A broad look at Web Application Performance


So does your web application perform well? Is it responsive enough so that it increases the web user experience? It is a fact that user can relate only to what they see and what they click, that is the user interface (UI) and the experience with it (UX) largely depends on the performance of the overall
application. But what is the complaint that we usually hear, “this page is slow”, “this grid takes for ever to load”, “we click on that button and then we go for a coffee break”…and UI is always blamed first. We all know that it has to do with the perception a lot, but a services guy ignores it, database developer closes his eyes too, till someone proves where is the actual bottleneck.
Performance Tips
My cose friend and ex-colleague called me the other morning asking me help to figure out why the application was timing out, users are screaming and cursing this particular search page, it started timing out suddenly last week. They did all the usual tricks, iisreset, checking logs, checking db connections, even few server restarts, no avail. So this morning, we started looked thru the logs, stack traces and we could easily spot the culprit, a stored procedure in SQL Server, has gone rogue and slow. So in this case, we need to look closely at that proc, probably data has grown so much in one of the table that now a bad join has created performance issues.

I shared that experience, to point to fact that performance and thus performance tuning has to start at the very tail end (or head depends the way your look at the layered architecture :). So now lets take the layers, there is presentation layer, business layer, database access layer (persistence layer) and then there is the database itself.

Lets start with Database: What are the performance optimizations that you can think of, here is some starters.

  • Good logical design, helps in how db is physically stored, think thru your normalization levels depending on your usage, make sure you dont over-normalize, it will kill your joins and thus leads to bad performing queries.
  • Good key selection, have a good candidate for primary key, have proper foreign keys that are indexed (create non-clusterd index)
  • When writing queries, do not return all columns (i.e., no select * …)
  • Use SET NO COUNT ON, think about providing sql hints
  • There is 2 camps about using temp tables, some hate it, some love it, according to me temp table can be very useful when you are dealing with a large amount of tables to join. Create temp tables to divide and conquer. Always think about “number of rows” in the table when you make these decisions and potential growth of data. Check query plans, avoid table scans.
  • Use sp_executesql when you want to execute a dynamic sql
  • Reduce round trips to the server, try to do bulk operations.
  • This is just few I can think of now, Google database tuning and I am sure you can find a tons of good tuning tips from database gurus.

    Now how about Data Access Layer (DAL)? So what is DAL? DAL is nothing but a mechanism that provides us some CRUD capability along with some transactions, querying and concurrency control. In simpler systems, ADO.Net provides an database independent way to achieve this, So ADO.Net is our DAL layer. Or you might choose to absract the core features provided by ADO.Net into another library where you have tighter control and ability to switch databases at will, a DB Services Factory. If you are following a Domain Driven Design, then your OR/M plays the role of DAL.

    Lets see some ADO.Net related performance tips,

  • For connection pooling to work, always use the same connection string. By default pool size is 100, you can increase this (rarely you will need to)
  • Use db server IP rather than the server name in connection string to avoid a domain name resolution lookup. (This is a performance tip, but not a governance tip, for better governance, always use DNS name)
  • Always close the connections explicitly
  • Connect using a service account. Use windows authentication.
  • Use stored procedures, return only what you need.
  • Use some caching strategy to cache frequently used static data.
  • Again this is tip of the iceberg on what performance considerations you can use, for a more complete ADO.Net performance tuning tips, refer to “Improving .Net Application Performance and Scalability” from Microsoft patterns and practices.

    We will continue to explore performance from a OR/M layer perspective, then a Business Layer and finally UI layer in a future post.

    Until then Cheers and Happy Programming.

    Leave a Reply

    Fill in your details below or click an icon to log in:

    WordPress.com Logo

    You are commenting using your WordPress.com account. Log Out / Change )

    Twitter picture

    You are commenting using your Twitter account. Log Out / Change )

    Facebook photo

    You are commenting using your Facebook account. Log Out / Change )

    Google+ photo

    You are commenting using your Google+ account. Log Out / Change )

    Connecting to %s