How to Optimize Rules for Facebook

There's definitely a lot of optimization that has to happen because many of these rules have to run synchronously when the user is performing some action on Facebook. So you need to, well, we run all the rules in parallel for one thing. That's where Haskell is useful because it runs everything together in parallel and combines all the data fetches and does memorization and so on. And then there are a number of rules that run offline as well. They also have performance constraints because we have a total number of resources that we can bring to bear on running Sigma rules. Anything that's far too expensive will just contain too many machines and reduce the resources we have available to run other rules.