Thursday, October 30, 2008

VS2010 and Architecture

I had a little time left and started playing around with Visual Studio 2010 and I noticed some cool features. There's a new project type called the Modeling Project. I was a bit sceptical about this, because the modeling in VS2005 and VS2008 was not the experience I expected...

In VS2010 it is now possible to create Use Case diagram, Activity diagram, Component diagram, Sequence diagram and Layer diagram. Even versioning is possible with your favorite source control product. (TFS). Now I must have been asleep for a while or busy doing other things, but Microsoft has joined OMG, so the new UML diagrams are al UML 2.4! Now this I hadn't expected.

It's also possible to create Sequence diagrams from existing code, which is a big help if you're a new architect on a existing project. Talking about acrhitectional discovery VS2010 adds a new Architecture Explorer which creates Sequence diagrams, project dependency graphs and all sorts of handy diagrams, which can help you understanding the system.

From my point of view the Architecture Explorer is the best! It's a cool new tool in VS2010, which helps me get a better understanding of the system I'm working with. I can't wait to use it in real projects.

.NET Services and The Service Bus

The Service Bus is a new messaging infrastructure part of the new Windows Azure Services. It addresses a few Connectivity challenges we all are aware of:

  • IPv4 Address Shortage
    • Dynamic IP address allocation
    • Network address translation
  • Firewalls layered over firewalls over firewalls

So how does the service bus address these problems? There are a couple of components the service bus consist of:

- Naming: Because of the practical constraints of DNS (High update propagation latency, DNS names hosts and not instances etc.) the service bus naming system works in a different way i'm not going to dive into it now, but a few features are: R/W access with Access Control through the Service Bus Registry, Updates are reflected instantaneously and it names endpoints, not machines.

- Service Registry: The service registry is registry for service endpoints, not a general purpose directory. The registry is layered over the naming system and the services should publish themselves in the registry. The registry also provides the family of bindings for the service bus. Right now the service bus can be used with the following bindings:

  • BasicHttpBinding
  • WebHttp
  • WSHttp
  • WS2007Http
  • WSHttpContext
  • WS2007Httpfederationg
  • NetTcp
  • NetTcpContextBinding

So how does the service bus works with these bindings? If you use NetOnewayRelaybinding then the Receiver creates an outbound connect bidirectional socket and register in the cloud (Registry of the Service Bus) (TCP/SSL 828 outbound open required) The sender outbound connect one-way net.tcp (tcp/ssl 808/828) to the Service Bus. The sender sends a message the Service Bus creates a route to the receiver.

The NetEventRelayBinding allows muliple receivers, so you can send a message and both/more receivers will receive the message.

NetTcpRelayBinding / Relayed is the one you should almost always use, because it got the highest throughput. A connection starts just like NetOneWay, but instead of connecting to the Service Bus we connect to the load-balanced front-end nodes. These nodes are some sort of Socket-Socket Forwarders the nodes sends a control message to the cloud and the cloud sends a control message to the receiver. The receiver then connects to the socket-socket forwarder and then the receiver and sender are connected through the socket-socket-forwarder.

The last part I haven't discussed is security, this is where the Relay Access Control Model comes in.

- Relay Access Control Model:

RECEIVER MODEL: The receiver goes to the acces control service (azure) and acquire access token and asked for a claim for listen permissions on the service bus relay. Then this access token with subscription is passed to the service bus relay. The service bus relay will evaluate that token.

SENDER MODEL: The same way: get a token send it to the relay, relay evaluate, and now the relay will pass the normal message to the receiver.

There's even a possibility for you if you don't want to use the Access Control Service just add the following line in your BindingConfig <security relayClientAuthenticationType="None"/>. I don't think it's a good idea, but you can.

Wednesday, October 29, 2008

"Huron" Sync enabled cloud data hub

To my surprise the Microsoft Sync Framework team has teamed up with the SQL Services Lab to produce "Huron".

"Huron" is a combination of SQL Data Services and Sync Framework. Which actually means that an arbritary database can be published to the SQL Data Services (which is a part of the Windows Azure Services) and after that Sync Framework kicks in.

For example: You have some sort of SQL Database which contains information that you want to share with your colleagues (on mobile devices for example). It's fairly easy to publish this database to the "cloud" (SQL Data Services). The Data Service will store your database as a blob in the cloud. After you've uploaded your database, sync framework does a first sync of the data in the database. Another user can subscribe to your database, this means that a template of the database is downloaded to their local drive and sync framework does a first sync of al the data. This user also get a range of primairy keys he can use for adding rows. Which means that a user can add a row and the primairy key of that row will not be the id of the last row + 1, but will be the first number in the primairy key range. This little trick is done to reduce the sync logic. (Two users adding a row which gets the same id, sync framework can't determine which row is should be persisted in the real database).

