Dan HonNov 8, 2022

Personal instance hosting update:

This is what my queues look like after I do a Thread:

* Pull queue, 404 items, 412.01 latency/6 minutes
* Push queue, 2,958 items, 341.69 latency/5 minutes

I've now upgraded to the $19/mo Star tier at masto.host: "medium" federation capacity, 8 threads. This is a single-user instance (me) with 2.79k followers and 1.29k following.

Show threadDan Hon
ugh I forgot to add tags #mastodon #personalinstances #personalhosting #mastohost #performance
Nov 8, 2022 at 9:50pmWeb
Show threadDan HonNov 8, 2022

Ahahaha, I replied to a bunch of replies, and now:

* pull queue 2 / 0.75s latency
* push queue 2,977 / 1,242.45s latency/20 minutes

#mastodon #personalinstances #activitypub #mastohost #performance #latency

Show threadDarius KazemiNov 8, 2022
@danhon good lord how do you have 2.7k followers in just a few days? did you migrate them over from another account? or are you just that popular
Show threadJesse VincentNov 8, 2022
@darius @danhon is a rockstar
Show threadDan HonNov 8, 2022

@jesse @darius

a wordcel [UGH]

Show threadDan HonNov 8, 2022

@darius I like following lots of people and some of them follow me back, and then I write things and then other people share them and then...

I mean it's not like I'm *trying* very hard?

Show threadJordan CarlsonNov 8, 2022
@danhon @darius but you are very good at it. i found your work via someone's newsletter in... 2014? 2015? then found you on twitter, and we've interacted occasionally since. so a follow here was pretty quick
Show threadScott JensonNov 8, 2022
@danhon Yeah, I think you're discovering what 'being popular' does to the ActivityPub protocol. I'd guess that having lots of users shares that cost across the entire instance. As a solo instance with one high traffic user, it's all on you.
Show threadDan HonNov 8, 2022

@scottjenson To be honest, I did this deliberately to see what performance would be like and to validate masto.host's claims about its tiers and their suitability for various sizes of federating capacity.

It's quite... illuminating and I have to admit, kind of what I expected.

Show threadScott JensonNov 8, 2022
@danhon That's actually pretty cool (and insightful) I'd like to get some Mastodon dev to comment though so we get a clearer idea of what is going on. For example, I've heard one person is running their instance on a Raspberry Pi! (I doubt they have many followers) My point is that it DOES work if X (user count) is small enough. What is good rule of thumb for X?
Show threadblaineNov 9, 2022
@scottjenson @danhon obviously a different scenario, but less different than you might think, and hopefully a useful data point: https://mastodon.social/@blaine/109310597624930102
Show threadDan HonNov 8, 2022

@scottjenson That's what was confusing for me: clearly there's a difference between active users on an instance and the following/followers of each of those instances. When I was on the $6 single sideqik thread plan, I had 2k followers/following 881 and the 5.2k items in my pull queue had a 4 hour latency.

I think X is more like <150 users, to pull a Dunbar out of my ass...