AJ's blog

March 29, 2009

Visual Studio 2010 Architecture Edition

Today I’d like to share another left over from the SDX Talk I mentioned earlier: Basically some screenshots from Visual Studio 2010 Architecture Edition (VSArch from now on). Don’t expect something fancy if you already know VSArch, I just couldn’t find all that much information on the Web beyond the two screenshots on the Microsoft site.

The main new things within VSArch include the Architecture Explorer, UML Support, and the Layer Diagram.

Architecture Explorer

Note: To make the following more tangible I loaded a sample project I use regularly as test harness and took respective screenshots while I analyzed it. Click the images for a larger view…

The Architecture Explorer is about getting a better view into existing code. Whether you join a project that is under way, whether you have lost control over your code, or whether you just need to spice up your documentation. Architecture Explorer helps you by visualizing your solution artifacts and dependencies. Artifacts include the classical code artifacts (classes, interfaces, etc.), as well as whole assemblies, files, and namespaces.

Architecture Explorer lets you select those artifacts, display graphs with dependencies, and even navigate along those dependencies and in and out of detail levels.

The following screenshot shows VSArch. The docked pane on the bottom contains the Architecture Explorer that acts as “navigation and control center”. This is where you select your artifacts and visualizations. It could certainly use some improvement from a usability perspective, but it does the job anyway.


The screenshot shows two different visualizations of the assembly dependencies in my solution, a matrix view and a directed graph. Just to stress the fact: This was generated out of the solution, by analyzing the project dependencies.

The next screenshot shows a mixture of various artifacts, including classes, interfaces, even files, across parts of or the whole solution.


Depending on what filters you set, this graph could give you a high level overview of certain artifacts and their dependencies. For example you could easily spot hot spots, like the one class your whole system depends upon. Or make sure the dependencies are nicely managed via interfaces and find undue relationships. Even spot unreferenced and therefore dead graphs.

Once you go one level deeper, you may want to cluster the artifacts by some category.


The image shows again artifacts and their dependencies, but this time grouped by the project to which they belong. It also shows what kind of relationship a line represents and lets you navigate along that dependency.

The Architecture Explorer should help getting a better understanding of your code. It helps you to detect code smells or may guide your refactoring.

UML Support

Yes, UML like in, well UML. Not extensively, but it includes activity diagram, component diagram, (logical) class diagram, sequence diagram, and use case diagram. I didn’t spend much time investigating them, just drew some diagrams in order to take the screen shots. Generally I can say that Microsoft can draw boxes and lines (big surprise here) but there is a lingering feeling that those diagram editors may not be finished yet (again, hardly surprising on a CTP).

Creating a new diagram is easy enough. Just create a new project of type “Modeling Project” and add an item:


Everything starts with a use case, so here is our use case diagram:


One can draw the diagram as he likes. As you can see from the context menu, there is something being worked on. Namely the “Link to Artifacts” entry shows the Architecture Explorer, yet I couldn’t quite figure out what’s behind this. Also note the validate entries which didn’t do very much, but we’ll see them later in the Layer Diagram.

Next on the list is activity diagrams:


Works as expected, no surprises, no hidden gems that I’ve found.

The same is true for the component diagram:


Just a diagram, no surprises.

The logical class diagram gets more interesting:


As you can see, it contains very .NETy stuff like enumerations. It also has these menu entries that hint on more to come in the future — right now the selected menu entry brings up the error message asking for a stereotype, yet I didn’t even find a way to set those. Also the editor may still need some work, e.g. one cannot drag classes in and out of packages.

As a side note: The relation between this logical class diagram and the already existing class diagram escapes me. At least they are a little redundant.

Next on the list is the sequence diagram. Rather than drawing one myself I reverse engineered the existing code:


Quite nice and again, used this way it can help you documenting or just plain understanding existing code.

Note: If you want to try that yourself, the CTP has a bug: You need to have a modeling project and at least one diagram before the menu entry “Generate Sequence Diagram” appears. And while you will be presented with a dialog asking what call depth to analyze, it usually works only for one level.

Layer Diagram.

