Hacker News new | ask | show | jobs
by comboy 2999 days ago
I have a few dedicated boxes for almost 10 years now. I highly recommend it. I use it to run bitcoinity.org since it was created in 2011. Because of relatively big database (2TB) and many concurrent websocket connections (peaking around 20K) cost would be prohibitive elsewhere. Great uptime, instant responses, eKVM and blah blah I'm not their marketing team, I'm just a really happy customer.

I've tried OVH I think on two or three occasions, mostly VPS which was really sucked. But that was a few years ago maybe they got their s..ervers together.
1 comments
truetraveller 2999 days ago
How many servers do you have with Hetzner to manage 20,000 websocket connections? Also, did you ever have data loss in your large database because of hard drive failure?
link
swsieber 2999 days ago
Theoretically, they could get away with just one: http://phoenixframework.org/blog/the-road-to-2-million-webso...

That said, it depends on the stack, the level of optimizations and how much each websocket requires...

I'd love to know too though (both the # of servers and the other details)

link
comboy 2998 days ago
Yup, I switched to phoenix lately. Works as advertised. I didn't even mess much with any kernel options. I still have most of my stack on ruby (on rails), but I bridge more and more things with elixir app (and phoenix), mostly using redis queues. As mentioned just 2 boxes are doing ws pushing currently, they also serve the app, one additional server also serving app but mostly static files, DB server, DB-slave and the app logic (which in my case is mostly fetching data and calculating some stats).

Unfortunately, even though there are some fallbacks and it seems to be pretty reliable, these servers are pets rather than cattle.

link
comboy 2998 days ago
Two are doing just fine with the current stack (elixir), I actually have more problem with gigabit connections getting saturated at some points, but that's just my failure to optimize data that's being sent. Before that I was using nginx module (nginx-push-stream-module) which is also great and way ahead of any other out-of-the-box solution that I know.

Every server has RAID1 by default, plus I keep postrgresql slave on another box and I try to keep up to date backup. I actually never experienced a drive failure there (I currantly have ~10 boxes), but on my dev server at home which runs the same service I got I guess 5 HDDs failures already. Just two of them were server-grade though.

link