An Interview with John deVadoss on Patterns and Practices at Microsoft
Like this article? We recommend
Like this article? We recommend
Like this article? We recommend
Larry O’Brien: Can you give our readers some sense of the size and scope of the "patterns & practices" group at Microsoft and your role within it?
John deVadoss: The patterns & practices group is part of the Developer Division at Microsoft and is tasked with delivering applied engineering guidance (patterns, practices) to help our customers build better applications. We find that customers that use our guidance achieve benefits in a number of areas, including increased developer productivity and improved application performance. Customers that use patterns & practices guidance tell us that it removes 30 to 40 percent of the work and lets them focus on their core business logic. It is my privilege to lead the patterns & practices team; and by the way, this is one of the best jobs at Microsoft, in my humble opinion.
Larry: Microsoft's target developer audience includes people of all levels of experience and sophistication. A few years ago there was a flurry of discussion about the use of developer "personas" called "Mort, Elvis, and Einstein," as an organizing principle. Are personas still used in that way and, if so, can you talk a little about the concept?
John: Our customers include people of all levels of experience and sophistication as you call it, and we continue to use personas to help us better understand our customers’ needs and aspirations and to create a shared context for us to ensure that we give them the best possible developer experience.
We constantly evolve these personas, keeping in mind the changing technology and business landscape—and the key point here is not a static or dynamic set of personas, such as you might have heard about and reference in the question, but that we are committed to meeting the needs of our developer community, across multiple levels of sophistication and experience, and across multiple technologies.
Larry: In addition to the varieties of experience and sophistication, Microsoft wants to support the development of every size of application from a handful of lines to codebases of multiple millions of lines of code. Does "p&p" try to serve the entire scope of development or do you focus on a particular size and type of development?
John: p&p’s customers span small business, mid-market and enterprise, as well as ISVs and SIs. In fact, there are many product teams inside Microsoft that build on our deliverables—the Exchange team and the BizTalk team are two that I will highlight today. Our goal has been and continues to be to help our customers build better applications, whether they are an ISV, an SI, a small or medium or large organization—or a product team inside Microsoft. If the guidance is relevant and applicable, then by all means please use it. Having said that, much of our guidance is targeted towards building line-of-business applications.
Larry: The Nucleus Research report on "p&p" (http://msdn.microsoft.com/en-us/practices/ee406167.aspx) says that "Developers adopting patterns and practices can expect an average productivity increase of 25 to 40 percent." That's a huge win. Can that be broken down at all into components, or does that type of success rely on a synergy between patterns, practices, sample architectures, and so forth?
John: While we can try to break this down into components, I would emphasize this is truly the result of a synergy across patterns, proven practices, usage of our Reference Implementations, our scaffolding and so forth.
You should definitely expect to benefit with respect to application planning and development, as well as application testing, along with improved application management (lower cost of maintenance for example), but I would also add that your mileage may vary with respect to the increased developer productivity—depending on your organization, your business domain, level of experience and your team/development culture etc.
Larry: When shouldn't people use "p&p" guidance?
John: That is a good question. As I highlighted earlier, p&p’s customers span small business, mid-market and enterprise, as well as ISVs and SIs – and in fact there are many product teams inside Microsoft that build on our deliverables. Once again, if the guidance is relevant and applicable then by all means use it. The caveat is of course, as Maslow said, "if you only have a hammer, you tend to see every problem as a nail."
Larry: The first principle of the Agile Manifesto (agilemanifesto.org) is "Individuals and interactions over processes and tools." As a software vendor, Microsoft obviously wants to promote its technologies, both APIs and general purpose tools like, say, SharePoint. Is there a tension between the Agile principles and commercial imperatives and, if so, how do you address it in the guidance coming out of "p&p"?
John: No, not at all. The manifesto actually says “That is, while there is value in the items on the right, we value the items on the left more.“ Tooling is valuable but you need to recognize that people and their process or approach is often more important. Now, tooling continues to have a role in enabling people and the process they are following. In many cases—for example distributed or large scale team environments—tools are a key part of the solution. We in the patterns & practices group have been using agile development techniques for over five years—and continue to share with the community at large many of the resources that we have used to make our teams more successful. I would also encourage readers to peruse Ade Miller’s (our development manager) talks on this topic here – and in particular this paper on Distributed Agile development at patterns & practices. By the way, Grigori Melnik, who is the lead PM for Enterprise Library was the Program Chair of Agile 2008.
Larry: Microsoft is always going to tend to overshadow others in the "Windows and .NET ecosystem." Should people expect "p&p" guidance to incorporate options from beyond the shores of Lake Washington or should they expect that the context of "p&p" guidance is "using Microsoft technologies"?
John: Our customers see us as the trusted source for guidance on the Microsoft platform, and hence our primary focus is on helping them build better applications using Microsoft technologies. Having said that, design and architecture patterns transcend technology platforms, and we believe that a meaningful proportion of our guidance can be of value to you even if you are not on the Microsoft stack.
Larry: What about with something like, say, NHibernate, where there's not an option from Redmond that's directly comparable?
John: Yes, good question. For example, we are working on best practices on using jQuery to build web applications—and as you are aware, jQuery is not built by a Microsoft product team. We continue to take these on a case-by-case basis, guided almost always by the input and prioritization from our customers (again via customer advisory councils and Codeplex).
