[Back-end] How to scale my side project to handle 100k users?

@prasanna, based on the comments about architecture (nodejs, mongodb, Digital Ocean)... here’s what I’d suggest you have covered:

• put a load balancer in front of your servers so you can scale horizontally with demand
• migrate your DB to Mongo Atlas... last thing you want is to manage DB crashes, scaling, and backups
• serve static assets from a CDN designed to handle scale (Netlify, Cloudflare, AWS CloudFront)

That should give you scalability on demand.

Depends, on lots. But if your APIs are heavy on 'read' requests you can just cache those. But make sure to expire that cache correctly, when that should be depends on your application.

If you need any help feel free to email me.

I've worked for a number of banks and never underestimate the benefits of transferring a CSV.

You can run the import multiple times, and at your own pace, sorting out any issues that pop-up while doing so.

It's a lot easier than going back and forth with your client asking them to retry their API requests that failed because of a data issue or a rate limit.

Hi, really hard to say without knowing your app.

But some general advice is make sure you have pagination on any list screens and stuff.
I'd also favour csv/json import over API if it makes sense for what your model. Those import features will be easier for more customers to use and can be re-used for bulk updates.

Are you doing any blocking operations? Are your services compute heavy? Perhaps going with a micro service architecture might be a solution. It all depends on the problem and what is causing it.

What scaling you need depends on the expected usage pattern. Is the usage throughout the day, only during work hours or spikes at particular times (e.g. lunchtime)?

I would personally not worry about scaling too much since you don't know the bottle necks just yet. (EDIT: assuming you don't know them - if you know them, that's great)

What will make your life easier to make sure you have the capability to throw money at the problem. Scale your servers horizontally (i.e. use a load balancer) and your database vertically (just get a bigger size). Lastly, have some kind of monitoring on the basics (CPU, memory) logged somewhere.

Also does the client plan to pull the switch at once or migrate users over time? It might be easier to ask if they can stagger adoption.

Best case scenario is that nothing breaks and your current setup is fine. Worst case, you eat some upfront costs to scale things quickly up using money (with a little bit of downtime) and you get to discover where your bottlenecks are (this is where logging comes in).

Congrats on the success!

You should stress test your setup. If you say 'you don't know' it means perhaps there's nothing to worry about. Once you know, you can describe your scaling problem better. Caching smartly, buying a bigger server, load balancing, moving DB, going serveless...

For migrating, do whatever will be quicker for you that's acceptable to your user. I, personally, wouldn't feel comfortable sharing plain text files of 50k customers if they include sensitive data.

It really depends on the bottleneck. Are you sure you can’t handle the load? Is the db the problem? Number of concurrent users? Number of users period?

Splitting the logic apart into something that can scale could be challenging. Sounds like you need to refactor so that more hardware makes more users possible. But any other deets would tell us more

Scaling , first step is to put the database on a different server. For 100 k users this may be ok. If the load is still too much shard the database. On code side find the busiest sections and put those on multiple servers with a load balancer.

Have you considered writing a load test?

Then the next step is to monitor which resource is the bottleneck, optimize, repeat.

Good luck!

Where is your site hosted? Scaling up on a service like Netlify for example is easy but will just cost you more. Scaling up on AWS is a it trickier depending on your setup but adding a load balancer with a couple beefy servers there would probably cover you.

As far as importing users, one option is to build yourself an admin interface that uses an internal API. Then you can add an import users from CSV interface you will be able to use.

Hey Prasanna, It depends on how you're storing your data today and where are you planning to move it. Maybe you could just scale up your server as a quickfix and work in a better architecture in the meantime.

I'm running GetAdvice (https://getadvice.github.io), you could schedule a free one-to-one session call with one of our volunteers to help you analyze the problem and design a solution. The project was created with this type of uses cases in mind so feel free to check it out.

My profile: https://getadvice.github.io/adriano.galello.html

Hi,

If your target is a more robust/HA infrastructure on AWS or GCP we can maybe assit you. We are cloud experts offering an affordable and more approachable cloud consultancy. We designed several infrastructure for our clients that are flexible, elastic and so on and operated loads of cloud migrations. We would need more information about your existing setup to figure out what would be the best option. Send us a mail at info@cloudbuddies.io if you are interested :)