Exam 5: SOA Design & Architecture Lab (S90-09A)

Services A, B, and C are non-agnostic task services. Service A and Service B use the same shared state database to defer their state data at runtime. An assessment of these three services reveals that each contains some agnostic logic, but because it is bundled together with the non-agnostic logic, the agnostic logic cannot be made available for reuse. The assessment also determines that because Service A and Service B and the shared state database are each located in physically separate environments, the remote communication required for Service A and Service B to interact with the shared state database is causing an unreasonable decrease in runtime performance. You are asked to redesign this architecture in order to increase the opportunity for agnostic service logic to be reused and in order to decrease the runtime processing demands so that performance can be improved. What steps can be taken to achieve these goals?

Service A is a task service that is required to carry out a series of updates to a set of databases in order to complete a task. To perform the database updates Service A must interact with three other services, each of which provides standardized data access capabilities. Service A sends its first update request message to Service B (1), which then responds with a message containing a success or failure code (2). Service A then sends its second update request message to Service C (3), which also responds with a message containing a success or failure code (4). Finally, Service A sends a request message to Service D (5), which responds with its own message containing a success or failure code (6). You've been given a requirement that all database updates must either be completed successfully or not at all. This means that if any of the three response messages received by Service A contain a failure code, all of the updates carried out until that point must be reversed. Note that if Service A does not receive a response message back from Services B, C, or D, it must assume that a failure has occurred. How can this service composition architecture be changed to fulfill these requirements?

Service Consumer A sends a message to Service A. Before the message arrives with Service A, it is intercepted by Service Agent A (1). which checks the message for compliance to Policy A that is required by Service A. If the message fails compliance, Service Agent A will not allow it to proceed and will instead write the message contents to a log. If the message does comply to the policy, it continues to be transmitted toward Service A, but before it arrives it is intercepted by Service Agent B (2), which validates the security credentials in the message header. If the security credential validation fails, the message is rejected and a runtime exception is raised. If the security credentials are validated, the message is sent to Service A. Upon receiving the message, Service A retrieves a data value from a database and populates the message header with this data value (3) prior to forwarding the message to Service B. Before the message arrives at Service B. it is intercepted by Service Agent C (4) which checks the message for compliance with two policies: Policy B and Policy C. Policy B is identical to Policy A that was checked by Service Agent A. To check for compliance to Policy C. Service Agent C uses the data value added by Service A. If the message complies with both of the policies, it is forwarded to Service B (5), which stores the message contents in its own database. You are told that Policy B and Policy C have changed. Also, in order to carry out the compliance check of Policy C, Service Agent C will now require a new data value from the Service B database. How can this service composition architecture be changed to fulfill these new requirements?

Service Consumer A invokes Service A (1). The logic within Service A is required to retrieve three independent data values from Services B, C, and D and to then return these data values back to Service Consumer A. To accomplish this, Service A begins by sending a request message to Service B (2). After receiving a response message with the first data value from Service B, Service A sends a request message to Service C (3). After it receives a response message with the second data value from Service C, Service A then sends a request message to Service D (4). Upon receiving a response message with the third data value from Service D. Service A finally sends its own response message (containing all three collected data values) back to Service Consumer A. Service Consumer A and Service A reside in Service Inventory A. Service B and Service C reside in Service Inventory B. Service D is a public service that can be openly accessed via the World Wide Web. The service is also available for purchase so that it can be deployed independently within IT enterprises. Due to the rigorous application of the Service Abstraction principle within Service Inventory B, the only information that is made available about Service B and Service C are the published service contracts. For Service D, the service contract plus a Service Level Agreement (SLA) are made available. The SLA indicates that Service D has a planned outage every night from 11 PM to midnight. You are an architect with a project team building services for Service Inventory A . You are told that the owners of Service Inventory A and Service Inventory B are not generally cooperative or communicative. Cross-inventory service composition is tolerated, but not directly supported. As a result, no SLAs for Service B and Service C are available and you have no knowledge about how available these services are. Based on the service contracts you can determine that the services in Service Inventory B use different data models and a different transport protocol than the services in Service Inventory A. Furthermore, recent testing results have shown that the performance of Service D is highly unpredictable due to the heavy amount of concurrent access it receives from service consumers from other organizations. You are also told that there is a concern about how long Service Consumer A will need to remain stateful while waiting for a response from Service A . What steps can be taken to solve these problems?