In his Modelling Overload blog entry, Piyush Pant has questioned the usefulness of SML and has wondered if SML addresses a problem that no one is trying to solve. Harry Pierson, my colleague at Microsoft, has responded by explaining the benefits of using a formal language for operational modeling.
SML is based on nearly 3 years of experience within Microsoft in using SDM for building products that solve real problems, and on the collective experience of other industry leaders: BEA, BMC, Cisco, Dell, EMC, HP, IBM, Intel, and Sun. SML has not been created in vaccuum; it is based on solid real-world scenarios and working code.
SDM is being used in VS 2005 to allow architects and developers to build models of their distributed systems and IT environments, and use these models to verify the deployment of a distributed system on a specific IT environment. Having personally witnessed the problems that crop up when a system is moved from development environment to test/staging/deployment, I think that this an important real world problem that needs to be solved.
The next versions of MOM and SMS will use SDM models. MOMv3 will use SDM models to capture the structure and relationships of monitored services, and SMSv4 will use SDM models to enable desired configuration management. Again, a modeling language that allows an adminstrator to model the desired configuration of a machine, and use that model for desired configuration management solves a real problem.
Sure, SML needs to be backed by working code for supporting above scenarios. You need a runtime or an SML model validator that can validate instances against some SML schema and Schematron rules, you need the ability to discover real-world data and use it to populate instances, etc. Most members of the SML WG have confirmed their intent to use SML in their products. Microsoft has already shipped VS 2005, and is working on using SML in next versions of MOM and SMS, the Server Manager UI tool for Longhorn Server, and the recently announced Service Desk product.