Getting Started Now on SOA for BI
Information Management Magazine, May 2007
Michelangelo once said, "Genius is eternal patience." For those of you who believe that service-oriented architecture (SOA) is genius, your eternal patience is about to be rewarded. According to software industry analyst David S. Linthicum, "Most activity around SOA has been limited to discussion, study, planning and small projects. 2007, however, will witness a significant surge in SOA spending as early adopters evolve proof-of-concept implementations into more robust deployments and late adopters buy into the architectural shift."1
SOA for business intelligence (BI) is following a similar adoption path, moving rapidly from prototypes and single projects to broader, enterprise-wide deployments.
Where is your company on the SOA for BI journey? Have you done your planning and proof of concepts? Have you advanced your learning curve for both front-end BI services, such as reporting and analytics, and back-end BI services, such as connectivity, transformation and integration? Do you have strategies for taking SOA for BI to the next level where SOA's interoperability and reuse benefits accelerate? Or, are you like many companies that are still unsure where to start?
Advertisement
SOA for BI is Accelerating
Ventana Research recently completed a survey of more than 488 business and IT professionals that evaluated the status and plans of SOA for BI in medium- and large-sized enterprises.2 More than 81 percent of respondents said SOA was important to BI due to its powerful combination of business and IT benefits. From the business side, nearly 50 percent felt that BI services would help make information more broadly available, while also improving the business's ability to respond faster to change. From the IT side, nearly 66 percent felt that BI services would help IT better respond to business needs, with another 33 percent viewing ease of integration and lower lifecycle management costs as key benefits.
Across business functions, customer operations (83 percent), finance (73 percent) and IT operations (73 percent) were identified as the highest value opportunities. When asked about deployments within the next 12 months, 43 percent of respondents cited customer operations and 32 percent cited finance as immediate focus areas.
SOA for BI covers both the front-end (visible to users) BI services and back-end (visible only to IT) BI services. More than half of the survey respondents planned to implement front-end query and reporting services by October 2007. In addition, approximately one-third of the respondents planned to implement other front-end services such as analytics, dashboards and alerting during the same time period. Regarding back-end BI services, also known as data services, more than half of the survey respondents planned to implement data services for source connectivity by October 2007, closely followed by transformation and integration data services.
Michelangelo also said, "Grant that I may always desire more than I can accomplish." Are these respondents too optimistic? Or do they recognize that the SOA for BI efforts pioneered over the past several years in both the vendor and user community have sufficiently evolved and are now ready for enterprise deployment? Do you share this optimistic view and aggressive implementation strategy? Or are you still in the exploratory stage?
When implementing SOA for BI, you must be careful to avoid common pitfalls. Surprisingly, these pitfalls are not technical. At this point, the technology works. BI vendors have SOA-enabled their BI front ends. Data services vendors provide robust data services middleware proven in large enterprise deployments.
The biggest hurdles for deploying SOA for BI are organizational. BI teams are typically user focused, although they rely on DBAs and other data specialists to support the back end. SOA efforts have been typically driven by new technology assessment teams, often with a focus on e-commerce or business process re-engineering. Frankly, infrequent collaboration is the norm across these groups.
Figure 1: SOA for BI Plans of Surveyed Businesses
Further, business imperatives and large project backlogs don't allow "risky" new approaches. Few companies can afford to allocate precious resources pioneering the intersection of SOA and BI. Even in the functional areas such as customer operations and finance, identified as the highest priorities, care for customers and cash are paramount.
In spite of these aggressive plans and proven technologies, many companies are still unsure about where to start their journey toward SOA for BI. The best way to avoid known pitfalls is to narrow your focus.
Department focus. Do not try to make SOA for BI work for the entire company out of the gate. Focus instead on one line of business or department (customer operations or finance), which has been listed as the highest priority. Once this department or line of business is successful, others will happily follow.
Project focus. Within a department, pick a project or two that you can slightly overstaff, providing enough slack to work through your learning curve. Again, once you have several successful projects, other project leaders will apply this new, now-proven approach.
Staffing focus. Select team members who like to try - and later evangelize - new technologies and approaches. Because the BI for SOA technology is relatively easy to learn and use, these attributes are more important than specific technical skills. And make sure these pioneers can stay focused on SOA for BI for at least 18 months to allow their expertise to develop and evangelism to spread to other projects and departments.
BI services focus. Do not try to execute on both front- and back-end BI services at the same time. While the survey seems to show a slight preference toward front-end BI services, focusing on the back end first is actually the best place to start.
Why prioritize back-end BI services ahead of front-end BI services? Source connectivity is the primary reason. Kathleen Wilhide and Dan Vesset from IDC sum up this issue as follows, "Of organizations with $500 million or more in revenue, 34 percent indicated that they had 10 or more operational source systems feeding their business analytic solution."3 Each of these many sources has its own access mechanisms, syntax and security. How can anyone support this kind of complexity? Before data services, IT would try to solve this problem by writing custom code, passing periodic extracts or even building various data warehouses. You can stop this one-off approach by using state-of-the-art data services middleware to write reusable services that access the required data in an optimized fashion.
Page 1 of 2.
