Category / Coding

by Aaron

Visual Studio 2010 Remote Desktop Performance Tips

I was disappointed, but certainly not surprised, to see that Microsoft employees blogging that Visual Studio 2010 performance over remote desktop would almost match, or be slightly slower than Visual Studio 2008 performance.

However, they got some good advice on how to optimize your RDP experience with Visual Studio.

http://blogs.msdn.com/jgoldb/archive/2010/02/27/optimizing-visual-studio-2010-and-wpf-applications-for-remote-desktop.aspx

Slow connections:

  1. On slow connections, change remote desktop connection settings to use 16-bit color.
  2. If running on high latency connections, select a “high latency” connection speed in RDC options:
    image
  3. On XP/Vista, grab the new remote desktop client, version 7.0.
  4. Disable all check boxes except Persistent bitmap caching.
  5. Also recommended is using a smaller window size (but that may be the most difficult to do).

WPF Application Optimization for Remote Desktop

In addition to the performance tips for Visual Studio, there’s a very nice set of best practices to follow for how to optimize any WPF application for use over remote desktop. Until the long-ago Longhorn promise of truly remoted-WPF applications support is added to the platform, we’ll need to optimize your WPF applications if they may be used over remote desktop.

The worst offender that I’ve encountered in the past is that animations, even hidden behind an opaque element, still cause that area of your application to be sent for each frame of the animation! (You might need to think about how silly that seems).

by Aaron

Resource Intensive WPF Progress Bar (animation)

I’m using a progress bar in a small WPF application I’m working on and noticed that the Private Working Set for my application seemed higher than I expected.

My application, once simplified down to it’s most basic element, consisted of:

Window, Grid, ProgressBar

On my Windows 7 x64 machine, running the application uses around 29MB (private working set).

Removing the animation only from the visual template drops the private working set to 19MB.

So, it appears that the animation alone causes an extra 10MB of private working set to be needed. Sorry, but that’s crazy (especially for my application)!

If anyone has any specific theories – speak up! Here’s the basic starting Window I created: 

<Window
    xmlns="http://schemas.microsoft.com/winfx/2006/xaml/presentation"
    xmlns:x="http://schemas.microsoft.com/winfx/2006/xaml"
    xmlns:Microsoft_Windows_Themes="clr-namespace:Microsoft.Windows.Themes;assembly=PresentationFramework.Aero" x:Class="PerfProgressBar.Window1"
    Title="Window1" Height="300" Width="300">
    <Grid>
        <ProgressBar x:Name="progressBar1" Value="100" />
    </Grid>
</Window>

I’m not going to post the entire ProgressBar template, but if you want to reproduce the problem (and see how the animation impacts the memory requirements, just comment out the “Animation” rectangle in the control template (and in the trigger as well): 

<Rectangle x:Name="Animation" Fill="{TemplateBinding Foreground}" Grid.ColumnSpan="3" Grid.RowSpan="2">
    <Rectangle.OpacityMask>
        <MultiBinding>
            <MultiBinding.Converter>
                <Microsoft_Windows_Themes:ProgressBarHighlightConverter/>
            </MultiBinding.Converter>
            <Binding Source="{StaticResource ProgressBarIndicatorAnimatedFill}"/>
            <Binding Path="ActualWidth" ElementName="Background"/>
            <Binding Path="ActualHeight" ElementName="Background"/>
        </MultiBinding>
    </Rectangle.OpacityMask>
</Rectangle>

<Trigger Property="IsIndeterminate" Value="false">
    <Setter Property="Fill" TargetName="Animation" Value="#80B5FFA9"/>
</Trigger>

Apparently, the ProgressBarHighlightConverter is intense (a little double checking in Reflector confirms!).

(All animations have some price of course – as an experiment I rotated the progressbar in a storyboard and that used about 7MB).

The moral of the story here really is to make sure you do some sanity checks against your [fill-in-the-blank] technology so that you understand how it uses system resources such as memory, CPU, etc, as it may have an impact on the success or failure of your application. I’ll leave out the gratuitous animations in my application so it doesn’t use memory unnecessarily.

by Aaron

Disabling automatic Sys.UI.Control attachment

If you’re using the Microsoft Ajax Library (learn), you may not always want to start the automatic “attach” process that takes place when the page loads. It’s easy to disable, but not yet documented any place I could find easily. 

<script src="Scripts/MicrosoftAjax/Start.debug.js" type="text/javascript"></script>
<script type="text/javascript">

    var ajaxPath = "";

    Sys.activateDom = false;

All you must do is set Sys.activateDom to false as shown above (make sure this is set after the new Start.js JavaScript file loads, otherwise your code will crash when you try to set the Sys object before it has been properly constructed).

Then, to begin the attach process, just call Sys.activateElements: 

Sys.activateElements(document.documentElement);

In the code line above, though I’ve specified that I want the entire HTML document activated, you could provide any element you want as a starting point (for example to optimize the use of the library and prevent unnecessary DOM searching for example).

I’m adding the delay in some JavaScript code because I wanted to set up a few variables in advance of the attach occurring. I tend to write my JavaScript code in an object oriented fashion these days (using the prototype pattern), including code that is interacting with the DOM. In this case, I’ll create a class that represents the logic of the page rather than following the typical purely functional model that is done on many JavaScript pages. But, when using the “eval” syntax of the Microsoft Ajax library “{{ code }}”, occasionally, I’ll need to delay the eval or the page will crash.

From my recent post on making a simple command extension to the Microsoft Ajax library, I wanted to make that more object oriented by referring to an instance of my class, rather than pointing directly to a function: 

<body sys:attach="wpc" 
    wpc:onbubbleevent="{{$view.onCommand}}"  
    xmlns:sys="javascript:Sys" xmlns:wpc="javascript:WiredPrairie.Commanding">

$view represents the instance of my page’s behavior. However, if the attach were to occur too early, this variable is not yet set. I’m using the slick script loading functionality of the ajax library, specifying the various JavaScript libraries and their dependencies, including my page’s behavior. It’s not until that JavaScript code is loaded that the code can create an instance – and that could be AFTER the page has already done the attach logic. The attach happens before Sys.onReady for example. (Sys.onDomReady happens before onReady, but not all JavaScript files may have been downloaded). 

Sys.onReady(function() {
    $view = new WiredPrairie.MainView();
    
    Sys.activateElements(document.documentElement);

When using the sys:attach attribute, note that the attach and instantiation process happens before any code you’ve specified in onReady is executed (Microsoft currently uses the same method for determining when everything is ready by adding a function call to onReady – but their call is first in the queue).

by Aaron

A key management issue.

If this isn’t a head-slapping coding-moment I don’t know what is…

Encryption busted on NIST – certified Kingston, SanDisk and Verbatim USB flash drives

The crack relies on a weakness so astoundingly bone-headed that it’s almost hard to believe. While the data on the drive is indeed encrypted using 256-bit crypto, there’s a huge failure in the authentication program. When the correct password is supplied by the user, the authentication program always send the same character string to the drive to decrypt the data no matter what the password used. What’s also staggering is that this character string is the same for Kingston, SanDisk and Verbatim USB flash drives.

I can’t imagine the security firm SySS’s reaction when they found this:

Analyst 1: “Ah Houston, we have a problem here. Dudes, this thing isn’t secure at all. It doesn’t even use my password for the encryption!”

Analyst 2: “Get the #$@!# out!! I’ve got 4 weeks scheduled to look at this thing. It’s day 2. No @$%!@# way.”

Analyst 1: “Seriously. It data is always encrypted with the same string.”

Analyst 2: “Cool, I get the next 4 weeks off.”