Parse-server: Google Cloud + Mlab vs Digital Ocean

[nickkuh]

Hi There

I've been running a production app via Google Cloud with the DB hosted on mlab's servers.

However, the costs are starting to escalate - seems that mlab's $15 per month quickly increases if you go over 2GB database (last month's bill was $50 just for mlab storage).

Now that I'm out of the initial free 2 months on GoogleCloud Platform it also looks like I'm going to be paying an additional $50 per month to Google based on my app's resource usage.

I'm thinking of switching both my database and parse server to be hosted on a single Digital Ocean droplet which amounts to about $20 per month for 2GB CPU / 40GB SSD / 3TB Transfer

Just wondering what other people's experience of Digital Ocean has been like? It's potentially a big saving but I don't want to jeopardise the performance of my app.

It seems to me if my parse server is accessing mongodb on the same server it should make cloud code queries faster (a lot of my app runs in quite heavy cloud code functions).

Any thoughts? Anyone got an experience of running a parse server production app on a Digital Ocean Droplet?

Apologies if this is miss-use of the issues section of the github repo but it seems like a good place to ask this question.

Cheers

Nick

[ajamaica]

Hi!

I Have experience with both. There are 2 types of projects I manage to run. Some ones use a lot of Cloud Code to work and others need more Fast Queries for serve. You can manage to run (1 thread) Parse-server and MongoDB (not clustered) in a 512 mb of ram droplet if you use the SSD of a Digital Ocean VP it works well for 10-20 reqs per second without any complicaded cloudcode. Works for most of my clients. This cost 5USD month. NGINX-FOREVER + MongoDB

In the other hand there are the apps that need 50+ or more request per second. This was due alot of cloud code. This migration was for a client that pay about 1700 USD at parse.com. This company relay a lot on Jobs at parse.com, so there was another component that add more request. What we did was to make a separated MongoDB instance + 2 8v High CPU machines at GCP. + Other Parse that runs KUE for the jobs. A small router server was needed, we put NGINX. this setup is about 800 USD. So the customer was a big powers with less than a half the price. This database was huge.

With these experience I select g1-small (1 vCPU, 1,7 GB) with pm2 with nginx and mongo runing ubuntu. There cost cheaper than the appengine one and I manage to get up to 30+ req/seq.

A hot trick is to change express to production mode NODE_ENV=production this will incresse the performance by 1/3.

• I recoment forever-service for 1 thread and 1 CPU and PM2 for multiple Cores in vCPU High CPU.

Arturo

[Knana]

I would recommend to you NodeChef. We help customers with their parse server apps. The performance we offer and the price can't be beat. Our databases run on 8 physical cores that is the equivalent of either c4.2xlarge or r3.2xlarge on AWS on bare metal. We have customers performing 150 + req/sec performing complex adhoc queries using just 2 app containers of 512 MB ram.

[nickkuh]

To others considering it... I made the leap about 3 weeks ago to running both my mongodb and parse server app on the same Digital Ocean droplet.

Experience so far has been incredibly positive.

I store a lot of data in my app and the costs for mlab rocket up as soon as you go over 2GB. I think I his 6GB and they billed me $100 for last month's storage.

Google's Cloud Engine is much more reasonably priced for running Parse Server (and very fast and stable) however I couldn't figure out how to run mongodb locally to the Parse Server installation for this setup.

I migrated to the $20 per month / 40GB Digital Ocean Droplet and have saved myself over $100 in the last 3 weeks alone.

By running mongodb and parse server on the same server the api calls to my app have become much much faster, so not only am I saving money but the experience is greatly improved for my customers.

If you're thinking of making the same change you will need to play more of an admin role than with the serviced options. But in light of the fact that we're all having to learn a lot more back end code I don't consider this such a bad thing.

[hramos]

We're closing this issue due to inactivity.

If this is a bug you care about that is not getting attention, consider opening a pull request with a fix.

[shizar1159]

Hello all,

I know it is a high-level topic. I hope I do not cut your conversation. Since I am all new to mbaas, I am having some difficulties. I have gone through a series of tutorials that involved Parse. But Parse is being retired now and I don't know how to work or eve start on other options.
Can anyone help me learn how to work with google cloud engine?

Thanks!