Sunday, November 14, 2010

WCF 4 - File Less Activation, Default Endpoints and Default Bindings

Started going through WCF 4 a few months back, however, its today I got a chance to write down something about it

So let me talk about something that was really interesting when I first read about it, File Less Activation of services.

In WCF 3.5 when you wanted to host a service in IIS, you had to go through, adding endpoints, bindings and also a SVC file so that IIS can pick the request up.

In WCF 4, this has been simplified so much, that you can get a service up and running in no time.
Lets take an example, I created a service Service1 in a namespace DefaultEndpointSVC, and this is my web.config file....

<configuration>
<system.serviceModel>
<serviceHostingEnvironment>
<serviceActivations>
<add service="DefaultEndpointSVC.Service1" relativeAddress="myService.svc"/>

</serviceActivations>
</serviceHostingEnvironment>
</system.serviceModel>
</configuration>

Now, if I go and deploy my service in IIS inside a virtual directory /DefaultEndpointSVC, I can access my service like this...

Note that I did not add any svc file called myService.svc, instead I have configure it in the serviceActivations element, the relativeAddress attribute specifies the relative .svc file and the service attribute specifies that service to activate when IIS gets a request for a "myService.svc" file and in our case, our Service1 will get activated.

Now, are we missing something here?....where is the endpoint tag? wasn't WCF all about the ABC (address, binding and contract)??

WCF 4 introduces the concept of DefaultEndpoints, that is if you dont configure an endpoint, WCF will add a default endpoint for you..., so now the question how does it do this..

WCF does this by looking at the addressing schema, in our case, we are accessing our service through http, and because we have not defined an endpoint in our config file, it will add a basicHttpBinding endpoint.

WCF has default binding for different transport protocols, for http the defualt binding is basicHttpBinding, for net.tcp it used netTcpBinding; you can get a list of the default bindings that WCF uses from the machine.config.comments.config file found in the folder
C:\WINDOWS\Microsoft.NET\Framework\v4.0.30319\Config\ ; within the section protocolMapping

In WCF 4.0, you can put your WCF configurations extensively in the machine.config file, so that it would effect all the services hosted in that machine. So, if all your services use wsHttpBinding, all you have to do is change the default protocol mapping to choose wsHttpBinding instead of the default basicHttpBinding like this...

<protocolMapping>
<add scheme="http" binding="wsHttpBinding" bindingConfiguration="" />
</protocolMapping>

Now, what if, for a new service that you are creating you need to expose it as basicHttpBinding, you put the above line your web or app config, this would override the machine level or the default protocol mapping.

If you have noticed, you would see there is no binding configuration, for each binding there is an associated default binding, if you want to change this at your application level, you just need to add that binding in the config file and you don't need to no longer associate the binding name on the endpoint (see config section below), if you place a netTcpBinding binding configuration, then any net tcp endpoint defined for that application will pick up that binding configuration, but you can still use named configurations for your endpoints.

The advantage of no longer needing to associate your binding configuration into your endpoint is powerful, now you can just add the binding configuration with the standard values for your applications into your machine config like this...

<netTcpBinding>
<!-- put your settings here -->
</netTcpBinding>

Notice that there is no name attribute, hence this binding configuration will be used in any service that is hosted in this machine that uses net.tcp endpoint but without a named binding.

The same concept is there for service behaviors, hence you can put a default service behavior again in your machine config file that can be globally used in all the services hosted.

You can also name this behavior and now if you specify the same behavior name in your application web.config, the settings for the behavior in the machine.config will be inherited.

Guess, that all I have time for now...need to go and finish up the movie I stared...

Sunday, November 7, 2010

Running .NET code on a 64 bit Machine

Needed some information on how to develop .NET application for 64 bit machine, so I was doing the usual, reading a little bit of it, so before I forgot, i thought of posting it here for future reference.

The main advantage of using a 64 bit machine over a 32 bit machine is that memory is not constrained to 4GB; in a 32 bit machine, the highest address that the CPU understands is around 4GB, but with a 64 bit machine, the address space become much more then this in the order of 2^64.

Now lets talk about running .NET code, code that is 100% managed has no problem when it executes in a 64 bit, this is due to the fact that the CLR takes care of this.

Going back to the basic, when you compile a piece of C# or VB.NET code, the compiler produces MSIL (Microsoft Intermediate Language) code, that contains enough information for the CLR (Common Language Runtime) to start execution, like meta data and the types used and so...

When you start a program the windows loader, peeks at the assembly and sees if the application needs the CLR to be loaded, and the version of CLR to be used and the platform (X64 or X84) the code was meant to run in.
and then the CLR is loaded, the MSIL would be executed, but this code has to be first interpreted (JITTED) into native code according to the current machines instruction set so that the CPU can understand and execute the program.

So, building your managed code in a 64 bit or 32 bit machine produces the same MSIL code, however generally for managed code ,its just when this code gets JITTED you need to worry about machine architecture.

When you install, the .NET framework on a 64 bit machine running a 64 bit OS, the installer installs both the version of the CLR, yes, there are 2 versions of the CLR (2 versions of the JIT as well), one for 32 bit and the other catering for 64 bit architecture. The windows loader will be responsible for choosing which CLR to be used. The loader chooses the CLR based on the information on the assembly the developer sets at compile time; if you note that when you build your .NET application in Visual Studio (2005 and higher), you can specify for which platform your are building your code against, possible values include, 64, 32, itanium and ANY

When you specify X64 for your build platform, then the loader will execute your executable in the X64 bit CLR, meaning that your MSIL code will be JITTED into X64 native code. This is same when you specify X32 for your build target.
By setting the above you are forcibly telling your loader which version of the CLR/JIT to be used, this would become very useful when you are loading some 32 bit DLLs into your application as well, which we will discuss in a few seconds...
When you specify "ALL" for you build target process, the loader will select the CLR/JIT according to the machine your code is running on, i.e if its on a 32 bit, the 32 bit CLR will be used and if its a 64 bit then the 64 bit CLR will be used.

Now lets discuss some important rules...
You cannot run a 32 bit DLLs with 64 bit applications in process or mix 64 bit and 32 bit assemblies in process.

Windows 64 bit OS was designed in such away that 32 bit applications can still run on it...64 bit version of Windows comes with an emulator called WOW64, and all 32 bit applications will be running on this emulator, as per MSDN there is not much of a performance implication on this, more on this here....
Something interesting to note is that, if you install Visual Studio on a 64 bit machine, it would install both the version of the CLR, however, Visual Studio is a 32 bit application, and hence will run on WOW64

So, if you are developing a pure 100% managed application you don't need to worry about porting your code to a 64 bit, an xcopy would just work fine.
However, you might need to review, if your application is...
1) using a third party dlls that are built for 32 bit machines
2) If you are using COM objects
3) using unmanaged code in your application.
4) Serelization of objects, this would be a problem, when you are sharing object state serlized from a 64 bit machine and consumed by a 32 bit machine, this can be overcomed by using XML serelization to an extent.

When porting .NET code from 32 bit to 64 bit, you need to review the above, do necessary changes, deploy and then test it out.

This is an old article that can be used as a guide.