Skip to main content

Avoiding Data Hotspots at Scale

There are two key choices when scaling a NoSQL datastore: choosing between a hash or a range based sharding and choosing the right sharding key. Any choice is a trade-off between scalability of read, append, and update workloads. In this talk, I present the standard scaling techniques, some non-universal sharding tricks, less obvious reasons for hotspots, as well as techniques to avoid them.