blue skies server testing ideas
It would be interesting to do a resource comparison on one Mastodon with 10,00 accounts on it vs. 10,000 single-user instances, under the same activity load.
I hear that a Mastodon roams the steppes better in a VM, with each of its body parts in separate containers. I wonder how many single-user containers you could run on one VM? On one physical server machine?
What kind of server infrastructure would you need to test this kind of thing?
Better in a VM than what?
Do you mean better single-user instances than multiuser instances, or better in a VM than in separate physical servers?
What advantages do you expect the single user setup to have?
@volkris
I expect that;
* 1 Mastodon instance with 10 users would use less server resources than 10 single-user Mastodon instances.
* 10 instances of software designed for a single-user would use less server resources than either.
I don't think that answers my question as to what advantage there is to single user instances.
I'm not trying to be belligerent or anything, I am honestly asking what you have in mind.
@volkris
> I don’t think that answers my question as to what advantage there is to single user instances
If you don't see using less server resources to support the same level of use as an advantage, I'm not sure what to tell you 🤷♂️
@helge
> bovine_herd should allow you to move off a multi user instance to another as long as you provide your own domain name
Cool, I hadn't heard of this. Takahē was designed to be multi-domain too:
https://jointakahe.org/
@helge
> the problem with server load in the Fediverse is peak load
sure exactly what you mean by "peak load". From what I've read, load goes up on a server as more follow/ followed relationship develop between accounts there and on other servers. Eg:
> My guess is that the server requirements to run bovine are pretty much the same for 1 user and a few hundreds
Would be good to test this too and know for sure.
My point is that it uses MORE resources though!
So my question is that, using more resources, what advantage does it bring? That's exactly the question I'm asking here.
I'm not trying to be argumentative, I'm just trying to figure out what you are saying.
Is it that you did not realize, then, that it uses more resources to have single user servers?
@volkris
> uses more resources though
Why would an instance of a RDBMS-less lightweight like Epicyon (or a not-as-light microblog.pub) have the same resource usage as that of a RDBMS-backed Mastodon?
It is an interesting query @strypey has, on the numeric factors of this ratio.
I'd add that network load might also be different enough to merit attention.
@volkris
> Is it that you did not realize, then, that it uses more resources to have single user servers?
1) Are you talking about single-user Mastodon servers or instances of software designed for single-user servers?
2) Do you have evidence for this, based on a repeatable testing method, or are you just assuming?
> I’m not trying to be argumentative
Neither are you trying to read my posts in the context of the whole thread. Please do or I'm likely to lose interest.
@volkris
> Better in a VM than
... on a shared server.
> Do you mean better single-user instances than multiuser instances
That's what the test would be looking to find out. Not only comparing multi-user to single-user Mastodon under the same load of user activity, but comparing multi-user Mastodon to single-user instances running software designed for this use, like microblog.pub.