Caching RESTful API requests with Heroku’s Redis Add-on

Michael Bogan - Apr 17 - - Dev Community

Most software developers encounter two main problems: naming things, caching, and off-by-one errors. 🤦🏻‍♂️

In this tutorial, we’ll deal with caching. We’ll walk through how to implement RESTful request caching with an open source-licensed version of Redis. We’ll also set up and deploy this system easily with Heroku.

For this demo, we’ll build a Node.js application with the Fastify framework, and we’ll integrate caching with Redis to reduce certain types of latency.

Ready to dive in? Let’s go!

Node.js + Fastify + long-running tasks

As I’m sure our readers know, Node.js is a very popular platform for building web applications. With its support for JavaScript (or TypeScript, or both at the same time!), Node.js allows you to use the same language for both the frontend and the backend of your application. It also has a rich event loop that makes asynchronous request handling more intuitive.

The concurrency model in Node.js is very performant, able to handle upwards of 15,000 requests per second. But even then, you might still run into situations where the request latency is unacceptably high. We’ll show this with our application.

As you follow along, you can always browse the codebase for this mini demo at my GitHub repository.

Initialize the basic application

By using Fastify, you can quickly get a Node.js application up and running to handle requests. Assuming you have Node.js installed, you’ll start by initializing a new project. We’ll use npm as our package manager.

After initializing a new project, we will install our Fastify-related dependencies.

~/project$ npm i fastify fastify-cli fastify-plugin
Enter fullscreen mode Exit fullscreen mode

Then, we update our package.json file to add two scripts and turn on ES module syntax. We make sure to have the following lines:

  "type": "module",
  "main": "app.js",
  "scripts": {
    "start": "fastify start -a 0.0.0.0 -l info app.js",
    "dev": "fastify start -p 8000 -w -l info -P app.js"
  },
Enter fullscreen mode Exit fullscreen mode

From there, we create our first file (routes.js) with an initial route:

// routes.js

export default async function (fastify, _opts) {
  fastify.get("/api/health", async (_, reply) => {
    return reply.send({ status: "ok" });
  });
}
Enter fullscreen mode Exit fullscreen mode

Then, we create our app.js file that prepares a Fastify instance and registers the routes:

// app.js
import routes from "./routes.js";

export default async (fastify, opts) => {
  fastify.register(routes);
};
Enter fullscreen mode Exit fullscreen mode

These two simple files—our application and our route definitions—are all we need to get up and running with a small Fastify service that exposes one endpoint: /api/health. Our dev script in package.json is set to run the fastify-cli to start our server on localhost port 8000, which is good enough for now. We start up our server:

~/project$ npm run dev
Enter fullscreen mode Exit fullscreen mode

Then, in another terminal window, we use curl to hit the endpoint:

~$ curl http://localhost:8000/api/health 
{"status":"ok"}
Enter fullscreen mode Exit fullscreen mode

Add a simulated long-running process

We’re off to a good start. Next, let’s add another route to simulate a long-running process. This will help us gather some latency data. In routes.js, we add another route handler within our exported default async function:

 fastify.get("/api/user-data", async (_, reply) => {
    await sleep(5000);
    const userData = readData();
    return reply.send({ data: userData });
  });
Enter fullscreen mode Exit fullscreen mode

This exposes another endpoint: /api/user-data. Here, we have a method to simulate reading a lot of data from a database (readData) and a long-running process (sleep). We define those methods in routes.js as well. They look like this:

import fs from "fs";

function readData() {
  try {
    const data = fs.readFileSync("data.txt", "utf8");
    return data;
  } catch (err) {
    console.error(err);
  }
}

function sleep(ms) {
  return new Promise((resolve) => {
    setTimeout(resolve, ms);
  });
}
Enter fullscreen mode Exit fullscreen mode

With our new route in place, we restart our server (npm run dev).

Measure latency with curl

How do we measure latency? The simplest way is to use curl. Curl captures various time profiling metrics when it makes requests. We just need to format curl’s output so that we can easily see the various latency values available. To do this, we define the output we want to see with a text file (curl-format.txt):

    time_namelookup:  %{time_namelookup}s\n
        time_connect:  %{time_connect}s\n
     time_appconnect:  %{time_appconnect}s\n
    time_pretransfer:  %{time_pretransfer}s\n
       time_redirect:  %{time_redirect}s\n
  time_starttransfer:  %{time_starttransfer}s\n
  -------------------  ----------\n
          time_total:  %{time_total}s\n
Enter fullscreen mode Exit fullscreen mode

With our output format defined, we can use it with our next curl call:

curl -w "@curl-format.txt" \
     -o /dev/null -s \
     "http://localhost:8000/api/user-data"
Enter fullscreen mode Exit fullscreen mode

The response we receive looks like this:

  time_namelookup:  0.000028s
        time_connect:  0.000692s
     time_appconnect:  0.000000s
    time_pretransfer:  0.000772s
       time_redirect:  0.000000s
  time_starttransfer:  5.055683s
                       ----------
          time_total:  5.058479s
Enter fullscreen mode Exit fullscreen mode

Well, that’s not good. Over five seconds is way too long for a transfer time (the time it takes the server to actually handle the request). Imagine if this endpoint was being hit hundreds or thousands of times per second! Your users would be frustrated, and your server may crash under the weight of continually re-doing this work.

