In working to better understand the role of data in the service oriented world (SOA) it is often difficult to see the role. Data is a major base consumable like metals in car manufacturing. So in the example or metaphor for a car SOA is the standardization of the base components that make up a car (drive train, transmission, etc). When the manufacturer is building the vehicle, they will look to their component suppliers and say design to build this item. The manufacturer will only be loosely thinking about the base material consumption.
So the difficulty in Decision Support services is the need to be able to plan for how the data will be consumed. You may build a structure to support a service, but knowing the end use of the next layer up service is difficult. So if you don’t know how do you plan for capacity and performance?
In the Think Big conference Stephen Brobst briefly touched on this issue and here is what I took from his thoughts.
Think Integration of Decision Services to the Enterprise (exported). This is not a hard thing to think about, but when talking with many customers they start to worry about access management and tracability. So this requires:
- Middleware – Enterprise Application Integration, Enterprise Information Integration…
- Enterprise Solutions – SAP ESA, BEA Aqualogic
- Enterprise Platform – SAP Netweaver, BEA WebLogic, IBM Websphere and maybe even .NET
- Extreme Data Freshness
- Move your batch data transfer into more of messaging architecture. Messaging feeds data integration as the data is created.
- Up to date data
- Move data marts and operational data sources into the Enterprise data warehouse
- Governance and Stewardship
- Requester, Broker and Provider alignement
- Publish to the enterprise, so manage the data access at the service
- Reference Architecture at the business level
With this you can now support things like:
- Application Business Services
- Segment customers
- score products, channels, sales regions, offers
- Compute risk exposure, product profitability
- analyse inventory
- Event Notification Services
- Event detection – airline example or fraud detection
- polling vs triggering – travelocity is a great example
- Event Driven Business Processes
- Application Integration
- Web services integration