Now for the most dreadfully looking diagram (though Microsoft has a more colorful one on its site…): Some boring connected blocks, meant to represent the layers of your architecture.


Actually this is one of the most interesting features for any architect and dev lead: It’s a living beast! 👿 

You can add assemblies as well as other artifacts to the bleak boxes. Afterwards you can actually validate whether the dependencies between those artifacts match or violate the dependencies implied by the diagram. In the screenshot you can see that I deliberately misplaced an assembly and consequently got a respective error. Using this feature an architect can ensure that all layer related architectural decisions are honored during development.

To conclude…

The Architecture Explorer is certainly a worthwhile feature and I also like the validation feature of the Layer Diagram. That’s certainly something new and not to be found in other products.

Generating sequence diagrams is nice but it remains to be seen whether this will allow roundtrip engineering. The logical class diagram doesn’t yet meet my expectations and it’s not quite clear to me how it will evolve. The other diagrams? Well, they just work. However in this group is nothing exciting for you if you already have another modeling tool like Enterprise Architect (no advertising intended, just happens to be the one I’ve used recently…). And a dedicated tool probably will provide a more complete UML coverage. UML 2.0 has 13 types of diagrams, including state diagrams, which is in my opinion the biggest gap in VSArch UML support.

Anyway, if that caught your attention and your interested in more details there are two options: One, download the CTP and try for yourself. Two, if you want it more condensed and avoid the hassle with a VPC, watch a video with VSArch at work. For that there are two links I can provide:

  1. Peter Provost’s talk at the PDC. Go to the timeline on the PDC site, search for TL15 and you should find “TL15 Architecture without Big Design Up Front”, which is about VSArch, despite the title. His talk was the role model for my analysis of VSArch, yet seeing it live could still give better insights.
  2. Visual Studio Team System 2010 Week on Channel 9 has a bunch of videos, especially the “Architecture Day” ones. “top down” and “bottom up” show VSArch at work.

The final question however will be if all those features are compelling enough to actually buy the … Visual Studio Team Suite (i.e. the “you get everything” package). Why not the Architecture Edition? Well, if you are a developer as well as an architect, the Architecture Edition lacks too much in the other areas. Given that there is usually quite a monetary gap between dev edition and team suite, that gap might very well be used to buy a 3rd party UML tool instead… .

That’s all for now folks,

kick it on DotNetKicks.com

March 27, 2009

Being parallel?

Filed under: .NET, .NET Framework, Software Architecture, Software Development — Tags: — ajdotnet @ 4:40 pm

As a follow-up to the last post, I’d like to review one question I got after the talks I held. That one was particularly interesting, because it highlights a certain aspect of the changes that we have to face with multiple cores and some misconceptions some people might have:

Doesn’t today’s multithreaded software profit from more cores automatically?

The answer is a definite yes and no. Server software like IIS or SQL Server has always been optimized for throughput (performance only secondary) and as long as the processing is CPU bound it will certainly put more cores to good use. That should result in better throughput (i.e. more requests handled at the same time), but not necessarily in better performance (i.e. the time to complete one request).

The more interesting observation however is on the client: Client applications follow different use cases for multithreading, actually mainly two of them:

  1. Avoid blocking of some task while processing something else. Examples include keeping the Windows UIs responsive (i.e. work around the UI’s thread affinity), Windows Services (their service control handler is invoked by the SCM and has to return in a certain amount of time).
  2. Be able to spend processing time on something valuable while one task is blocked waiting for something to happen. Blocking regularly happens during some kind of I/O, especially network calls. This is the „call WebService asynchronously and do something worthwhile until the call returns“ scenario.

Just different sides of the same coin actually. Please note that neither use case has been about actual performance, as in doing things really faster. The first one is about perceived performance by reducing latency for some favored task, the second one improves performance, albeit not by doing things faster but by avoiding unnecessary wait times. Anyway, this works quite well even (or rather especially) on single core machines.