Everyone who have been working with Sync Framework won't see anything new here accept the fact that the database is now out in the "cloud" and not in the company. So actually nothing new here, just a hype to sell Live Services. I understand that for smaller companies which don't have the money to provide the cost of administration, deployment and availability this could be a good solution. For bigger companies who already have the network and the maintenance departments (and the budget) for supporting these kinds of technologies, this can be an option if you want your data up in the "cloud", but I think there are better options.

First contact with Oslo

The first presentation about the language Oslo was by Don Box and David Langworty. It was really good to see that I don't have to know T-SQL anymore!

Why "M"

  • We want creating and interacting with oslo content to be simple and natural
    • Having a box-and-line design experience is an important enabler
    • Having a complementary textual experience is equally important
  • M is how we achieve the latter

What is "M"

  • "M" is a language for defining domain models and textual domain-specific languages (DSLs)
  • M domain models define schema and query over structured data
    • Values, constraints, and views
    • Natural projection to SQL
  • M DSLS define projections from Unicode text to structured data
    • Rule-based transformation
    • Grammar driven text editor integration

What "M" is not

  • An object-oriented language
    • No polymorphism, virtual dispatch
    • "ls-a" determined based on structural subtyping, not stipulation
  • A data access technology
    • M domain models compile down to T-SQL
    • Tool chain supports course-grained loading/unloading of schemas and values - not an OLTP solution
  • A replacement for T-SQL
    • Far less expansive feature set
    • Tool chain supports linking/invoking T-SQ

So Oslo is an abstraction on T-SQL and although it is not finished yet, it is already fun to play with. A little example:

First to start with a normal table:

Module Contacts

{

People

{

{ Name="persona", Age=29 },

{ Name="personb", Age=28 }

}

}

This actually means that we defined a schema and filled the table with two rows.

Next up constraints:

Module Contacts

{

-- Constraints:

People :

{

Name : Text where value.Count <= 30;

Age : Integer32;

}*;

People

{

{ Name="persona", Age=29 },

{ Name="personb", Age=28 }

}

}

What you see here is a constraint on the People table, which in T-SQL would look like this:

Create table [Contacts].[People]

{

[Age] int not null,

[Name] nvarchar(28) not null

}

It's still another language to learn, but I like this more then T-SQL.

Tuesday, October 28, 2008

WPF on Windows 7

Scrott Guthrie just presented some new features in WPF on Windows 7. They added some more interesting controls like a jumplist. A jumplist add some contextual support for your app. So when you right-click on your app in the taskbar, it shows the jumplist with it's available actions. They also improved the theme support and Deep Zoom will be in the WPF 4.0!

They enabled multitouch, which you can use to make Surface-like applications (if you got the hardware!)

Just to prove they believe in WPF (How could they not) they even built Visual Studio 2010 on WPF! This means that it's going to be really easy to extend Visual Studio specially on the visualisation part.

Today there will be a new WPF toolkit out, with a new datagrid (finally), datepicker, calender, ribbon and a preview of the Visual State Manger available in WPF in .NET 4.0. The Visual State Manager introduces two concepts: State transitions and visual states. A control, such as a datepicker, has multiple visual states that define how the datepicker should look like when the mouse is hovering over it, or when it is pressed etc. Transitions are used to define how visuals dynamically change as the control moves from one state into another.

The new WPF Toolkit will be available here.

WF 4.0 vs WF 3.0

A fast blog for now: the best improvements for WF 4.0!

Activity:

  • Authoring is simpler and takes much less code
  • Fully declarative workflows and activities
  • Alignment across Expressions, Rules, and Activities
  • Seamless Composition Across Flow Styles

Runtime

  • 10-100x Performance Improvements
  • Full control over persistence
  • Flow-in Transactions
  • Partial Trust Support
  • Integrates with WCF, WPF, ASP.NET

Tools

  • Designer Performance and usablility
  • Rehosting Improvements
  • Unified Debugging Experience

I've seen the new Designer and it rocks! It looks better and even has a new zoomfunction which works much better then the old! But they added more they added external Arguments you can assign which goes in and out. And my personal favourite it even shows which variables are in scope! And it's possible to use expressions on those variables.

WF 4.0 is a part of "Dublin" for a complete overview on "Dublin" check it out here!

Visual Studio 2010 CTP

Visual Studio 2010 and .NET Framework 4.0 CTP is out!