Distributed SQL Fundamentals: Horizontal Scaling

This was a live webinar that was recorded and later re-published on Yugabyte’s Vimeo.

Scaling your SQL database can be difficult. Most traditional, legacy SQL databases are monolithic – they are constrained to a single server, VM, or container. If your database exceeds the CPU, memory, or storage resources, you either have to move to a new, more powerful instance or upgrade the existing one. Vertical scaling works fine until your database or transaction volumes get large. Since you can only increase the resources for an instance so far, your database can only scale so far. Then you will inevitably start running into latency and stability issues with your database.

Horizontal scaling addresses the problems with vertical scaling. Instead of having a single, monolithic instance, you have multiple distributed instances – called nodes. So if you need more CPU or memory, you add another node. This helps ensure availability, reduce latency, and scale far beyond what is possible with vertical scaling.

In this talk, we will discuss horizontal scaling: what it is, how it compares with vertical scaling, its strengths and weaknesses, and why it should be a requirement for your next application database.