Build Stuff

What can software development learn from the motor racing industry? In the 1970’s we began to see over specialisation of drivers and designers leading to drivers who knew little about how their cars worked.  Jackie Stewart, 3 times F1 world champion, coined the phrase “Mechanical Sympathy” as a term for the driver and the machine working together in harmony.  This can be summarised in that a driver does not need to know how to build an engine but they need to know the fundamentals of how one works to get the best out of it.For software development, have we reached the point of over specialisation whereby developers no longer understand the fundamentals of how a computer works? Are we more influenced by fashion than science these days? Is fashion just a poor proxy for stylish design that can co-exist with science?In this session we will explore these questions and how we can strike a balance between elegant design and the application of science in the development of great modern software.

Most of us work in languages that provide automatic memory management of some kind. In fact, it's become so commonplace that we take it for granted. But should we? Are garbage collectors really so magical that we can blindly trust them to work efficiently for us? Or are they like any other piece of software solving a hard problem: a best effort, trading off lots of different factors against each other?

This session will take a deep dive into garbage collection. I'll demystify a bunch of the terminology surrounding the topic: reachability analysis, generations, parallel collection, concurrent collection, copying, pinning... I'll also discuss the kinds of trade-offs a GC designer has to make, having played that role over the last year. Finally, there will be some tips on how to write code that is more GC-friendly, for those times when you need to squeeze out a few more drops of performance.

Many systems give you wonderfully simple mechanisms for writing to a file that are some of the leakiest abstractions out there. There are many layers of abstraction between you can the disk. How to make things fast? How to make sure data is actually on the disk? What things can go wrong? How can we benchmark and be sure our benchmarks are reasonable? All of these questions and more will be covered in this talk as we look in detail at everything you never wanted to know.