Larry: Software project management seems driven by anecdote and very scant data. For two decades, process was driven by Barry Boehm's assertion (ref. Software Engineering Economics) that the cost of defect correction rises exponentially over the development lifecycle; Extreme Programming asked "what if it's cheap to correct a defect later in the lifecycle?" (ref. XP Explained) There's very little published data on the fundamental question of how programming talent is distributed—how good the best developers are compared to the median. Microsoft's Developer Division has been using Team Foundation Server for years now and the "p&p" Website says your guidance is based on working with teams working in the real world. Microsoft Research's Empirical Software Engineering and Measurement Group publishes regularly, but peer-reviewed journal articles are necessarily conservative and narrow in scope. To what extent does the guidance coming from "p&p" rely on data and to what extent on the intuition of experts?
John: At patterns & practices, we use an approach that we call Customer Connected Engineering (CCE). Customer Connected Engineering is a process for engaging customers during the planning, development, and release of deliverables. Instead of simply collecting customer requirements up front, or getting feedback after the fact, it’s continuous involvement of customers throughout our life cycle. By involving our customers in our process, we enable a greater degree of transparency and increase the probability of shipping what’s most critical to our customers. By partnering with customers, we improve our ability to understand end to end scenarios as well as priorities. By shortening the cycles of feedback we also improve our ability to learn and adapt our deliverables as we clarify the wants and needs of our customers.
At the heart of customer connected engineering is our customer advisory board for every program; a set of customers that influence and direct what we deliver. The customer advisory board helps identify relevant scenarios; validate, refine and prioritize our backlog, and provides timely feedback during planning and development. Using CCE means that customers better understand our tradeoffs and have more visibility into our process.
Larry: Microsoft does retain, does it not, a high-level pattern for structuring a product team (Product Unit Managers, Product Managers)? Does "p&p" include guidance on that type of organizational pattern?
John: Any guidance that we in patterns & practices have delivered with respect to organizational patterns has been in the domain of agile development. I would refer readers to this paper in particular to get insight into some of our proven practices inside p&p.
Larry: The "p&p" Website describes the process your group uses internally as a combination of "Scrum-style management with XP development practices." As an organizational Pattern, what are the Applicability and Consequences of that structure?
John: Agile approaches to software development naturally lend themselves to projects where the requirements are unclear as are the potential solutions. The Customer Connected Engineering approach used by p&p fits very naturally with Scrum and XP which both place a lot of emphasis on involving customers throughout the project; soliciting their feedback and incorporating it into future releases very rapidly.
In today’s business environment, being able to respond quickly changing customer requirements is of great advantage. Obviously this approach needs to be tailored to individual deliverables. For example a small team building a customer facing web site would approach this very differently from a company building device drivers. However, in each case there is great value in being able to frequently evaluate working software with customers.
Larry: What are some of the patterns coming out of "p&p" that you think are particularly notable?
John: Rather than describing it as ‘patterns that come out of p&p’, because we are not in the business of creating patterns for the sake of creating patterns, I would describe goal in life as helping our customers apply and derive business value from said patterns.
If I had to highlight some of our recent work, then I would certainly call out our guidance on applying the MVVM pattern, and in particular on building rich, composite applications using Silverlight and WPF. This is intended to build loosely-coupled composite applications. I am really proud of the work that we have done to make it easier to build applications using WPF and Silverlight, with a single code base.
I would also call out the ESB Toolkit that we released recently. By the way, with the release of version 2 of the ESB Toolkit, and given the maturity of the pattern, we have now transitioned this work to the Business Platform Division inside Microsoft.
And last, but not the least, I would call out Enterprise Library, our collection of application blocks designed to help developers with common enterprise development challenges. In particular, I am very proud of the work that we have done with the Unity application block, in delivering a lightweight, extensible dependency injection container that supports constructor injection, property injection, and method call injection as well as interception.
Larry: F#s been productized, Erik Meijer is giving a course on Haskell on Channel 9 (http://channel9.msdn.com/shows/Going+Deep/Lecture-Series-Erik-Meijer-Functional-Programming-Fundamentals-Chapter-1/), and C# and VB are gaining functional constructs. Would you say that object-orientation remains at the heart of "p&p"s guidance or are you incorporating functional programming? (Does this tie in with the earlier answers regarding data vs. intuition, cutting edge vs. tried-and-true, etc.?)
John: Our customers, in the form of our various advisory councils, as well as on Codeplex, tell us where to focus and what to work on. It is probably fair to say that our focus has predominantly been on object-orientation, and that is primarily because the many of our customers have tended to be focused here. While we have opinions and perspectives, and dare I say, biases—at the end of the day, the items in our backlog and their prioritization continue to be driven by our customers; and I expect this to continue in the domain of functional programming.
Larry: How disruptive will the many-core era be in terms of software patterns and practices? Do you anticipate something very dramatic in terms of languages and approaches or do you think it will be more incremental and evolutionary?
John: A great question. In fact, we have been working with a number of teams inside Microsoft on this very topic—and delivered a workshop on Patterns of Parallel Programming at the Microsoft Professional Developers’ Conference in LA.
The transition from single-core to multi and many cores is altering computing as we know it—enabling increased productivity, and powerful energy-efficient performance—presenting an unprecedented opportunity for us as developers to design new and innovative software experiences.
Our goal at Microsoft is to simplify parallel programming for both native and managed code developers to safely and productively build robust, scalable and responsive applications. Microsoft’s Parallel Computing Initiative is taking a comprehensive and integrated approach spanning solutions from local to distributed/cloud computing and from task concurrency to data parallelism. Microsoft plans to deliver a solution stack consisting of a concurrency runtime, programming models, language extensions, libraries and tools. Beta 2 of Visual Studio 2010 includes new programming models for concisely expressing concurrency. This includes new .NET Framework libraries such as the Task Parallel Library and Parallel LINQ, as well as the Parallel Pattern Library and Concurrency Runtime for developing native applications with C++.