What infrastructure does this site run on?

[Elysium]

Hi!
Sorry if this is not the right place for this question, but I'm curious, what sort of hardware or VM provider is hosting this site.
From a quick snoop I could tell that this is an nginx webserver running on Ubuntu hosted by DigitalOcean.
I just wonder what sort of IT infrastructure does it take to run and administer a forum in this day and age.

 

[overscan (PaulMM)]

It's a Digital Ocean droplet running Ubuntu and the software stack is nginx, PHP, MariaDB and redis. The forum software is Xenforo which is commercial. It is backed up to my home Linux server.

It's very simple for me to manage, but I am an IT engineer with 30+ years experience with UNIX OS.

 

[overscan (PaulMM)]

Once the server is running, it's pretty much just installing updates.

Every few years I migrate the forum to a new host with the latest Linux and Xenforo CMS.

 

[Elysium]

Thanks for the info. So is everything (db, forum, hosting content) running on the same machine, or is it more elaborate?

It's crazy that with all this cloud stuff going on, where if you asked a cloud guy to design a forum architecture, you'd end up with a cluster of machines running docker, a separate cluster for the DB, a third party CMS for the content and a load balancer in front of it all, managed by something like Kubernetes. No availability zone splitting, no multiple geographic locations, etc.

It's refreshing to see that you can run a production website used by many users on such straightforward infrastructure.

 

[overscan (PaulMM)]

Everything is on one server.

For the next major update I will switch to storing attachments in an object store (DigitalOcean Spaces, an S3 clone) which will be cheaper and expandable indefinitely, but it will stay single server.

If it was a commercial enterprise making money, I could indeed have a highly available Kubernetes cluster, failover database cluster, load balancer etc. I work with these kind of technologies at work.

However, the downtime caused by OS patching on a reliable, single purpose virtual server is no more than a minute per month. So long as nothing goes wrong at the Cloud provider level, this is a level of reliability that is very hard to improve on.

The additional complexity of a clustered, multi-server solution is as likely to cause more outages (and more complicated failure scenarios) as it is to prevent them, and the cost would be 3 times greater at least. It would probably be more work to manage too.

If the forum audience was a lot larger, my single server solution might not scale to fit.

 

[Archibald]

You two might as well speak klingon or aramean, but I can see the point nonetheless. The proverbial KISS - Keep It Simple, Stupid.

 

[Elysium]

Well, Kubernetes is apparently the Greek word for 'helmsman', so if you've said 'This sounds Greek to me', you'd have been correct And while throwing all these buzzwords around might make it sound Star Trek-y, in practice it's tremendously boring and mundane.
Jokes aside, managing all this complexity IRL is a huge pain in the butt, and as overscan mentioned, a simple KISS architecture is far more reliable in practice than this.

I wonder if this idea has real world aerospace implications, where world-class engineers working with blank cheques make sure to 'do everything correctly' by making everything doubly-triply redundant with a ton of failsafes and contingencies, with the resultant design ending up far more capricious and hard to keep in working order than a simple system would've been.

 

[aim9xray]

A: No, you don't don't understand. Ask your question again.

Q: All right, what does this website run on?

A: Funding! No bucks, no Buck Rogers!

 

[overscan (PaulMM)]

Well, Kubernetes is apparently the Greek word for 'helmsman', so if you've said 'This sounds Greek to me', you'd have been correct And while throwing all these buzzwords around might make it sound Star Trek-y, in practice it's tremendously boring and mundane.
Jokes aside, managing all this complexity IRL is a huge pain in the butt, and as overscan mentioned, a simple KISS architecture is far more reliable in practice than this.

I wonder if this idea has real world aerospace implications, where world-class engineers working with blank cheques make sure to 'do everything correctly' by making everything doubly-triply redundant with a ton of failsafes and contingencies, with the resultant design ending up far more capricious and hard to keep in working order than a simple system would've been.

I would say Sidewinder was an exercise in making the simplest possible missile. It seemed to do pretty well compared to its more complex contemporaries

 

[overscan (PaulMM)]

A: No, you don't don't understand. Ask your question again.

Q: All right, what does this website run on?

A: Funding! No bucks, no Buck Rogers!

That's true.

The new version of the forum software is up to Beta 3, so hopefully will be here in a few months. This will be a good time for a new server and a spring clean.