Redis to the rescue!

Caching your responses is the first line of defense to reduce your transfer time (assuming you’ve addressed any of the poor programming practices that might be causing the latency!). So, let’s assume we’ve done everything we can do to reduce latency, but our application still needs five seconds to put this complex data together and return it to the user.

In our scenario, because the data is the same every time for every request to /api/user-data, we have a perfect candidate for caching. With caching, we’ll perform the necessary computation once, cache the result, and return the cached value for all subsequent requests.

Redis is a performant, in-memory key/value store, and it’s a common tool used for caching. To leverage it, we first install Redis on our local machine. Then, we need to add Fastify’s Redis plugin to our project:

~/project$ npm i @fastify/redis
Enter fullscreen mode Exit fullscreen mode

Register the Redis plugin with Fastify

We create a file, redis.js, which configures our Redis plugin and registers it with Fastify. Our file looks like this:

// redis.js

const REDIS_URL = process.env.REDIS_URL || "redis://127.0.0.1:6379";

import fp from "fastify-plugin";
import redis from "@fastify/redis";

const parseRedisUrl = (redisUrl) => {
  const url = new URL(redisUrl);
  const password = url.password;
  return {
    host: url.hostname,
    port: url.port,
    password,
  };
};

export default fp(async (fastify) => {
  fastify.register(redis, parseRedisUrl(REDIS_URL));
});
Enter fullscreen mode Exit fullscreen mode

Most of the lines in this file are dedicated to parsing a REDIS_URL value into a host, port, and password. If we have REDIS_URL set properly at runtime as an environment variable, then registering Redis with Fastify is simple. After configuring our plugin, we just need to modify app.js to use it:

// app.js

import redis from "./redis.js";
import routes from "./routes.js";

export default async (fastify, opts) => {
  fastify.register(redis);
  fastify.register(routes);
};
Enter fullscreen mode Exit fullscreen mode

Now we have access to our Redis instance by referencing fastify.redis anywhere within our app.

Modify our endpoint to use caching

With Redis in the mix, let’s change our /api/user-data endpoint to use caching:

  fastify.get("/api/user-data", async (_, reply) => {
    const { redis } = fastify;

    // check if data is in cache
    const data = await redis.get("user-data", (err, val) => {
      if (val) {
        return { data: val };
      }
      return null;
    });

    if (data) {
      return reply.send(data);
    }

    // simulate a long-running task
    await sleep(5000);
    const userData = readData();


    // add data to the cache
    redis.set("user-data", userData);


    return reply.send({ data: userData });
  });
Enter fullscreen mode Exit fullscreen mode

Here, you see that we’ve hardcoded in Redis a single key, user-data, and stored our data under that key. Of course, our key could be a user ID or some other value that identifies a particular type of request or state. Also, we could set a timeout value to expire our key, in the case that we expect data to change after a certain window of time.

If there is data in the cache, then we’ll return it and skip all the time-consuming work. Otherwise, do the long-running computation, add the result to the cache, and then return it to the user.

What do our transfer times look like after hitting this endpoint two more times (the first one to add the data into the cache, and the second one to retrieve it)?

   time_namelookup:  0.000023s
        time_connect:  0.000560s
     time_appconnect:  0.000000s
    time_pretransfer:  0.000729s
       time_redirect:  0.000000s
  time_starttransfer:  0.044512s
                       ----------
          time_total:  0.047479s
Enter fullscreen mode Exit fullscreen mode

Much better! We’ve reduced our request times from several seconds to milliseconds. That’s a huge improvement in performance!

Redis has many more features that may be useful here, including having key/value pairs timeout after a certain amount of time; that’s a more common scenario in production environments.

Using Redis in your Heroku deployment

Up to this point, we’ve only shown how this works in a local environment. Now, let’s go one step further and deploy it all to the cloud. Fortunately, Heroku provides many options for deploying web applications and working with Redis. Let’s walk through how to get set up there.

After signing up for a Heroku account and installing their CLI tool, we’re ready to create a new app. In our case, we’ll call our app fastify-with-caching. Here are our steps:

Step 1: Login to Heroku

~/projects$ heroku login
...
Logging in... done
Enter fullscreen mode Exit fullscreen mode

Step 2: Create the Heroku app

When we create our Heroku app, we’ll get back our Heroku app URL. We take note of this because we’ll use it in our subsequent curl requests.

