wispy sand
hey remix runners! 💨
i have a question relating server usage and scale.
Remix shifts a lot to the server away from the client/browser. Do we know how high traffic platforms with a lot of authenticated users impact the server?
I am curious what would the costs be like.
i guess I was wondering if there are examples like:
I tried to get an answer on general channel but maybe here people can still add data and i can find it later on too.
server usage - scale
storm vine
Making a few connections here to some specific people who have large sites:
@candid mirage do you have any insights to share about the costs of hosting 2M+ views per month on Remix for https://udisc.com/
@sturdy knoll do you have any cost insights you can share about Daffy?
@deft shoal do you have any costs insights you can share about your ~1M page views per month on https://kentcdodds.com
deft shoal
My costs are around $75/month for 500k visitors a month. Pretty inefficient use of memory on my site and I've scaled it more than I need probably.
urban bolt
My costs are around $75/month for 500k visitors a month. Pretty inefficient use ...
what vendor are you using?
sturdy knoll
I don’t know how much we paid at Daffy to run the Remix app, but I know we run 8 pods of k8s (I don’t know the size) and we used to run 16, also we over provisioned so we most likely could run less pods
sturdy knoll
what vendor are you using?
Kent’s websites uses fly.io
urban bolt
I don’t know how much we paid at Daffy to run the Remix app, but I know we run 8...
isnt k8 main shtick to have the correct provisions of server and spin more up?
sturdy knoll
isnt k8 main shtick to have the correct provisions of server and spin more up?
You can also configure k8s to always have a min amount of instances running
urban bolt
arnt we all hoping the tumultuous traffic we will hit one day so we can bare the load!
I'm reffering to the over provisioning here 😆
candid mirage
We are currently running 14 instances on fly.io (of 1x shared cpu/512mb RAM) across 7 regions (2 per region).
Our bill was $~500 last month (our highest yet, but only $52.61 of that was instance costs, and $450 of bandwidth (22TB...).
The bandwidth was a bit surprising, and it turns out I had misconfigured the cache-control value for font files (which we recently started self-hosting).
So, aside from that bandwidth mistake, it's quite reasonable. I've only had to scale beyond 14 on a few very large spikes of traffic that were causing some crashes in the servers (due to a now-fixed bug).
deft shoal
I think I'd say that in general fly is extremely reasonable for what you get
candid mirage
Worth mentioning that our previous implementation (meteor/node) was running on 2CPU/4GB ec2 instances and we needed ~8 or more of them for the same or less traffic. (and had to scale up way more often, plus that was only a single region)
deft shoal
Worth mentioning that our previous implementation (meteor/node) was running on 2...
Interesting. So you're saying your resource usage went up a great deal when migrating to remix?
sturdy knoll
I think it’s the inverse
deft shoal
Oh, great!
sturdy knoll
From 2CPU/4GB to 1 shared CPU/512MB
Being able to run with a shared CPU and half GB of RAM it’s really nice
candid mirage
ya, correct @sturdy knoll - quite a bit less resource usage.