Service Consumer A sends Service A a message containing a business document (1). The business document is received by Component A, which keeps the business document in memory and forwards a copy to Component B (3). Component B first writes portions of the business document to Database A (4). Component B then writes the entire business document to Database B and uses some of the data values from the business document as query parameters to retrieve new data from Database B (5).
Next, Component B returns the new date* back to Component A (6), which merges it together with the original business document it has been keeping in memory and then writes the combined data to Database C (7). The Service A service capability invoked by Service Consumer A requires a synchronous request-response data exchange. Therefore, based on the outcome of the last database update, Service A returns a message with a success or failure code back to Service Consumer A (8).
Databases A and B are shared, and Database C is dedicated to the Service A service architecture.
There are several problems with this architecture. The business document that Component A is required to keep in memory (while it waits for Component B to complete its processing) can be very large. The amount of runtime resources Service A uses to keep this data in memory can decrease the overall performance of all service instances, especially when it is concurrently invoked by multiple service consumers. Additionally, Service A can take a long time to respond back to Service Consumer A because Database A is a shared database that sometimes takes a long time to respond to Component B. Currently, Service Consumer A will wait for up to 30 seconds for a response, after which it will assume the request to Service A has failed and any subsequent response messages from Service A will be rejected.
What steps can be taken to solve these problems?