Let’s dig deeper into an example: Say a lengthy CPU intensive calculation is triggered by a button. The time spend on updating the UI and doing the calculation is denoted in the following picture, the red line represents the executing thread:

The calculation is done on the UI thread, giving it peek performance, but at the same time freezing the UI until the calculation has finished. The typical “optimization” is putting the calculation on a second thread, e.g. using the BackgroundWorker component. That way, the UI keeps updating itself rather than degenerating to one of those white and unresponsive windows:

The UI and the calculation run on different threads (yet still on one core), thus the UI can update itself even during the calculation is still running. However, now it has to deal with that intermediary state (denoted by the striped blocks). And every time the UI thread uses the core, that time is lost to the calculation, so the time to completion will actually be longer than with A.

Now switch gears to our new, say quad core, machine…

What happens on a quad core is that both tasks now get executed in parallel:

As you can see, actual processing time is now back to the same it was with A, yet it manages to avoid freezing the UI as in B; again at the expense of having to deal with the intermediary state. However that’s about as good as it becomes. But the application doesn’t get faster, neither will a third or fourth core be utilized at all (other than ensuring that other applications can run there rather then interfering with the cores we just occupied).

So, while the potential in “classical multithreading” lies in shifting or arranging calculations of disparate tasks, the potential of parallelism with many cores lies in doing more calculations of CPU bound tasks at the same time. Like so:

This time the calculation has been split into parts and been distributed on different threads on top of the UI thread. This eventually caused better performance than A. However this is not the way today’s multithreaded applications are written, because on a single core it actually leads to worse performance, due to memory pressure if used excessively, thread context switches, synchronization and contention, etc..

The verdict…

Today’s multithreaded client software may profit from more cores, but to a far lower degree than one might think at first. And what’s more: It might even stop running altogether because actual parallelism opens the door for error conditions that were simply theoretical threats on a single core system but become a reality on multi cores. Something like accessing some memory while another core is still in the middle of an instruction doing the same.

PS: And by the way: the future is now!

I just started in a new project and got my new machine last week: An HP wx6600, with 2 quad cores @ 3GHz, 4GB 😀

This is a task manager screenshot, showing 8 nice little cores, sitting under my desk, waiting for me to send them on one or the other errand…

A single core is such a lonesome entity… 😉

That’s all for now folks,

kick it on DotNetKicks.com

March 21, 2009

Going parallel…

The so-called „many core shift“ is happening. It’s not a thing of the future, it’s not „just around the corner“, it has already begun. And it will change our developers’ life.

Last week we had some customer events, containing some talks about PDC and other stuff and how it will affect the near term future. Among other things I tried to describe the many core shift as well as its consequences. Curiously the audience was largely aware of the facts, yet the consequences had yet escaped the vast majority. So let me try to repeat the gist and see whether this is maybe a common symptom…

What’s the many core shift, anyway?

Moore’s Law states that the number of transistors on an integrated circuits doubles every two years. Until not long ago, and accompanied by more complex designs and higher clock speeds, that meant faster CPUs. This was sometimes called the „free lunch for developers“, because if one happened to write a slow application… not that anyone ever would 😉 , all he had to do was wait two years and it ran twice as fast.

However this evolution has reached its physical limits (clock speed, power consumption, etc.). Yet, still the number doubles… .
So instead of building faster and more complex CPUs, the manufacturers started placing more CPUs, read cores, on a chip. It started 2006 with Intel’s dual cores, today you won’t find a single core desktop machine anymore. High end consumer machines come with quad cores, and servers with 16 cores (delivered as 4 quad cores). Have a look at the extrapolation:


The “lower” line shows cores on a CPU, starting 2006 with 2 cores, while the steeper one assumes 4 CPU sockets. And just in case the conclusion escaped you: Five years from now we will have between 32 and 128 cores. And remember, we are talking “consumer grade stuff”, that is the box under your desk, not something special! Impressive?

So that’s the many core shift. But what does it mean?

Well, it probably means that today’s software runs a bit faster. Not much, mind you, certainly not the 32 times faster a 64 core machine is supposed to be compared to my dual core. Why is that? Well, have a look at the following task manager of a 64 core machine:

(That’s a fake of course, but have a look at Marks’s blog for a real one.)

