1. Object Initialization
A page's controls (and the page itself) are first initialized in their raw form. By declaring your objects within the constructor of your C# code-behind file, the page knows what types of objects and how many to create. Once you have declared your objects within your constructor, you may then access them from any sub class, method, event, or property. However, if any of your objects are controls specified within your ASPX file, at this point the controls have no attributes or properties. It is dangerous to access them through code, as there is no guarantee of what order the control instances will be created (if they are created at all). The initialization event can be overridden using the OnInit method.
2. Load Viewstate Data
After the Init event, controls can be referenced using their IDs only (no DOM is established yet for relative references). At LoadViewState event, the initialized controls receive their first properties: viewstate information that was persisted back to the server on the last submission. The page viewstate is managed by ASP.NET and is used to persist information over a page roundtrip to the server. Viewstate information is saved as a string of name/value pairs and contains information such as control text or value. The viewstate is held in the value property of a hidden <> control that is passed from page request to page request. As you can see, this is a giant leap forward from the old ASP 3.0 techniques of maintaining state. This event can be overridden using the LoadViewState method and is commonly used to customize the data received by the control at the time it is populated.
3. LoadPostData Processes Postback Data
During this phase of the page creation, form data that was posted to the server (termed postback data in ASP.NET) is processed against each control that requires it. When a page submits a form, the framework will implement the IPostBackDataHandler interface on each control that submitted data. The page then fires the LoadPostData event and parses through the page to find each control that implements this interface and updates the control state with the correct postback data. ASP.NET updates the correct control by matching the control's unique ID with the name/value pair in the NameValueCollection. This is one reason that ASP.NET requires unique IDs for each control on any given page. Extra steps are taken by the framework to ensure each ID is unique in situations, such as several custom user controls existing on a single page. After the LoadPostData event triggers, the RaisePostDataChanged event is free to execute (see below
4. Object Load
Objects take true form during the Load event. All object are first arranged in the page DOM (called the Control Tree in ASP.NET) and can be referenced easily through code or relative position (crawling the DOM). Objects are then free to retrieve the client-side properties set in the HTML, such as width, value, or visibility. During Load, coded logic, such as arithmetic, setting control properties programmatically, and using the StringBuilder to assemble a string for output, is also executed. This stage is where the majority of work happens. The Load event can be overridden by calling OnLoad.
5. Raise PostBack Change Events
As stated earlier, this occurs after all controls that implement the IPostBackDataHandler interface have been updated with the correct postback data. During this operation, each control is flagged with a Boolean on whether its data was actually changed or remains the same since the previous submit. ASP.NET then sweeps through the page looking for flags indicating that any object's data has been updated and fires RaisePostDataChanged. The RaisePostDataChanged event does not fire until all controls are updated and after the Load event has occurred. This ensures data in another control is not manually altered during the RaisePostDataChanged event before it is updated with postback data.
6. Process Client-Side PostBack Event
After the server-side events fire on data that was changed due to postback updates, the object which caused the postback is handled at the RaisePostBackEvent event. The offending object is usually a control that posted the page back to the server due to a state change (with autopostback enabled) or a form submit button that was clicked. There is often code that will execute in this event, as this is an ideal location to handle event-driven logic. The RaisePostBackEvent event fires last in the series of postback events due to the accuracy of the data that is rendered to the browser. Controls that are changed during postback should not be updated after the executing function is called due to the consistency factor. That is, data that is changed by an anticipated event should always be reflected in the resulting page. The RaisePostBackEvent can be trapped by catching RaisePostBackEvent.
7. Prerender the Objects
The point at which the objects are prerendered is the last time changes to the objects can be saved or persisted to viewstate. This makes the PreRender step a good place to make final modifications, such as changing properties of controls or changing Control Tree structure, without having to worry about ASP.NET making changes to objects based off of database calls or viewstate updates. After the PreRender phase those changes to objects are locked in and can no longer be saved to the page viewstate. The PreRender step can be overridden using OnPreRender.
8. ViewState Saved
The viewstate is saved after all changes to the page objects have occurred. Object state data is persisted in the hidden <> object and this is also where object state data is prepared to be rendered to HTML. At the SaveViewState event, values can be saved to the ViewState object, but changes to page controls are not. You can override this step by using SaveViewState.
9. Render To HTML
The Render event commences the building of the page by assembling the HTML for output to the browser. During the Render event, the page calls on the objects to render them into HTML. The page then collects the HTML for delivery. When the Render event is overridden, the developer can write custom HTML to the browser that nullifies all the HTML the page has created thus far. The Render method takes an HtmlTextWriter object as a parameter and uses that to output HTML to be streamed to the browser. Changes can still be made at this point, but they are reflected to the client only.
