Tumblr surpassed 10 billion posts only a month ago, and today we have over 11.4 billion. At peak times, over 900 posts are published on our platform every second.
Ever wonder how we manage all that data?
Although we’re increasingly making use of newer, non-relational data stores like HBase and Redis, the bulk of our data is currently stored in a heavily partitioned MySQL architecture.
On November 8, I’ll be giving a talk at Velocity Europe in Berlin about how we’ve scaled our MySQL databases at Tumblr, with a particular focus on how, when, and why we sharded our dataset. The talk will also cover the tools we built to manage and move billions of rows of MySQL data for millions of blogs, and how to avoid some nasty pitfalls along the way.
Register for Velocity with our discount code veu11sc20 and save 20%. Hope to see you there!