New on LowEndTalk? Please Register and read our Community Rules.
All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
Server Recomendation
The thing is, I don't know what kind of server I needed. I am talking about 5000 simultaneous users.
My Current setup consist of
- Nginx as load balancer and caching, use about 256MB of RAM and 1 GHz of CPU.
- Apache webserver (frontend), use about 2GB of RAM and 10GHz of computing power
- mysql server, use about 2 GB of RAM
- Nodejs (backend), use about 2GB of RAM 1.6GHz of CPU.
Should I cramp them all in 1 dedicated server, if so, what spec am I looking for ?
or should I buy a few vps in 1 provider and keep them separated / isolated ?
My budget is around $20 - $50 a month.
Any suggestion is welcomed
Comments
Is your current setup working okay? Do you foresee growth in users?
I would prob just get a dedi from Soyoustart with SSD drives.
Since you're running node, get the best single core performance you can get since node is single threaded language.
I made the same mistake already. I got a server with 6 cores 12 threads and it just slowed down just to find out node actually sucks when it comes to multithreaded workload.
Weird, there was just similar question to this ... and i've seen other things where 2 similar threads popup very close to each other today.
You will want high I/O (ie. EPYC, Ryzen) system and work on your config.
I'd put everything under the same dedicated. There doesn't seem to be a need to separate MySQL for now.
I dont see the advantage of having them in 4 separate VPS under the same provider. You're most likely under the same node, so if the node goes down for whatever reason all 4 of them go down. If the software crashes in any of those 4 VPS, your app/website would lost likely not load anyways.
Running 4xOS is an overhead and having a bunch of open ports is not the best security practice.
Put everything under one decently powered dedi, move away from Apache to Nginx if possible.
Here are some suggestions:
Hetzner's AX41-NVME would have more than enough power to handle everything but the setup seems to be delayed.
SYS Game-1 would lost likely be also decent if you dont need ECC memory and a lot of storage.
SYS E3-SSD-1-32 similar to Game-1, but less CPU power and more RAM and Storage. Still no ECC memory.
It's working okay. But there has been incident and it has been down for 2 hours. Good thing, it's down in the middle of the night. Even then I have to explain to my clients why their app is down.
Imagine the chaos, when it's down in work days, I personally have experienced it. Saying it's a Disaster is an understatement.
This is what you called Low End Mission Critical App. I need contingency.
Is it possible to put load balancer and spawn multiple Node Server ?
Thanks for the recommendation. How is SYS tough ? are they reliable ?
you can use Litespeed webserver instead of Nginx or apache so your current server works well and you don't need to move somewhere else
It is possible. You can start with PM2's cluster mode to test it out.
My current server and setup works well. Just need more availability. Moving to litespeed is not an option. I just spent the last month to learn how nginx works.
Also apache is just there to server static content, which is cached by the nginx anyway. Any processing is done by the nodejs.
Then you should go with someone who promises sub 2hr response times (super rare and expensive) - and/or build in redundancy by design. Or in other words, spread the application across multiple DCs etc.
You are not going to find what you are looking for with any LET regular provider. Easiest is just to go with something like AWS, Azure, Upcloud. In any case expect to add an zero (or several) to your monthly expense.
Hell, even some banks do not reach sub 2hr response times during server issues and just last week i believe a major major brokerage was down roughly ~close full day in 2 consecutive days during most important market days.
Hetzner, OVH (SYS) have the head count, but they have so many customers they just cannot respond to you in 2 hours typically - and in case of OVH they might just swap a HDDs without asking you even if they are fine (just had this happen with our stupidly old wiki server).
So your best bet is to distribute your load across many servers across many providers.
Get a dedicated, if you want to archive the best uptime, you can only do it with that.
Compared to VPS's they just run, the impact of other people doing something is nearly zero.
Use the VPS's as backup if your machine goes into maintenance or goes down.
Mainboard/PSU can fail out of nowhere, but its very rare.
A few key things
Learn to monitor and analyse server resource usage. If you do not measure and monitor, you'll have no real idea of your requirements and where you current bottlenecks are.
Learn to load test/benchmark your site. I'd also breakdown individual testing for your Nginx, Apache, MySQL and Nodejs layers as well. Combined with monitoring, you'll be able to accurately measure your current performance and when you tip that breaking point when you need to upgrade your server.
Not all dedicated/vps servers are created equal as are not all web servers created equal due to how they're configured - hardware and software wise. If you learn the first 2 items, you'll be able to spot that more easily. Example in my 13-way VPS benchmark comparison for Upcloud, DigitalOcean, Linode, Vultr and Hetzner https://community.centminmod.com/threads/13-way-vps-server-benchmark-comparison-tests-upcloud-vs-digitalocean-vs-linode-vs-vultr-vs-hetzner.17742/
For high concurrency traffic on a budget, caching (both frontend + backend) is key and you'll only be able to perform as well as the slowest server in your current VPS setup.
As your budget is restrictively low for 5k concurrent visitors, I'd try hourly billed VPS providers with a test copy of your setup and load test/benchmark and measure how they perform. As they're hourly billed VPS, you could end up testing several VPS providers within the same month for much less than the cost a full month usually. Only way to do it, as no one here knows accurately your real usage requirements and performance of your current setup. Then you can take what you have learnt and decide on final server destination, whether it be dedicated or a VPS provider.
HTH