Serving 400 DB writes per minute with bare minimum infrastructure

To give a brief intro - Blanq is a custom URL shortener
that supports link redirection, analytics and more.

It is primarily a DB write-heavy application since analytics requires every redirection to be saved in the database.

When I first started, this is how my stack looked like on Digital Ocean:

1. A single $5 instance that ran both the Java application and the Mysql.
2. Rollbar for error monitoring

Pretty basic I would say but, it made sense since the only user was me.

After posting on Indiehackers, I landed up a couple of pretty large customers - which was totally unexpected.

These customers had in total a million followers on social media each.

It hit me then! My app cannot handle the huge amount of traffic coming my way.

I increased the droplet size to $25 that provided me with enough headroom in case there was a spike in the traffic.

And it worked!

It could now handle 200 writes per minute without dropping any request.

But...

It was not perfect. I was constantly worried about losing database or getting hacked. I had no confidence in my backup script. I did not want to part away with my data by using third-party backup services. They were also a bit expensive. Also, I was not sure about how my application was performing in the real world.

So I,

1. Migrated the DB to Digitalocean's managed database. It comes with free daily backups and security patches. Peace of mind ensured.
2. Installed New Relic agent that has a decent free tier plan. Now I can see what the hell is happening in my application and database(Free).
3. Tweaked my application code to take in a lot more requests.
4. Cached all the redirection requests.
5. Set up Rollbar alerts in case things go wrong(Free).
6. Integrated Slack alerts with downtime monitoring service(Free).

Though Managed database's entry plan is not the best - it is good enough to start with and, a lot better than running my own cluster.

I reduced the instance size to $15 plan since I no longer had to run MySQL DB. I plan to further reduce it to $10 instance since it provides me with plenty of headroom for growth and spikes in traffic.

My current costs are $30 per month and the application serves 400 requests per minute(DB writes) during peak traffic hours. Quite decent I guess.

What's next:

1. DB tables do not have indexes on all columns. Need to fix that.
2. Add a slave to the MySQL cluster as a failover.
3. Further optimise reads to lower response times.

I prefer to keep things simple and postpone as much as possible to save costs and time.

It was also a great learning experience for me in the last 4 months and a roller coaster one.

Thanks for reading :)

👉 Try Blanq to shorten your links.