Tim ChambersNov 21, 2022

#Admin #Indieweb.social

This instance now has 9,347 people with this as their home base...With being so busy on admin work, I missed us crossing that line.

From instnaces.social: this makes us 0.15% of total 6,150,381users they scan on the whole #Fediverse.

Here's to the medium-small instances of the world, the #Fediverse needs to scale horizontally, with thousands more of this size.

Show threadRaphael Luckom
@tchambers One set of stats that I haven't seen is the traffic volume. Bytes in/out per user per day is one of the biggest potential go / no go questions in most public cloud deployments I can think of. I saw some docs suggesting that the volume makes s3 media storage non-cost-effective but having it quantified would be very useful because it applies to CDNs as well.
Nov 21, 2022 at 6:37pmWeb
Show threadKelly GriffinNov 21, 2022
@raphaelluckom @tchambers I turned up an instance to play my part in the overall scheme. What I don't find is what are the storage requirements per user. What should I be prepared for in the future? Two users today, but what do I need for 1K, 10K, etc.?
Show threadTim ChambersNov 21, 2022
@unwired @raphaelluckom Nice! I've seen estimates that 100 users is about 10 GB.
Show threadKelly GriffinNov 21, 2022
@tchambers @raphaelluckom Wow! Much better than what I feared.I started with 4TB of storage and unlimited 1Gbps bandwidth. It sounds like I am good for awhile. Making the presumption that my server will provide resources for new users. #Presumptuous #WillingToHelp
Show threadAnsel HalliburtonNov 21, 2022
@tchambers @unwired @raphaelluckom FWIW, here's where I'm at with 64 users:
Show threadKelly GriffinNov 21, 2022
@anseljh @tchambers @raphaelluckom Thanks for sharing. Do you think the media is pictures, GIF, etc. or do you think there is quite a bit of video?
Show threadRaphael LuckomNov 21, 2022
@tchambers @unwired I wrote up a blog post describing the #terraform #aws deployment I ended up with for testing and some cost estimates / questions: https://raphaelluckom.com/posts/Notes%20on%20a%20test%20Mastodon%20deployment.html
Notes on a test Mastodon deployment

Show threadSimon ColesNov 21, 2022

@unwired @raphaelluckom @tchambers I'm not so sure it's storage that's the thing.

I don't know what the tipping point of people and followers is, but at some point you end up needing to tune Sidekiq and your database. And that's CPU with a bit of RAM I think.

I'm running my instances using S3 for media and don't anticipate needing to increase my servers' storage as more people join.

You might find this interesting https://nora.codes/post/scaling-mastodon-in-the-face-of-an-exodus/?ref=upstract.com

Scaling Mastodon in the Face of an Exodus | Nora Codes

Show threadKelly GriffinNov 21, 2022
@simonc @raphaelluckom @tchambers With regards to CPU and RAM, wouldn't a distributed model within your own instance be better? Limit each server to "x" number of users?
Show threadSimon ColesNov 21, 2022

@unwired @raphaelluckom @tchambers My understanding is scaling a Mastodon instance involves splitting out Sidekiq and Postgres onto different machines and then tuning it.

I don't know when you need to do that but I think it's 1k+ users.

What's interesting is how Mastodon both technically and by convention encourages lots of smaller instances. So by all means create instances of 10k plus people but the moderation burden will likely kill you before the technical stuff does.

Show threadSimon ColesNov 21, 2022

@unwired @raphaelluckom @tchambers With my limited experience as a Twitter refugee I think Mastodon is about communities gathering on an instance and then federating with others. So lots of small (under 1k users) instances working together.

This "Community gathering on an instance" approach makes moderation easier because there's more of a social contract. It's also technically more achievable because everything fits on a single server, no heroics.