.

For Fail Over, this means that the IDOL server's database takes over as the sole destination for actions (as it takes precedence over the IDOL™ database that is currently serving results).

DAH Benefits

Load Balancing

The DAH improves performance in very large systems by distributing service requests across several IDOL servers and delivers greater application flexibility and service transparency by combining different databases (Virtual Databases) of information into a single logical database.

Increased Action Performance

Due to the advantages of load balancing that the DAH provides, organizations can optimizes the Intelligent Data Operating Layer™ (IDOL™) performance, enabling more action to be processed through intelligent distribution.

Saving Valuable Processing Time

Autonomy's modular architecture further enables each ACI Server to be located on different machines and platforms. In this way the load per machine can be minimized, saving valuable processing time.

Linear Scaling

Facilitating distributed service requests allows performance to remain constant with growing data sets, further enhancing linear scalability.

High Availability

Including the DAH as part of the Autonomy infrastructure delivers the benefits of ensuring a high availability system that can send simultaneous service requests to multiple (replicated) IDOL server™ instances across a number of machines. Therefore, in the event of hardware failure, the DAH can continue sending service requests in fail-over mode to other IDOL servers guaranteeing total availability.

Autonomy in Global Infrastructure

The Distributed Action Handler enables the system to submit requests simultaneously to multiple IDOL servers, regardless of geographic location. DAH therefore complements a high performance infrastructure that provides location transparency to both the IDOL servers™ and the application software in a global infrastructure.

The DAH contributes significantly towards ensuring a software infrastructure that delivers both high performance and high availability within an automatic environment. Positioned as an ACI Option, the DAH is an add-on module to existing Autonomy infrastructure such as the Intelligent Data Operating Layer™, Classification Server™ and User Agent Server™, in addition to all Service modules (Categorizer, Mailer) that Autonomy offers.

Undoubtedly, with the inclusion of the Distributed Action Handler, organizations will immediately benefit from improved performance through load balancing, reliable location transparency, replication and redundancy advantages.