What's new in Data for SaaS? October 2024

October is almost over and the nights are drawing in. re:Invent may be just around the corner, but that didn't stop October from being an action-packed month in the world of data.

On November 19th we're running a deep-dive on PostgreSQL performance and scaling, with a focus on challenges software companies and SaaS providers often face.

PostgreSQL is one of (if not the) most popular relational database used in modern software applications, but it can easily become the bottleneck as your application grows. Performance tuning for PostgreSQL doesn't need to be a dark art, and in this session we lift the lid on the unique internals that make PostgreSQL tick, how to use the native tools available and how AWS builds on this by providing platform functionality and observability to get the most out of your database.

In the morning, we'll be going through a series of deep dives covering PostgreSQL internals, a deep-dive into one of the 3 critical database scaling dimensions: Connection Management (the other 2 being vertical and horizontal scaling), and finally a look into what Amazon Aurora brings to the table on top of open source PostgreSQL.

Stick around until after lunch and we'll get into hands-on labs putting what we learned in the morning into practice.

Become the PostgreSQL hero your company needs!

In this post, Druva describe how in 12 years of running DynamoDB in production, they never had an outage due to database maintenance. Most database engines will require you to perform periodic updates to apply critical security and stability patches, not to mention more major updates to get the latest functionality. This most often means restarting the database process, resulting in at least some lowered availability. DynamoDB abstracts all this away so you're always running on the latest version, seamlessly without interruption.

Amazon Neptune is a graph database that is popular for workloads with highly related data such as fraud detection, knowledge graphs and social networks. Multi-tenant implementation in Neptune is highly dependent on the partitioning model and whether you are using a property graph or RDF. This prescriptive guidance has some great examples (particularly for pooled models) on how to approach this.

The RDS Data API is one of my favorite features we released this year, and now it's available for both PostgreSQL and MySQL. If you want to learn more about multi-tenancy with the Data API, check out this blog post.

We've had item-level access control in DynamoDB for a while, which is great for a pooled partitioning model. This new feature allows you to restrict access to certain tables based on tags, which is great for a silo partitioning model where you have a table per tenant (by the way, table-per-tenant is a scalable approach for multi-tenancy; you can have by default 2,500 tables per account, adjustable up to 10,000. More than enough for most SaaS applications).

It was a bit of undifferentiated heavy lifting to maintain 2 different sets of config for your primary and DR regions with Aurora Global Database (or worse, trying to manage the DNS yourself). Now with the Global Writer Endpoint, your applications can connect to a single endpoint that will automatically follow the primary cluster in the event of a Global Database failover.

If you're heading to re:Invent in a few weeks, you won't want to miss SAS306! This session will answer everything you ever wanted to know, but were afraid to ask about building multi-tenant data architectures.