Discover and read the best of Twitter Threads about #fitsinatweet
Most recents (1)
1/ People often wonder how tumbling (sometimes called “fixed”) windows work in stream processing. They’re an incredibly useful construct, and their implementation can be surprisingly interesting. Let’s dig in! 
There are lots of problems where a piece of data is only relevant to other pieces of data that occur close to it in terms of time. In other words, you need to group proximal events together based on when they happened. Windows are an abstraction for doing just that.
They group data together based on time. They can support fast updates and retrieval with O(1) operations, even in the presence of out of order data. They can also be made solidly memory efficient.
