Two Network Volumes
Id like to have an endpoint with a network volume but across data centers. I don’t mind if network volumes are in a different state as they act like a cache, however i’d like my traffic to be load balanced.
Is that possible ?
16 Replies
Not possible
You can only attach a single network volume to an endpoint
we are planning to allow this in serverless, my guess is sometime in march
Is this updated now?
No not yet
If I had the same need I would approach it by running a rsync server in a CPU pod. I would then create a serverless worker with 0 active workers and 1 max worker in every region with cheapest available CPU/GPU in that region and attach it to local storage in that region to it. You could then use cron on the rsync server to schedule syncs of the storage with the rsync workers. You could also configure your other serverless workers to connect to the pod and force a sync if they notice their models are not available. This solution with 10GB of global storage would cost ~ $52/month to run on RunPod. $43 of which is for the pod so adding additional storage is very inexpensive.
RunPod could implement the solution above and make it multi tenant and they would have true multi region cloud storage practically overnight. All the pieces are already in place.
not every region has cpu pod, yeah well depending on what config the sync is
every type has it own pros and cons
Yeah, EU-RO-1 would be where you host the rsync server so it could run on a CPU/cheapest($43.2/month) instance since it would need to run 24/7/365. RunPod should really give a discount if you run a CPU pod 24/7/365 since you can get equivalent from DO for $6-12/month. EU-RO-1 wouldn't need a rsync worker since the storage in EU-RO-1 would be the master copy. All the other 12 regions would have to be run on GPU instances. But since they are configured for active workers of 0 and max of 1 the majority of the time they will not running . On days when the disk is already in sync the workers will likely run under a minute per day, $0.144 on 16GB GPUs, for all 12 workers to check in. Imagine a heavy sync day that takes up to 5 minute to sync. 12 workers using 16GB GPU machines for 5 minutes would cost $0.72 for that day. That is why I love RunPod, it is the only service in the world, that I am aware of, that if you plan out a deployment you can scale down very near absolute 0!
haha detailed plan
sure sure feel free to apply that
I am currently in the building something phase. Once I get closer to releasing my Saas I'll need to develop something like that, if RunPod hasn't done so beforehand. So, I've been thinking about it. You're today's lucky soundboard participant, thanks for playing along! LOL
ooh what saas are you building?
It's a suite a tools for video editing... incorporates a lot of AI. Makes it easy for people to create content for social.
Wow that's a great idea
I'm curious how's it
???
One of the tools I am nearly finishing up with converts mp4 movies into ascii art videos. One video I posted on TikTok a couple days ago got 23.5k view and 5K likes. https://www.tiktok.com/@artisanascii/video/7378275383113731371?is_from_webapp=1&sender_device=pc&web_id=7377057546446472746
TikTok
TikTok · Artisan ASCII
5424 likes, 26 comments. “[Meme] - She's the bomb in anime ASCII - Artisan ASCII”
I'm calling it Artisan ASCII
Ooh nc
Alot of miku fans