Tutorial: Design a multitenant database with elastic clusters (preview)

APPLIES TO: Azure Database for PostgreSQL - Flexible Server

In this tutorial, you use Azure Database for PostgreSQL with elastic clusters to learn how to design a multitenant application in order to benefit from horizontal scale-out.

Prerequisites

Create an elastic cluster in one of the following ways:

• Create an elastic cluster with ARM template
• Create an elastic cluster using CLI
• Create an elastic cluster using the Portal

Use psql utility to create a schema

Once connected to the elastic cluster using psql, you can complete some basic tasks. This tutorial walks you through creating a web app that allows advertisers to track their campaigns.

Multiple companies can use the app, so let's create a table to hold companies and another for their campaigns. In the psql console, run these commands:

CREATE TABLE companies (
  id bigserial PRIMARY KEY,
  name text NOT NULL,
  image_url text,
  created_at timestamp without time zone NOT NULL,
  updated_at timestamp without time zone NOT NULL
);

CREATE TABLE campaigns (
  id bigserial,
  company_id bigint REFERENCES companies (id),
  name text NOT NULL,
  cost_model text NOT NULL,
  state text NOT NULL,
  monthly_budget bigint,
  blocked_site_urls text[],
  created_at timestamp without time zone NOT NULL,
  updated_at timestamp without time zone NOT NULL,

  PRIMARY KEY (company_id, id)
);

Each campaign pays to run ads. Add a table for ads too, by running the following code in psql after creating companies and campaigns:

CREATE TABLE ads (
  id bigserial,
  company_id bigint,
  campaign_id bigint,
  name text NOT NULL,
  image_url text,
  target_url text,
  impressions_count bigint DEFAULT 0,
  clicks_count bigint DEFAULT 0,
  created_at timestamp without time zone NOT NULL,
  updated_at timestamp without time zone NOT NULL,

  PRIMARY KEY (company_id, id),
  FOREIGN KEY (company_id, campaign_id)
    REFERENCES campaigns (company_id, id)
);

Finally, we track statistics about clicks and impressions for each ad:

CREATE TABLE clicks (
  id bigserial,
  company_id bigint,
  ad_id bigint,
  clicked_at timestamp without time zone NOT NULL,
  site_url text NOT NULL,
  cost_per_click_usd numeric(20,10),
  user_ip inet NOT NULL,
  user_data jsonb NOT NULL,

  PRIMARY KEY (company_id, id),
  FOREIGN KEY (company_id, ad_id)
    REFERENCES ads (company_id, id)
);

CREATE TABLE impressions (
  id bigserial,
  company_id bigint,
  ad_id bigint,
  seen_at timestamp without time zone NOT NULL,
  site_url text NOT NULL,
  cost_per_impression_usd numeric(20,10),
  user_ip inet NOT NULL,
  user_data jsonb NOT NULL,

  PRIMARY KEY (company_id, id),
  FOREIGN KEY (company_id, ad_id)
    REFERENCES ads (company_id, id)
);

You can see the newly created tables in the list of tables now in psql by running:

\dt

Multitenant applications can enforce uniqueness only per tenant, which is why all primary and foreign keys include the company ID.

Shard tables across nodes

An elastic cluster deployment stores table rows on different nodes based on the value of a user-designated column. This "distribution column" marks which tenant owns which rows.

Let's set the distribution column to be company_id, the tenant identifier. In psql, run these functions:

SELECT create_distributed_table('companies',   'id');
SELECT create_distributed_table('campaigns',   'company_id');
SELECT create_distributed_table('ads',         'company_id');
SELECT create_distributed_table('clicks',      'company_id');
SELECT create_distributed_table('impressions', 'company_id');

Ingest sample data

Outside of psql now, in the normal command line, download sample data sets:

for dataset in companies campaigns ads clicks impressions geo_ips; do
  curl -O https://examples.citusdata.com/mt_ref_arch/${dataset}.csv
done

Back inside psql, bulk load the data. Be sure to run psql in the same directory where you downloaded the data files.

SET client_encoding TO 'UTF8';

\copy companies from 'companies.csv' with csv
\copy campaigns from 'campaigns.csv' with csv
\copy ads from 'ads.csv' with csv
\copy clicks from 'clicks.csv' with csv
\copy impressions from 'impressions.csv' with csv

This data spreads across worker nodes.

Query tenant data

When the application requests data for a single tenant, the database can execute the query on a single worker node. Single-tenant queries filter by a single tenant ID. For example, the following query filters company_id = 5 for ads and impressions. Try running it in psql to see the results.

SELECT a.campaign_id,
       RANK() OVER (
         PARTITION BY a.campaign_id
         ORDER BY a.campaign_id, count(*) desc
       ), count(*) as n_impressions, a.id
  FROM ads as a
  JOIN impressions as i
    ON i.company_id = a.company_id
   AND i.ad_id      = a.id
 WHERE a.company_id = 5
GROUP BY a.campaign_id, a.id
ORDER BY a.campaign_id, n_impressions desc;

Share data between tenants

Until now all tables distributed by company_id. However, some data doesn't naturally "belong" to any tenant in particular, and can be shared. For instance, all companies in the example ad platform might want to get geographical information for their audience based on IP addresses.

Create a table to hold shared geographic information. Run the following commands in psql:

CREATE TABLE geo_ips (
  addrs cidr NOT NULL PRIMARY KEY,
  latlon point NOT NULL
    CHECK (-90  <= latlon[0] AND latlon[0] <= 90 AND
           -180 <= latlon[1] AND latlon[1] <= 180)
);
CREATE INDEX ON geo_ips USING gist (addrs inet_ops);

Next make geo_ips a "reference table" to store a copy of the table on every worker node.

SELECT create_reference_table('geo_ips');

Load it with example data. Remember to run this command in psql from inside the directory where you downloaded the dataset.

\copy geo_ips from 'geo_ips.csv' with csv

Joining the clicks table with geo_ips is efficient on all nodes. Here's a join to find the locations of everyone who clicked on ad 290. Try running the query in psql.

SELECT c.id, clicked_at, latlon
  FROM geo_ips, clicks c
 WHERE addrs >> c.user_ip
   AND c.company_id = 5
   AND c.ad_id = 290;

Customize the schema per-tenant

Each tenant might need to store special information not needed by others. However, all tenants share a common infrastructure with an identical database schema. Where can the extra data go?

One trick is to use an open-ended column type like PostgreSQL's JSONB. Our schema has a JSONB field in clicks called user_data. A company (say company five), can use the column to track whether the user is on a mobile device.

Here's a query to find who clicks more: mobile, or traditional visitors.

SELECT
  user_data->>'is_mobile' AS is_mobile,
  count(*) AS count
FROM clicks
WHERE company_id = 5
GROUP BY user_data->>'is_mobile'
ORDER BY count DESC;

We can optimize this query for a single company by creating a partial index.

CREATE INDEX click_user_data_is_mobile
ON clicks ((user_data->>'is_mobile'))
WHERE company_id = 5;

More generally, we can create a GIN indices on every key and value within the column.

CREATE INDEX click_user_data
ON clicks USING gin (user_data);

-- this speeds up queries like, "which clicks have
-- the is_mobile key present in user_data?"

SELECT id
  FROM clicks
 WHERE user_data ? 'is_mobile'
   AND company_id = 5;

Next step

In this tutorial, you learned how to create an elastic cluster. You connected to it with psql, created a schema, and distributed data. You learned to query data both within and between tenants, and to customize the schema per tenant.