10. Disposal
After the page's HTML is rendered, the objects are disposed of. During the Dispose event, you should destroy any objects or references you have created in building the page. At this point, all processing has occurred and it is safe to dispose of any remaining objects, including the Page object. You can override Dispose, as well as Render by setting the appropriate selection in the object parameter.
The following are the some of the guidelines to create a good ASP.NET application.
* Disable session when not using it. This can be done at the application level in the "machine.config" file or at a page level.
* The in-proc model of session management is the fastest of the three options. SQL Server option has the highest performance hit.
* Minimize the amount and complexity of data stored in a session state. The larger and more complex the data is, the cost of serializing/deserializing of the data is higher (for SQL Server and State server options).
* Use Server.Transfer for redirecting between pages in the same application. This will avoid unnecessary client-side redirection.
* Choose the best suited session-state provider - In-process is the fastest option.
* Avoid unnecessary round-trips to the server - Code like validating user input can be handled at the client side itself.
* Use Page.IsPostback to avoid unnecessary processing on a round trip.
* Use server controls in appropriate circumstances. Even though are they are very easy to implement, they are expensive because they are server resources. Sometimes, it is easier to use simple rendering or data-binding.
* Save server control view state only when necessary.
* Buffering is on by default. Turning it off will slow down the performance. Don't code for string buffering - Response.Write will automatically buffer any responses without the need for the user to do it. Use multiple Response.Writes rather than create strings via concatenation, especially if concatenating long strings.
* Don't rely on exceptions in the code. Exceptions reduce performance. Do not catch the exception itself before handling the condition.
// Consider changing this...
try { result = 100 / num;}
catch (Exception e) { result = 0;}
// to this...
if (num != 0)
result = 100 / num;
else
result = 0;
* Use early binding in VB.NET and Jscript code. Enable Option Strict in the page directive to ensure that the type-safe programming is maintained.
* Port call-intensive COM components to managed code. While doing Interop try avoiding lot of calls. The cost of marshalling the data ranges from relatively cheap (i.e. int, bool) to more expensive (i.e. strings). Strings, a common type of data exchanged for web applications, can be expensive because all strings in the CLR are in Unicode, but COM or native methods may require other types of encoding (i.e. ASCII).
* Release the COM objects or native resources as soon as the usage is over. This will allow other requests to utilize them, as well as reducing performance issues, such as having the GC release them at a later point.
* Use SQL server stored procedures for data access.
* Use the SQLDataReader class for a fast forward-only data cursor.
* Datagrid is a quick way of displaying data, but it slows down the application. The other alternative, which is faster, is rendering the data for simple cases. But this difficult to maintain. A middle of the road solution could be a repeater control, which is light, efficient, customizable and programmable.
* Cache data and page output whenever possible.
* Disable debug mode before deploying the application.
* For applications that rely extensively one external resource, consider enabling web gardening on multiprocessor computers. The ASP.NET process model helps enable scalability by distributing work to several processes, one on each CPU. If the application is using a slow database server or calls COM objects that access external resources, web gardening could be a solution.
* Enumerating into collections sometimes is more expensive than index access in a loop. This is because the CLR can sometimes optimize array indexes and bounds checks away in loops, but can't detect them in for each type of code.
* JScript .NET allows methods within methods - to implement these in the CLR required a more expensive mechanism which can be much slower, so avoid them by moving inner methods to be just regular methods of the page.
* Do a "pre-batch" compilation. To achieve this, request a page from the site.
1. Avoid making changes to pages or assemblies that are there in the bin directory of the application. A changed page will only recompile the page. Any change to the bin directory will result in recompile of the entire application.
2. The config file is configured to enable the widest set of features. For a performance boost it is better to tune it to the requirements. Some key points here are:
o Encoding - Request/Response - The default is UTF-8 encoding. If the site is completely ASCII, change the option to ASCII encoder.
o Session State - By default is ON. If session state is not maintained then the value should be changed to OFF.
o ViewState - Default is ON. Turn it off if not being used. If ViewState is being used there are different levels of security that need to considered which can impact the performance of the application.
o AutoEventWireup - Turning off AutoEventWireup means that the page will not try and match up method names to events and hook them up (i.e. Page_Load, etc). Instead, if the application writer wishes to receive them, they need to override the methods in the base class (i.e. override OnLoad for the page load event instead of using a Page_Load method). By doing so, the page will get a slight performance boost by not having to do the extra work itself, but leaving it to the page author.
3. For efficient debugging Use ASP.NET trace feature to debug instead of Response.Write.
Conclusion
Performance is the main objective for most of the ASP.NET applications. Each time we request an ASP.NET page, we run through the same process from initialization to disposal. By understanding the inner workings of the ASP.NET page process and following the above mentioned tips, writing and debugging our code will be much easier and effective (not to mention less frustrating).
Monday, October 13, 2008
Subscribe to:
Post Comments (Atom)
0 comments:
Post a Comment