With a dozen years' experience at Microsoft, Satya Nadella has been put in charge of the company's Project Green initiative is to re- architect its business application offerings under a common, service-oriented architecture.
Recently, Nadella sat down with IDG and discussed where Project Green stood, outlined the roadmap for the initiative and described how Microsoft is developing its offerings for mid-market users.
With Microsoft CRM and acquisitions, you have all the elements of a mid-market ERP suite. Is that the plan?
Satya Nadella (SN): We got into this business through a series of acquisitions, and we did some homegrown development, such as Microsoft CRM. We have ERP products, Microsoft CRM and our small-business applications that are part of Microsoft Office. In ERP, we have Great Plains, Axapta, Solomon, Navision - those are the four major ERP brands for the mid-market.
How will these products evolve?
SN: People want things to be simpler, more flexible, and they want to drive down the total cost. But they also want lots of features within a given business domain. To make sense of all this, we first developed what we call the customer model. It has three elements. The first is people. (Users) need a bridge between their ad hoc communications and their more structured, transactional work. The second thing is ... business process complexity as defined by looking at an org chart. The number of people in a department sets the complexity, as opposed to the company size.
The last part is what we call work, or process. People in departments are working on some business process.
How will the move to a service-oriented architecture affect these programs?
SN: We found five horizontal attributes that customers are asking for. The first one is that end-users want simpler, task-oriented, role-based user interfaces that will help them navigate through information models they already have. Great Plains or Solomon have a pretty robust data model and object model underneath, but what (the user) is really saying is, "How are you going to help me get to the data I want?"
The next (attribute) is business insight. Yes, they want reports, but small and medium-sized businesses are really managed by exception. We call it operational BI.
The third piece is (being) connected. There is no such thing as a business application, an ERP application, living in isolation. The first level is to be able to open up our systems using Web services so you allow for these composite applications that can be built in a loosely coupled fashion using the new trends of SOA.
So the way we integrate between CRM and ERP is through an SOA-based approach to integration, which is a loosely coupled, asynchronous way to bring these systems together.
The next (attribute) is what we call adaptive process. All business applications today have gobs and gobs of business logic in code. The problem is, business processes are not static. What we've found is that in time, any business application gets out of sync with the actual process in the physical world, and that causes a lot of pain.
The real Holy Grail is to be able to take this thing that is written in code today and put it into a more modelled form. The (next challenge) is, how do we go into the system and start putting in models so we can increase the longevity of the system, and more importantly, how can we make the system more adaptive to change?
The last piece is the process centricity in our application design. That's where we're going, and that anchors our vision.
Is this where Project Green fits in?
SN: Project Green is one of those things that with a little help from us gets written up as different things by different people. It is a bunch of research we are doing on those design pillars I talked about. It is also actual product delivery of that research in the context of releases of Great Plains or Navision or CRM. Project Green is showing up in our products today. When we start taking the innards of the business logic of these apps and start putting models on them, putting them on a single model, that's when you'll start to see us having a convergence of our core code.
So, what is the product roadmap?
SN: Our roadmap is all about sequential progress on these five design pillars. You can measure (our progress) by the last release we did, and every 24 to 36 months we'll have another one.
How far are you willing to go with componentisation and disaggregation of applications?
SN: We absolutely believe in componentisation and disaggregation that doesn't break the final assembly. Just saying, "Let's take SOA and apply it mindlessly to the entire core application" (doesn't work). At the end of the day, there is a screen in front of the end-user where he wants to be able to post a transaction. You have to draw the granularity boundaries very well. Otherwise, you just have components that can't be assembled.
Won't this approach lead to commoditisation of components?
SN: The most important thing to me from a componentisation perspective is it allows me to make the systems I have much more agile to change. And that's the reality of business applications.
What challenges does this componentised world present?
SN: Is a Web services description an API, or is it not an API? If people assume Web services are just APIs, you call them like you called them in the past, then you build systems that are no different.
You have to build more message-oriented systems. You now need to think through the workflow and control logic in your applications so that you're resilient to message passing.
You can't have the control flow of your code be completely sequential and synchronous. You have to have a workflow, and you have to be in sync with it. That's a big mental shift.
Be careful, too, where you want to be asynchronous and message-oriented. You can't do a final transaction post in a journal in that form, because if you start doing that, you really are going to create all kinds of issues in terms of distributed transaction control and also the user experience.
Will componentisation enable users to go to best-of-breed applications and mix and match?
SN: I believe ... we will have more systems deployed in a decentralised fashion, and they will be easier to deploy and integrate. If they are not easier to integrate, it is easier for users to just buy one system. The beast that needs to be tamed in this case is all about integration cost.