About event management

["Durand, Jacques R." <JDurand@us.fujitsu.com>]

Here is some input from an off-line investigation from Dr. Cho, Dr. Woo and myself, about events:

(some other options have not been describd here)

 

1. Event Boards management:

 

- For efficiency reason, test scripts will need to deal with "small" event boards (or "Event Streams" per the EPTS terminology). For example, there could be an event board for each PurchaseOrder event, collecting all events related to this PO - i.e. sharing the same POref #.

 

- Input events may need to be automatically partitioned in such smaller Event Boards (or EB). The way EBs are implemented is abstracted: they could all be in the same event store, represented by a proper index (e.g. database index), or could be in separate event stores. The definition, creation and management of EBs is under control of the Event Manager (EM).

 

- A test script must know about EBs, can declare and create some, or reference existing ones managed by the EM. Some EBs are "read only", some others are "read-write".

 

- Event Adapters are mediating external events into Events that have proper structure for eTSM processing. E.g. they transform B2B messages into eTSM xml Events. They are acting as Event Sources (per the EPTS terminology).

 

- an Event Adapter is not concerned with specific Event Boards. It is associated with an Event Manager and sends events to it. the EM is in charge of assigning the event to the right Event Board(s).  Several Adapters can feed the same EM.

 

- If there is a need to separate event flows based on their source and not based on their content (e.g. different companies testing their own B2B platform which generate similar events) , then Company A and B will send their events to different event adapters (adaptA and adaptB). In a fully separated event processing, these adapters will feed events to different EM instances (EM-A and EM-B) that will create and manage two sets of EBs, which wil in turn be consumed by separate eTSM engines.

 

- a Test script should be able to work on different EBs in a generic way. E.g. a Test Case verifying that PO exchanges are conform to spec, must be able to work on a different EB at each invocation (e.g. depending on the PO ref #). The CATCH operator must be designed so that it can reference an EB in a parametrized way. For example, all EBs that collect events related to the same PO ref #, wil have an identifier of the form "POtrack:ref123" where the first part "POtrack" is a class of EBs (here the class of all EBs that track PO-related events), and second part "ref123" is a particular PO ref #. The Test script that verifies a single PO exchange will have a CATCH statement of the form: <catch ebclass="POtrack", ebinst="myPO/@ref">, or something equivalent.

 

 

 

 

2. "Script Container" design:

 

In the General design section of our wiki http://wiki.oasis-open.org/tamie/GeneralDesign

 

See last section:

Top-Level Script Containers (to investigate)

 

We attempt to define a structure that will provide some contextual info useful for overal test case execution.

This structure would contain also definition of Event types in use.

It may contain config info for for Event Adapters - i.e. contextual info that is assumed to be known from the Test environment, in order to run the test scripts.