ReadySet Cal.com demo

This repository contains instructions and scripts for trying out ReadySet with Cal.com. The goal is to demonstrate how easy it is to get up and running using ReadySet with a real-world, production-scale web application, and to show an example of the kind of improvements you might see in throughput and latency.

Prerequisites

In order to run Cal.com, you'll need the following installed on your machine:

• Node.js (version 15 or 16)
• yarn
• A PostgreSQL server (version 13 or 14)

In addition, in order to run ReadySet for this demo, you'll need a local release build of the ReadySet binary. See the readyset documentation for more information on prerequisites for building ReadySet.

Note that you must follow the above instructions to generate a release build; if you are using a development build, you may see significantly lower performance numbers than expected. The above instructions do include the --release flag, but this point is worth emphasizing here since some users who are already familiar with ReadySet may be used to creating debug builds, and might therefore gloss over the above instructions.

Step 1: Set up PostgreSQL and seed the database

The following steps expect the postgres user to exist, so make sure to create this user and either set it as the default (via e.g. .pgpass), or specify it manually in the psql flags of the commands that follow. You should also make sure to create the postgres user with the necessary permissions to create a new database.

First, create a database in your PostgreSQL server for Cal.com to use to store its data:

$ psql
localhost/postgres=# create database calcom;
CREATE DATABASE

This repository includes a SQL DB dump file containing a large amount of data, which we'll use later to benchmark a query to compare ReadySet against PostgreSQL.

To load this data - along with all the DDL necessary to run Cal.com - into the Postgres database, run the following command:

cat calcom.sql.gz.1 calcom.sql.gz.2 calcom.sql.gz.3 | gunzip -c | psql calcom

Finally, note that the wal_level setting must be set to logical. You can check the value of this setting in a psql shell by running SHOW wal_level;. If it is not set to logical, update the setting in your postgresql.conf file and restart Postgres.

Step 2: Run Cal.com

First, clone the Cal.com repository:

$ git clone [email protected]:calcom/cal.com

Next, install dependencies:

$ cd cal.com
$ yarn install

To configure Cal.com to point at your local database, first copy the .env.example file to .env:

$ cp .env.example .env

After that, use the openssl rand -base64 32 command to generate a key and add it under NEXTAUTH_SECRET in the .env config file. Similarly, use the command openssl rand -base64 24 to generate a key and add it to the CALENDSO_ENCRYPTION_KEY field.

Then edit the file such that DATABASE_URL contains the connection information for your PostgreSQL database. Note that this necessarily includes editing the default database string of calendso and replacing it with calcom. It also likely includes changing the port in the connection string, as PostgreSQL typically defaults to listening on port 5432, whereas the example Cal.com config file uses port 5450.

Now, to start the Cal.com dev server, run:

$ yarn dev

Once a bit of time has passed, you should be able to access Cal.com at http://localhost:3000

From there, you can log in using the username [email protected] and a password of password.

Step 3: Run ReadySet

To start ReadySet and have it connect to your PostgreSQL database, run the following command:

$ path/to/readyset \
    --standalone \
    --deployment calcom \
    --database-type postgresql \
    --upstream-db-url postgresql://<user>:<password>@127.1/calcom \
    -a 0.0.0.0:5435

Once ReadySet outputs a log line containing Streaming replication started, it's ready to use for the application. Note that it may take some time to reach this point.

Step 4: Connect Cal.com to ReadySet

To point Cal.com at ReadySet, edit the .env file to add port 5435 to the configured DATABASE_URL. Then, restart the dev server. Cal.com is now using ReadySet!

Step 5: Cache a query

The query from Cal.com that we'll be caching is:

SELECT
    min("public"."Booking"."startTime"),
    count("public"."Booking"."recurringEventId"),
    "public"."Booking"."recurringEventId" FROM "public"."Booking"
WHERE "public"."Booking"."recurringEventId" IS NOT NULL
AND "public"."Booking"."userId" = $1
GROUP BY "public"."Booking"."recurringEventId"

To tell ReadySet to cache this query, first open a psql shell connected to port 5435:

$ psql -p 5435

Then issue the CREATE CACHE FROM command with the query:

localhost/calcom=> CREATE CACHE FROM SELECT
    min("public"."Booking"."startTime"),
    count("public"."Booking"."recurringEventId"),
    "public"."Booking"."recurringEventId" FROM "public"."Booking"
WHERE "public"."Booking"."recurringEventId" IS NOT NULL
AND "public"."Booking"."userId" = $1
GROUP BY "public"."Booking"."recurringEventId";

Step 6: Benchmarking the query

Let's run a more thorough benchmark to compare the latency and throughput of executing that query against Postgres vs ReadySet. To do this, we'll use pgbench, which is a Postgres benchmarking tool distributed with Postgres itself.

This repository contains a file benchmark.sql, which can be used as a pgbench custom benchmark script. We can use this file to run a pgbench benchmark for the query we just created a cache for. First, let's run against Postgres itself, to establish a baseline:

$ pgbench -Mprepared -j32 -s32 -c32 -f ./benchmark.sql -T30 -U postgres calcom

Now, let's run against ReadySet to compare:

$ pgbench -Mprepared -j32 -s32 -c32 -f ./benchmark.sql -T30 -U postgres -h 127.1 -p5435 calcom