Question

I am using asp.net mvc 3 <-> WCF Web service <-> Entity Framework 4 with Azure web role. Here it is the issue :

I have the index page, where i am showing the List of Users in tiles. Each user has associate with more than 20-30 groups may be 100’s of. but I am showing per tile 2 groups only. Refer this fig.

I have mapping of Users and groups in User-Group-Mapping table
with compisit key of userid and groupid. Now prblem is that I have addedd pagination on my index page. and client requirement is list should be statefull . where as i am fteching the List and again for each User fetching List so looks psudo like this

Get Users 
-> For each user
-> Add Div 
 -> For each User Group
   -> Get 2 Associated Groups to this User.
   -> Add Lable 
   -> Add Group Name

doing this becouse of requirements, the List shuold be statefull. so for each user, encountering request to server to fetch its groups…it is really time taking in this architecture. and data transaction between WFC to mvc 3 is slow because of using EF4 i guess. I can not use strictly Jquery.javascript. So what should be the best approcah to perform this. within a second.

Answer 1

(You don’t mention the back-end storage behind your system; I’ll assume it’s SQL Azure, but the discussion below is going to be pretty much the same no matter what.)

When tuning performance for this kind of application I’ve found it important to remember two things:

• Minimize the number of queries in each request and
• Let the database do the heavy lifting

In particular, the time taken to perform a single SELECT query against SQL Azure can be several hundred milliseconds. That suggests that you want to do a maximum of 10 – 15 queries per HTTP request for sub-second response time.

One approach to help this can be to make sure you query for all the data you need in as few database queries as possible. That will mean you’re pulling back lots of data through EF in one go, but that’s generally OK: in the cloud, query execution may be slow but transfer of data is usually quick.

So, for instance, replace things like iterating over a collection on the client with a groupby that executes a single larger query on the server instead.

An extremely useful tool to profile your application and identify repeated queries is the MiniProfiler, http://code.google.com/p/mvc-mini-profiler/.

If that still doesn’t help then consider denormalizing your data. There’s many approaches to this:

• create an additional table on the server that’s re-populated when you make changes to the data: this table is a flat representation of what’s needed when you display the Web page
• create a table using Azure table storage that is populated when you make changes to the data: this might store a serialized representation of your data which can then be queried or sorted in memory with LINQ to XML
• create a view on the database (sometimes there’s no real performance gain over a query, though)

Besides the development time and storage space required, the biggest tradeoff here is that writes will be slower as the denormalized representation will need to be updated. This is a situation where it may be appropriate to update the denormalized representation asynchronously using a worker role and an Azure queue.