~/project$ heroku create -a fastify-with-caching
Creating ⬢ fastify-with-caching... done
https://fastify-with-caching-3e247d11f4ad.herokuapp.com/ | https://git.heroku.com/fastify-with-caching.git
Enter fullscreen mode Exit fullscreen mode

Step 3: Add the Redis add-on

We need to set up a Redis add-on that meets our application’s needs. For our demo project, it’s sufficient to create a Mini-tier Redis instance:

~/project$ heroku addons:create heroku-redis:mini -a fastify-with-caching
Creating heroku-redis:mini on ⬢ fastify-with-caching…
…
redis-transparent-98258 is being created in the background.
…
Enter fullscreen mode Exit fullscreen mode

Spinning up the Redis instance may take two or three minutes. We can check the status of our instance periodically:

~/project$ heroku addons:info redis-transparent-98258
...
State:        creating

Enter fullscreen mode Exit fullscreen mode

Not too long after, we see this:

State:        created
Enter fullscreen mode Exit fullscreen mode

We’re just about ready to go!

When Heroku spins up our Redis add-on, it also adds our Redis credentials as config variables attached to our Heroku app. We can run the following command to see these config variables:

~/project$ heroku config -a fastify-with-caching
=== fastify-with-caching Config Vars

REDIS_TLS_URL: rediss://:p171d98f7696ab7eb2319f7b78083af749a0d0bb37622fc420e6c1205d8c4579c@ec2-18-213-142-76.compute-1.amazonaws.com:15940
REDIS_URL:     redis://:p171d98f7696ab7eb2319f7b78083af749a0d0bb37622fc420e6c1205d8c4579c@ec2-18-213-142-76.compute-1.amazonaws.com:15939
Enter fullscreen mode Exit fullscreen mode

(Your credentials, of course, will be unique and different from what you see above.)

Notice that we have a REDIS_URL variable all set up for us. It’s a good thing our redis.js file is coded to properly parse an environment variable called REDIS_URL.

Step 4: Create a Heroku remote

Finally, we need to create a Heroku remote in our git repo so that we can easily deploy with git.

~/project$ heroku git:remote -a fastify-with-caching
set git remote heroku to https://git.heroku.com/fastify-with-caching.git
Enter fullscreen mode Exit fullscreen mode

Step 5: Deploy!

Now, when we push our branch to our Heroku remote, Heroku will build and deploy our application.

~/project$ git push heroku main
...
remote: Building source:
remote: 
remote: -----> Building on the Heroku-22 stack
remote: -----> Determining which buildpack to use for this app
remote: -----> Node.js app detected
remote:        
remote: -----> Creating runtime environment
...
remote: -----> Compressing...
remote:        Done: 50.8M
remote: -----> Launching...
remote:        Released v4
remote:        https://fastify-with-caching-3e247d11f4ad.herokuapp.com/ deployed to Heroku
remote: 
remote: Verifying deploy... done.
Enter fullscreen mode Exit fullscreen mode

Our application is up and running. It’s time to test it.

Test our deployed application

We start with a basic curl request to our /api/health endpoint:

$ curl https://fastify-with-caching-3e247d11f4ad.herokuapp.com/api/health
{"status":"ok"}
Enter fullscreen mode Exit fullscreen mode

Excellent. That looks promising.

Next, let’s send our first request to the long-running process and capture the latency metrics:

$ curl \
  -w "@curl-format.txt" \
  -o /dev/null -s \
  https://fastify-with-caching-3e247d11f4ad.herokuapp.com/api/user-data

     time_namelookup:  0.035958s
        time_connect:  0.101336s
     time_appconnect:  0.249308s
    time_pretransfer:  0.249389s
       time_redirect:  0.000000s
  time_starttransfer:  5.384986s
  -------------------  ----------
          time_total:  6.554382s
Enter fullscreen mode Exit fullscreen mode

When we send the same request a second time, here’s the result:

$ curl \
  -w "@curl-format.txt" \
  -o /dev/null -s \
  https://fastify-with-caching-3e247d11f4ad.herokuapp.com/api/user-data

     time_namelookup:  0.025807s
        time_connect:  0.091763s
     time_appconnect:  0.236050s
    time_pretransfer:  0.236119s
       time_redirect:  0.000000s
  time_starttransfer:  0.334859s
  -------------------  ----------
          time_total:  1.276264s
Enter fullscreen mode Exit fullscreen mode

Much better! Caching allows us to bypass the long-running processes. From here, we can build out a much more robust caching mechanism for our application across all our routes and processes. We can continue to lean on Heroku and Heroku’s Redis add-on when we need to deploy our application to the cloud.

Bonus Tip: Clearing the cache for future tests

By the way, if you want to test this more than once, then you may occasionally need to delete the user-data key/value pair in Redis. You can use the Heroku CLI to access the Redis CLI for your Redis instance:

~$ heroku redis:cli -a fastify-with-caching
Connecting to redis-transparent-98258 (REDIS_TLS_URL, REDIS_URL):
ec2-18-213-142-76.compute-1.amazonaws.com:15940> DEL user-data
1
Enter fullscreen mode Exit fullscreen mode

Conclusion

In this tutorial, we explored how caching can greatly improve your web service's response time in cases where identical requests would produce identical responses. We looked at how to implement this with Redis, the industry-standard caching tool. We did this all with ease within a Node.js application that leverages the Fastify framework. Lastly, we deployed our demo application to Heroku, using their built-in Heroku Data for Redis instance management to cache in the cloud.

Happy coding!

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .