Multi-Tenant Apps & Postgres That Scales Out

Is Your SaaS Application Hitting the Limits on Single-Node Postgres?

Are your database queries taking too long?

Do you have more data than you can store in single-node Postgres?

Are you worried about whether your database infra can handle new customer signups?

Are you worried you might have to give up SQL in order to scale out your database?

Does your database performance deteriorate with a high number of concurrent requests?

Are lots of your developers spending time micro-optimizing your database?

Are you looking for a way to scale that doesn’t require your engineers to figure out sharding?

Is it taking you too long to analyze metrics across your customer base?

Is your largest customer creating performance issues for the rest of your customers?

Scale Out Without Sacrificing SQL (Keep Postgres, too)

Most SaaS applications are inherently relational. By sharding on customer_id, Citus gives you the ability to scale out your SaaS database with full SQL coverage, so you don’t have to give up the relational semantics you need, like joins, foreign key constraints, transactions, ACID, consistency. By scaling out, your app gets much bigger memory, compute, and disk footprint—ensuring performance even in the face of growth. And since Citus is an extension to Postgres, you can leverage all of your existing expertise—plus all the Postgres reliability, rich data types (think: JSON and PostGIS), Postgres extensions (our favorites include HyperLogLog), and ongoing innovation in the amazing Postgres open source community.

Get an Elastic Database That Works With Rails & Django

So you want a database that can handle peak workloads and boatloads of concurrent users; act as a system of record and provide business analytics; offer high availability and provide elastic scale as your customer base triples in size. Check. The icing on the cake is for the database to integrate seamlessly with your frameworks like Rails, Django, Hibernate, and Sequelize. Because Citus is an extension to Postgres, we’ve been able to build on top of the robust, existing Postgres integrations and make it easy for multi tenant apps to use Citus: recent additions include an activerecord-multi-tenant gem for Ruby on Rails, and a Python django-multitenant library for Django. More to come.

Shard by Tenant to Scale Out Your Multi-Tenant Application

While you may have a lot of data and your SaaS database may soon be too big for single-node Postgres, it’s likely that any given customer of yours can get the database performance they need from a single node. With our Citus extension to Postgres, the database is scaled out and yet data for each of your SaaS customers is co-located on the same worker node. As you grow and sign up new customers, adding nodes (and memory, compute, and storage) to your Citus cluster is simple. And our Citus shard rebalancer balances the workload across the cluster as you add nodes—with zero downtime.

Analyze Activity with Powerful Cross-Shard Queries

As of Citus 10 you can run Citus on a single node, or on multiple worker nodes—with Postgres and Citus installed on each node in the cluster. If you’re running your app on Citus in the cloud and you have HA turned in Azure Cosmos DB for PostgreSQL, then standby nodes will exist, too. Citus shards the database tables based on the distribution key, and then distributes shards across the worker nodes. The pièce de résistance: Because Citus can parallelize query fragments across multiple nodes in a cluster, you can run powerful cross-shard queries across all of your customer data quickly and easily.

screenshot of Grafana for Citus tenant monitoring

Monitor Tenants

Find noisy neighbors in your cluster with CPU usage insights and query counts from citus_stat_tenants. Citus maintains a resource utilization snapshot for top N tenants you can use with Grafana to visualize congestion & locate roadblocks.

Why Citus for Multi-Tenant SaaS Applications?

Need a database with performance that can support the growth of your SaaS business.
Want to take advantage of Postgres reliability, data types, extensions, toolset, & expertise.
Would like an open source solution (do not want lock-in).
Fully-managed cloud database fits strategy of focusing on core application & product.
Citus is easy to try out: Can install & setup dev environments locally with Docker.

Don’t want to give up JOINS, transactions, & consistency of relational database semantics.
Don’t want to divert developers from core business app to spend eons sharding.
Can’t afford opportunity cost of re-architecting SaaS app to go to NoSQL.
Easier to run on a database that has direct integration with Rails & Django frameworks.
Need to enable developers to experiment & analysts to analyze, without production impact.

Multi-Tenant Applications