Now look at your own desktop and count the open applications. Outlook, Word, perhaps PowerPoint, Internet Explorer, Acrobat Reader? OK, say half a dozen applications, add 10 more for the OS stuff actually doing something. That’s 16 applications, using the upper row of cores, perhaps even to 100% and yearning for more, while the other 3 rows just sit there and twiddle their thumbs. That sad truth is: Most of today’s applications simply are not capable of employing these cores appropriately. Consequence: In order to leverage these cores we have to change the way we write our software!

Two questions come to mind: Do we actually need that kind of processing power? And if so, how do we open it up?

Seriously, does the average user need 64 cores?

Well, yes he does. If for no other reason, he did need the increase in processing power during the last years, why should that change?

So, what does he need it for? Gamers are always at the forefront of processing power demand. We have the generally increasing demand in UI technology: 3D, animations, visual effects are becoming mainstream with windows, WPF and Silverlight. The trend to digital photography has had its effect on the demand for graphics software. On my dual core, DXO needs about 1 hour per 150 pictures, so there’s certainly room for improvement (I brought 2500 pics from my last vacation in Tanzania. Do the math 😉 ). Background encryption, compression, virus scanning, etc. also add up.

Even if you are an “ordinary business user”: Word just sits and waits for your input most of the time? Well, open a non trivial 100 page document and see how long Word takes for pagination, spell checking, or updating the TOC. Change the formatting and watch again. So while Word mostly does nothing exciting, there are „burst times“ when it could really need those cores.

And I did not even mention Visual Studio and the compiler yet…

How do I, Joe Developer, put 64 cores to good use?
And how do I make sure the app doesn’t degrade on an old dual core beyond reasonable limits?

Here we are right at the center of the problem: Multithreading is not exactly something new, we’ve had that for more than 20 years on PC’s now. So why do I even have to ask that question? It’s because we didn’t actually use multithreading within our applications if we didn’t have to. Because it’s laborious, error prone, awkward. You have to deal with thread synchronization, race conditions, dead locks, error management, communication between threads. You can’t debug it, tracing doesn’t help very much either. In short: It’s a pain in the …, well.

So let’s face it: Most developers have avoided multithreading altogether (perhaps the lucky ones). And those who did do multithreading probably did it just for optimizations in very distinct areas.

But what we need to leverage those cores is quite the opposite: We need multithreading to become mainstream, kind of ubiquitous. For that it needs to be easier to employ parallelism. Complexity has to be pushed out of our code into the platform. Somewhat like nobody thinks any longer about virtual memory (while some of us are old enough to remember the days of physical addressing).

In other words: In order to deal with the parallelization demands, we need new patterns, libraries, and tools.

Microsoft is going to give us a first delivery on that with the next Visual Studio 2010 and .NET wave. Optimized runtimes (e.g. the thread pool), better tools (e.g. the debugger) and not the least, new libraries, introducing new patterns (e.g. Task Parallel Library). There’s more in the unmanaged world (e.g. Parallel Patterns Library), more on the server side (CCR), more on the language front (F#), more in the research area (transactional memory).
Microsoft even devoted a whole “developer center” to parallel computing (look there for more details). And quite rightly so, because there is no single solution to parallelization, it comes in different flavors (e.g. data parallelism vs. task based parallelism) and we can expect further developments in this area in the future.

Also it’s noteworthy that the OSes, namely Windows Server 2008 R2 and Windows 7 which share the same kernel, can manage 256 cores. Compared to what they supported before this is quite a jump.


So, parallelization is here to stay and we are going to have to deal with it. If anything, the trend is going to accelerate. It’s reasonable to assume that eventually processor manufacturers will trade single core performance for number of cores, i.e. put more but less capable cores on a chip, in order to save power consumption (green IT and mobility being two other major trends).

Looking even further, the many core shift may reach a break even where standard desktop systems will cease to profit from additional cores (how parallel can you become after all?), the problems of memory access may limit the amount of cores. Asynchronous multi cores may evolve, e.g. having cores optimized for certain tasks…

See, there’s a lot to look forward to. Our profession certainly remains interesting 🙂

PS: See Being Parallel? for some thoughts on whether today’s multithreaded client software will profit from additional cores…

That’s all for now folks,

kick it on DotNetKicks.com

Create a free website or blog at WordPress.com.