Demo of Using Postgres & Citus for Lightning Fast Analytics Queries, also ft. Rollups

A demo to learn how you can use Citus to scale out Postgres and provide fast analytics queries. Using a 4-node Citus database cluster, this demo will show you how Citus shards Postgres to give you lightning fast performance, at scale.
-----------------------------------
Video bookmarks:
► 1:45 Why use Postgres and Citus for real-time analytics?
► 4:12 Overview of the Citus Architecture
► 6:34 What are rollups and why are they useful
► 9:01 What is hll?
► 10:16 Demo of fast analytics queries with Postgres and Citus
► 25:19 Performance comparisons of Citus vs. single Postgres node
► 25:44 Benefits of using Citus for demanding analytics workflows
► 29:33 How Heap Analytics uses a 70-node Citus database cluster
► 30:19 How Algolia uses Citus for real-time search analytics
► 31:29 How Microsoft uses Citus for their Windows telemetry data and VeniceDB
-----------------------------------
This demo of scaling out Postgres with the Citus extension to Postgres is useful for teams building customer-facing analytics dashboards who need to ingest data in real time while delivering sub-second responses; and for teams looking for a way to consolidate database platforms, to avoid having separate stores for your transactional and analytics workloads.
Citus is available as a database as a service, as enterprise software, and as open source.
✅ See how enterprises and SaaS businesses use Citus for real-time analytics today:
www.citusdata.com/customers/
✅ Join the Citus database community’s public slack channel:
slack.citusdata.com/
✅ Star the GitHub repo for the open source Citus database:
github.com/citusdata/citus
✅ Learn more about the Citus distributed database and how Citus scales out Postgres for real-time analytics dashboards:
www.citusdata.com/product

[@nikhiljha8717]

Hi, your video was very informative and has made me eager to use this in my application. May you please make a step by step video demonstrating how can i install and configure postgresql in hyperscale environment on a local cluster.

[@shweta3644]

i have a question. what is the resource for single instance of postgres...?...in distributed we see 4 nodes with 8 vcpus and ~ 60 GB RAM each node? is single node 32 vCPus and 240 GB RAM or 1/4?

[@milomiller161]

Still, nodes only make sense when you can't scale the SQL server any further. If you used the same RAM, CPU's etc on a single server it would outperform the distributed system.
However, the price on smaller hardware may make it cheaper to build a distributed system as well, so keep this in mind when designing the system.
The main point is this, NO SQL is not the solution based on scalability. No SQL databases are useful when the use cases supported the architecture of the DB in question. Different No SQL database, therefore, have different uses cases.
If the use case is a SQL DB, this is, however, the way to go when scaling the single server further is not possible or too expensive.
PS: Oracle RAC is doing a similar job, however for a premium license fee.

[@ahmettek315]

Are you sure of that. I am seeing better performance even on a single node comparing our SQL server data to Citus data generated from that (and I even favor SQL server keeping the data small enough to fit in memory - maybe there is a switch in SQL server that I miss to make it faster).
Even Microsoft itself found out postgreSQL and citus is the answer to some problems they are having and they acquired Citus. Check Cosmos Db for postgteSQL as they named it.
(and both Citus and postgreSQL are open source)