Book review : Writing High-Performance .NET Code

Cover-Tall-2000x2828-212x300
Ben Watson has been a software engineer at Microsoft since 2008. On the Bing platform team, he has built one of the world’s leading .NET-based, high-performance server applications, handling high-volume, low-latency requests across thousands of machines for millions of customers.

More recently Ben Watson posted two articles on codeproject, both are “Best C# article of the month”. So you can trust this guy.

There are so many books, so many articles & blogs, so many tweets talking about .net performance. Why do we need another book in 2014 and even why do we still need to understand .net internals today when our applications may be hosted on servers with 16 cores and 112 GB of RAM (Azure A9) ?

First, optimizing your code will scale much more better than scaling up/adding new hardware. For sure, you can add more cores, more RAM to your production servers for few €/$ per month, and this is may be a pertinent approach in some cases.  But don’t be naive, a bottleneck in your application is like the sword of Damocles “with great fortune and power come also great peril and anxiety”

More than ever, it’s difficult to predict the performance of a single line of code. By integrating nuget packages, OSS librairies, commercial components … everywhere (“don’t reinvent the wheel, use a framework, they all say”) … our code becomes some kind of magic. In an Agile world, our codebase changes very often and it’s the same of our dependencies: Any commit could potentially destroy your application performance. Good developers have full tests suites (unit, integration, e2e, …) Why not performance tests and benchmarks ?

More than 13 years after the first version of .net, it’s still there, very competitive & efficient in many areas. We can use .net to develop web sites, universal apps, android/iOs app, …. “One language to rule them all, and it’s C#”. Recent additions like async/await and the new features of C# 6 show us that it’s an active and dynamic language. Learning .net runtime internals is certainly not useless.

This is my feeling about dotnet performance today. That’s why I bought this book one month ago.

If you’re a bit concerned about performance, you may know that it’s a very broad topic. What is performance? response time, CPU on a front server, low available memory on a back service … The famous quote of Donald Knuth “Premature optimization is the root of all evil” is a really good argument for detractors “don’t spend time of code optimizations, it’s useless”

One simple rule, maybe the most important in this book is : Measure, Measure, Measure! Don’t try to guess, it’s impossible; don’t try to optimize everything, it’s also impossible. By measuring accurately, you will be efficient. Most of the time to reach your performance goal, you just have to change a few areas in your code, and not everything.

In this book, the author lists a wide range of tools that could help you when investigating performance issues: Visual Studio Profiler, PerfView, CLR Memory Profiler, MeasureIt, … This tool chain should be part of every Performance Chief Architect. “Optimizing performance is meaningless if you do not have effective tools for measuring it”.

This book is quite recent (summer 2014), so the content is quite up-to-date; For example, there are specific chapters on Windows Phone, TPL, .net 4.5 … By the way, it’s a delicious mix between .net internals, best practices and recommendations. It’s hard to talk about code optimization without talking about Garbage collection & Jit compilation.  It may not be the best explanation (in terms of details), but it’s certainly one of the easiest to understand. If you’re quite new in .net, it’s mandatory to understand these topics.

The middle of this book is like a catalog. What is a performance cost of throwing exceptions? Class or struct ? Why Enum.ToString() is so CPU-consuming ? How to work with strings efficiently ? Any difference between for & foreach ? …. You may have so many questions and there are so many answers in this book. Each topic is illustrated by code, IL, … and how to diagnose this problem with your new favorite tools (PerfView, CLR Profiler, Dump, …)

What I like also about this book is that it’s not only focused on code. The Performance effort should not be supported by only one guy (you or me) but by the whole team. Having always up-to-date production metrics, automated benchmarks & profiling, educated product owners, is very important for a successful performance strategy.

It will definitively be part of my library about performance, near to others famous books on this topic. I recommend it to every .net developer.

Writing High-Performance .NET Code

Advertisements

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