Lead Image © Alberto Andrei Rosu, 123RF.com

Scale-out with PostgreSQL

Load Up

Article from ADMIN 73/2023

By Franck Pachot

The world of scale-out is stateless; unfortunately, databases are not. YugabyteDB solves this dilemma for PostgreSQL.

If you want to respond to higher load with more instances instead of a more powerful server, you typically need stateless applications. Horizontal scaling, whether in the cloud or with multiple servers, virtual machines, Kubernetes pods, or containers, creates the elasticity and resilience that business applications need – especially in the cloud, where every component needs to be replaceable in the event of failure and where elasticity ensures cost reductions without performance loss.

Hardly any application can do without a database to store data and ensure consistency, even with concurrent transactions. However, a database is likely to be stateful, and that presents a challenge when it comes to scaling. YugabyteDB is a cloud-native database, is fully open source, and aims to achieve PostgreSQL compatibility, which means the ability to run the same applications and tools with identical application behavior while handling complex online transactional processing.

Monolithic RDBMSs

You could ignore the modern challenges of relational database management systems (RDBMSs) if you just opened them on a single server. This restriction allows database processes simply to use the operating system's shared memory, which is protected by latches (spinlocks or mutexes).

Database users often think their database's shared memory pool is only useful as a cache for boosting performance, because it removes the need for I/O access to the data file in many cases. In fact, one of the main reasons for the existence of the shared memory pool is that it serves as a single source of truth when different sessions need read and write access to a given block of data without corrupting the block. However, a side benefit is that a larger shared memory pool also reduces read and write access to the hard disk for frequently used blocks. PostgreSQL in particular partially delegates this cache function to the

...

Use Express-Checkout link below to read the full article (PDF).