The soda war of software
Will’s talk at BugBash was, at heart, about how and why we as software engineers can make software better. On some level, that’s an embodiment of what everyone at Antithesis is trying to do – not just because more and more of the world runs on software, but because it matters to us, very directly, as people and as professionals. Because software is the thing we do for fun and the thing we do for money and the thing that keeps us in darkened rooms in front of screens for way too many hours of our waking lives.
Part of making software better is making sure it works, and if you’re a working engineer at a company where software is part of the value chain, you’re definitely familiar with at least two of the main approaches (Will spoke about five) to doing this: testing and observability.
In some sense these are conceptual opposites – testing attempts to find problems before software is deployed, observability looks for problems in deployed software. The number one argument for emphasizing observability as a way of building reliable software is that testing outside of production doesn’t work very well. At the same time, it will forever be cheaper and better to discover and fix problems before they take your system down.
So the ideal observability platform is one that tells you what’s happening to your system before your users notice. Savvy SREs try to do this, setting alerts and tripwires that tell them that their system might get into trouble, well before it actually does.
But also – we’re constrained by the state of our tools. Testing in production is an attractive paradigm today because It’s extremely difficult and extremely expensive to build a testing apparatus that reasonably approximates production, while there are many sophisticated, highly capable observability platforms out there. You’ve got to have observability anyway, why not really lean into it?
In this talk, Will asks: what if you do have a testing setup that realistically simulates the chaos of production? Something like Antithesis? What if you could observe how your system would behave in production, or in an environment vastly more chaotic and hostile than prod, without a single user being affected, without a single alarm actually going off?
Watch on, and find out. We’re launching this feature in July.
P.S. I suggest watching the whole talk, but if you just want to hear about Pre-Observability, he starts around 32:00.
