UK-based software house Conchango was behind much of the development of a Chevron/Microsoft flagship ‘SEER’ business intelligence application that has been deployed on the Captain Field in the North Sea and on the Californian San Ardo ‘i-Field’ (OITJ Nov 2006).
In a recent blog*, Conchango’s Jamie Thomson notes that business intelligence often relies on a complex network of point to point, inter-application data transfers and Excel spreadsheets. In such situations, control over data formats, integration and quality has effectively been lost.
To address these issues and provide Chevron with usable business intelligence (BI), Conchango has leveraged web services and internet standards to ‘wrap’ legacy applications, creating new, web enabled workflows. For service-oriented business intelligence (SOBI) to work, you need common data exchange formats, master data definitions and trusted systems of record (SoR) for critical data, common applications and interfaces.
Above all SOBI demands a huge investment in data governance. As Thompson puts it, ‘Without a willingness on the part of the customer to clean up their SoR’s, SOBI can’t work.’ Data owners can’t hide sloppy implementations or poor data quality behind a data warehouse. Data integrity starts at the SoR.
Data is made available to consumers via services—a.k.a ‘facades’ or business objects. Facades serve data in a manner that can be understood by data consumers. Data can be composited from one or more SoRs into a single facade. Data is usually cached above the facades in a project data store for before use.
XML-based wrappers or ‘facades’ allow data to be shared between disparate data consumers. Despite some initial performance concerns, XML data transfer is a keystone of the project. Thomson says that with the right infrastructure, the system can even support real-time BI.
Mapping across different systems of reference with typically, different names for the same well was a major problem for Conchgango’s developers. The solution was to build a new system to handle master data like well names, along with mappings to objects in the SoRs and to define a shared data structure that was suited all stakeholders.
Chevron suggested using the Public Petroleum Data Model Association’s (PPDM) data model for the master data store which proved problematical. The PPDM data model has some 1,500 tables and 11,000 database objects making it far too big for the job in hand. There were also issues with the large number of foreign keys and the size of some primary keys. It would have also been necessary to add tables for behavior-based safety (BBS) information, absent from PPDM.
In the end, Conchgango abandoned PPDM and went for an abstract data model consisting of only three tables. This provided the flexibility required to model all of Chevron’s master data.
Did it really take the advent of SOA to tell us that cleansed, unduplicated data and a rigorous taxonomy are necessary for enterprise BI or even information management? Probably not. But SOA and XML have at least given us visibility of each other’s data. And projects like Conchango are successful in getting the data issues fixed once and for all at source, rather than by every Excel user in the community, whenever new data comes in. For real-time, it’s the only way to go.
© Oil IT Journal - all rights reserved.