Governance is the tail that wags the SOA dog. An organization that deploys Web Services without a solid governance program is headed for serious trouble in terms of reliability, security, and cost.

 While governance is a necessity for any IT endeavor, the open and potentially chaotic nature of SOA makes governance either a showstopper or a magic bullet for success, depending on how it's approached. This article will look at SOA governance from both a technological and business perspective, highlighting the need for a complete governance model - a closed loop that spans design time to runtime, policy definition, enforcement, and audit.

The Goal: Business Optimization through SOA
Encounters with today's endless discussions about Service Oriented Architecture bring to mind the old Hollywood phrase, "Cut to the chase." Forget the endless ruminations on containers, service buses, standards, and IDEs. What are we really talking about when we get granular with Web Services and their attendant minutia? We're talking about money.

SOA is about money, or at least, it should be. An SOA is a long-term technology initiative (it never really ends) that is meant to deliver a return on investment, whether you're investing hard dollars or equivalent resources. A non-profit with volunteer developers is just as much an investor in an SOA as General Motors.

In particular, SOA should be able to deliver on several of its well-hyped promises for business value, including a reduction in overall IT cost, business agility, and a faster application development and integration cycle. At the root of all these potential savings is the concept of software reuse. If a Web Service can be developed once, but reused many times across multiple consuming applications, then many cost reductions fall into place. The need to redevelop a comparable application drops off, and the cycle time for getting the new application up and running will be shorter if there's no development cycle to go through. Reuse also serves agility by enabling rapid extensions of systems into new uses based on dynamic shifts in strategic and operational requirements.

Getting There: A Path of SOA Governance
Reuse, the main driver of financial value from Web Services, is based on the highly flexible nature of Web Services. However, this flexibility is a double-edged phenomenon. The same openness and flexibility can also make Web Services chaotic and unreliable. How can you reach the financial results of reuse with such a potentially out-of-control mechanism?

Though it may seem counterintuitive, the path to the business benefits of SOA, to all that groovy openness and flexibility, leads right through a world of tight governance. The necessity of adopting strict but agile controls as a prerequisite for freedom is reminiscent of the classic Harry Truman line, "If you want to live like a Republican, you better vote for the Democrats." Only when you have a high level of confidence that your Web Services are well-managed, secure, and governed can you be confident that sharing them broadly won't cause more problems than the freedom is worth.

So, let's think old school IT governance and learn to govern our Web Services. SOA governance is a matter of getting your Web Services and overall SOA to behave in the way it's meant to, and preventing it from being misused. It's about control. If you design a Web Service for a specific purpose and set of consumers, you want to have an assurance (and the ability to audit) that it was only used for that purpose and consumer. And, you want it to be available, performing as intended, and secure. The transactions must have integrity, at least to the degree that you specify.

If you can't control your Web Service to this extent, you haven't governed it, and it will inevitably be misused, become unreliable and insecure. The risks of an ungoverned Web Service, and by extension an ungoverned SOA, include performance failures, security breaches, and breakdowns in data integrity. A low-performing, unreliable, and insecure SOA doesn't generate reuse or agility. It's just a costly pain in the neck. Let's not go there.

An SOA governance program delivers you from these risks. A good governance program will include provisions for the security and reliability of Web Services as well as sophisticated methods for controlling the reuse of Web Services. With governance in effect, you can exploit the reuse and agility potential of your Web Services. You can assign usage rights to Web Services to new consumers, confident that they're authorized and authenticated for use. You can meet security and reliability requirements set out for Web Services by the consumers you work with.

For example, if you develop a Web Service, and wish to extend it for reuse with a new business partner, you will likely want to be able to set, enforce, and monitor a service level for that Web Service. And, you'll probably want to establish, enforce, and monitor security policies that benefit both you and the partner, regarding the Web Service. Depending on your industry, you may need to encrypt or digitally sign the SOAP messages and establish an audit log of Web Service use and transactions for non-repudiation.

There are three main levers of SOA governance: policy definition, policy enforcement, and audit. To govern, you must be able to define the rules you expect to have followed. That's policy definition. And, to govern, you must be able to enforce the policies you define. It does little good to define policies if you can't enforce them. That would be like leaving a store unlocked and hoping that customers will have the decency to put money in the till on the honor system. Believe it or not, many IT governance programs rely to a degree on the honor system for policy enforcement. Your SOA doesn't have to be that way.

Finally, to satisfy various stakeholders that defined governance policies have in fact been enforced, you need a high-integrity audit record of what has gone on. Besides being able to audit the actual Web Service transactions, you also need to be able to show an audit log of the enforcement of the governance policies themselves. For instance, if you want to show a stakeholder that you specified that all SOAP message emanating from a specific Web Service must be encrypted, you have to be able to produce a bulletproof audit log of the establishment and enforcement of that policy.

Audit-worthiness is of interest to auditors of both the "capital A" and "small a" variety. There are auditors of various sorts in your company and its partners who need assurance that SOA governance is in place. There are also auditors (Capital A) who review your company's financial statements for the SEC and your internal controls for Sarbanes-Oxley.

It's only with these tight controls that the Web Service begins to have any reuse value. The challenge, of course, is to impose all this control without locking the whole SOA up so tight that it defeats its entire purpose. If you just wanted lockdown, you would have stayed with your existing systems. The trick, therefore, is to automate your SOA governance to the point that these controls are highly automated and seamlessly baked into the Web Service lifecycle itself. This is where the "loop" comes in.

Closing the Loop
Runtime SOA governance, the ability to define, enforce, and audit policies for Web Services at runtime, should be a benchmark of SOA maturity. Any serious enterprise SOA should be well governed at runtime. Of course, that itself is not always so easy, as Web Service endpoints and the "last mile" reaching from Enterprise Service Buses to the Web Services themselves can be out of the reach of many built-in ESB governance packages. Ditto for multi-vendor ESB to ESB mediation. However, assuming you've licked those challenges and adopted sound runtime SOA governance, you must still contend with the "loop."

The loop is the complete lifecycle for a Web Service from design time to runtime. The loop can either be "broken," meaning that you define policies at design time and hope they get enforced at runtime, or "closed," where definition, enforcement, and audit happen seamlessly in one solution. If you have two separate governance systems, one for design time and one for runtime, you might find yourself in a "define and hope" form of SOA governance, which is unsustainable and unreliable. Throw multiple platforms and enterprises into the mix and things can get downright out of control. "Define and hope," or loosely integrated design time and runtime governance systems, inhibit the realization of SOA business value.

The broken loop model of SOA governance carries with it a serious risk, which is the potential for a drop-off in governance between design time and runtime. Though this may not be a huge issue for a small organization, it can be a major problem in security and compliance terms for a large or complex operation. If the governance policies defined for a Web Service at design time can't be assured at runtime, and there's not an audit log, then the governance will be unsound. By extension, all of the reuse and agility you want from that Web Service will be very difficult to realize. At the very least, it will be quite costly to manually supervise the continuity of governance across the broken loop.

If you opt for a "closed loop" approach to SOA governance, you can define policies at design time and automatically enforce them at runtime. Closed loop SOA governance typically leverages the power of UDDI registry and a policy metadata repository that functions across the full life cycle of the Web Service. With a closed loop governance solution, your enforcement of defined policies will be seamlessly auditable from design time through runtime. With a closed loop approach, realizing the business value of SOA comes within reach. You have the ability to reuse Web Services and perform integrative acts of agility but stay confident that the governance policies you need to assure the integrity of your business remain in place.

Conclusion
SOA governance is a critical activity for an enterprise seeking to find the business return on its investment in Web Services. Though it may seem counterintuitive, the best way to find the financial upside in the openness and flexibility of Web Services is to implement tight governance. To work, though, the governance must not be overly restrictive of change. Otherwise, it would defeat the whole purpose of the SOA.

Best practices suggest that a comprehensive SOA governance solution should encompass both design time and runtime. Ideally, the solution should provide a "closed loop" approach that automates and connects the definition of governance policies at design time with their enforcement and audit at runtime. When the loop is closed, then the agility of SOA can take center stage with a minimum of governance overhead but a high level of confidence in reliability and security.