Avoiding SOA standards-based chaos
Can Mercury and Systinet help?
SOA (Service Oriented Architecture) is apparently nice and easy to understand - a bit of WSDL, some XML, SOAP, and a simple UDDI directory to select your services from.
So far so good – and then you find yourself in "standards based chaos", running services which don't do what they say on the tin; and even "rogue" services which you never wanted to run in the first place.
So you need testing – and then you find that your web application testing tools tend to assume you're testing a GUI, and individual services often don't have a GUI (composite applications do, and you can test them more-or-less conventionally, but the ideal is to have standalone services available).
What you need, possibly, is "service lifecycle management" tools, which can query a respectable UDDI-compliant registry and automatically build a test harness for your services based on what they find there.
Based on this, you can then integrate functional testing tools (do the services behave as they claim to in the registry?) and performance-testing tools (do they provide services fast enough?). This can the be carried forward into production monitoring, using the production version of the registry, proactively identifying emerging performance issues, and alerting operations to rogue applications (those not authorised in the registry).
One company that claims to support just this scenario is Mercury, which has built on its acquisition of Systinet at the start of 2006. In a comparatively short time, it has integrated Systinet (one of the more respected UDDI-compliant registries) with its Quality Centre and new Business Availability Centre, according to James Stevenson (VP and MD for UK Middle East and Africa at Mercury); thus addressing IT Governance for SOA in the round.
And Mercury's acquisition this summer by HP, with its OpenView infrastructure management framework complementing Mercury's existing application management (or "business technology optimisation") story, just makes its story stronger - although any merger this size must include some risk that those involved take their eye off the real ball (which is customer service).
Of course, since Mercury claims (with considerable justification) to be a leader in IT governance and application lifecycle quality management, it would be rather embarrassing if it hadn't managed to integrate Systinet quickly and efficiently, but such embarrassments do happen in the crazy world of IT. Although not (apparently) in this case - probably Mercury's run in with the SEC last year was embarrassment enough.
So, in an SOA-enabled world, I think Mercury's contribution to service governance will be welcomed. And the competition for this initiative? Please comment with (sensible) suggestions for tools you can actually buy now – not with possible UDDI-compliant competitors for Systinet (these are listed at www.uddi.org), but with holistic registry-based service governance frameworks.
However, I strongly suspect that the real competition is going to be the 80 per cent or so of companies that don't really automate testing or general defect removal in any systematic way at all. I sometimes wonder how they cope.
For the record, Mercury has announced a new version of Systinet 2; new Service Test and Service Test management products; a range of SOA management capabilities; and a strategy for Business Technology Optimisation (BTO) in the SOA space (see here).
And what is this BTO strategy for SOA? Well it's risk-based, applying IT governance to reduce risk rather than as an end in itself; and it's flexible and incremental (so you can have a green field SOA environment; or try out a few services; or introduce a full SOA environment to an existing installation). And it helps you plan, test, deploy, govern and manage SOA initiatives, which sounds fine, but the devil will be in the detail - probably why everyone is going SOA but few have got there yet.
As I've already implied, I think the key practical point of this announcement is the automated importation of services from the registry for the creation of test requirements and plans, and the automatic generation of test cases based on changes to services in the registry.
Companies are rushing to adopt SOA because, according to Toby Redshaw, vice president of IT strategy, architecture, and e-business at Motorola, it "makes our business more agile and efficient, resulting in faster project delivery and enabling our development groups to shift their focus from simply maintaining applications to composing services and applications to meet new business requirements".
Well, that's the promise. But Redshaw is certainly right when he says that, "to be successful at SOA, you need a foundation that supports governance, quality management, and management in production".
Mercury and Systinet are helping to provide that foundation. ®