Re: apt PARALLELISM
Scripsit Henrique de Moraes Holschuh <hmh@debian.org>
> THis is not something that would bother anyone if it is a single user... but
> if you have 10k users doing that, often close enough in time, well, things
> should get MUCH worse as far as I can see. If they are doing this at random
> times in the day, OTOH, it would not be that bad, I guess.
That's what I mean. People don't synchronize their updates - certainly
I don't synchronize with anybody, and I don't know of any mechanism
that I *could* use to sync with anybody if I wanted to.
Assume a situation where mirror bandwidth is the limiting factor, and
imagine a world with 3 mirrors. Say that during a certain time of the
day 600 users each minute start to download updated x.org packages.
Either they can do their download sequentially, choosing a random
server; then their download will be finished in 15 minutes, and each
server has a more-or-less constant 600/3*15 = 3000 connections
active. Alternatively each user can spread his load over all three
servers; his download now takes 5 minutes, and each server _still_
sees 600*5 = 3000 active connections at any time. Thus _all_ users get
it faster by parallelizing. We get the same result if only some users
parallelize - the mirrors do not see a diffence in load, the smart
users get things faster, and the sequentially downloading users get it
no slower than they would have otherwise.
The calculation becomes more murky if there is backbone congestion
which hits more than one mirror _and_ more than one end user. Then he
who opens more connections at a time (whether to one server or
several) will probably get an advantage at other users' expense.
But I don't think that backbone congestion is such a universal
condition that it should necessarily be the only scenario for making
moral decisions about what apt should be _able_ to do.
> Whether MY [a single individual] increased download speed is worth the extra
> load on the mirror network, and whether it WOULD increase the load on the
> mirror network is what we are asking here.
Hm, you are not even asking whether the mirror load would go up? What
_are_ you asking, then?
> (and for the people who can't read whole threads, my position is that we
> should never decrease the experience of a group of people to increase the
> experience of an individual).
I am questioning your assumption that doing parallel downloads will
necessarily decrease the experience of a group of people at all.
--
Henning Makholm "Vi skal nok ikke begynde at undervise hinanden i
den store regnekunst her, men jeg vil foreslå, at vi fra
Kulturministeriets side sørger for at fremsende tallene og også
give en beskrivelse af, hvordan man læser tallene. Tak for i dag!"
Reply to: