Performance, Measure and ANTS Profiler

Might need to create a separate page for notes on performance since I've been doing a lot of C# and database tuning lately but having them on this post so far. Here are some of my notes on performance [more]

* Before you optimize, ensure that your results are accurate first before optimizing. I would suggest Test Driven Development or at least some unit testing but that's another story. Just make sure your results are correct first otherwise your optimization is useless.

* Before you optimize, always MEASURE MEASURE MEASURE

* If you can measure without introducing extra code, go with it. Aside from saving time, it will keep your code clean and introduce what could be unnecessary complexity. And for this very reason I just purchase ANTS Profiler 4 from RedGate software. Another alternative is dotNetTrace from JetBrains and although I love Resharper from the same company I prefer ANTS profiler. I said prefer because as I haven't explored dotNetTrace that much, i would say it is really preference, UI/usability and being able to view the timings embedded in a window which shows the source code. Also like the call graph, drill down on significant methods in terms of time spent on it. I got a quote for a no support/update version of the software when I emailed RedGate about their pricing which is admittedly very high for a personal purchase when you are earning from a 3rd world country. So will be receiving fixes but not major upgrades (so as they say but haven't dug deep – besides the version looks pretty good and turns out to be very useful for me already). It really sucks to try to optimize something only to find out that it is not the bottleneck

* If your application involves database access and you're slow then more often than not your database is not tuned. And SQL Profiler and Database Tuning Advisor is a very good start. You can apply the recommendations or you can just evaluate them and make your own adjustments. In my experience however the recommended changes does make a lot of performance improvement. You think you know enough about indexes and database design? You might be surprised how much performance you can gain from these tools.

Again measure, measure, measure but if you can only identify some specific parts then data access seems like a good start.

* Then look into you application logic. Even if your code is optimized but your algorithm/logic is wrong, it will still turn out bad

* Also know about database/table statistics, indexed views and partition. If you can take advantage of partitions and involving considerable data access the improvement is quite impressive.

* <string>.SubString(…) does some considerable lifting so if you want to check if the first two chars in a string is equal to some other string then you should consider using <string>.StartsWith instead or avoid the SubString if you can

* Hashtable is faster than SortedDictionary, SortedList or List. I know I should provide stats for this (have a URL somewhere and will update this soon)

* When using nullable types, use <variable>.HasValue as much as possible than comparing using != null. Or even "!<var>.HasValue" vs. "== null". But avoid negation if you have to

Finally this list will grow soon hopefully and if feel like you disagree please feel free to comment and have other benefit from your thoughts too 🙂


Posted

in

, ,

by

Tags: