Re: Report from DSA Team Sprint in Oslo
]] Filipus Klutiero
> Tollef Fog Heen wrote:
> > ]] Filipus Klutiero
> > > It is concerning that we no longer receive enough hardware donations
> > > to meet our needs. I can only encourage these improvements to
> > > donations pages. Specifically, if we lack any significant hardware,
> > > let's add it tohttp://www.debian.org/misc/hardware_wanted.en.html
> > I don't think it's particularly concerning. Relying on donated hardware
> > rather than buying hardware ourself ties our hands with regards to what
> > kind of hardware we end up with. By buying hardware, we get the
> > hardware we want, at the time we want.
> I agree that imposing ourselves to rely exclusively on donations would
> be a constraint. Clearly, buying hardware ourselves gives us the
> optimal result in a sense, but it's also most costly, and money is one
> of our constraints.
In my experience, when asking big companies for money for
freedesktop.org, it was not particularly hard to get fairly big
donations with little effort. I think there are two important points to
note in that regard though: We had a concrete goal (both in terms of «we
need X» and «we are going to do Y with the money») and we solicited the
donations, we didn't rely on somebody coming up with the idea that maybe
they would want to give money to freedesktop.org.
> I was not really thinking about stopping to buy hardware, but rather
> to increase the hardware we're donated (and if we get enough, perhaps
> reduce our hardware acquisitions). I imagine that lots of our hardware
> needs are not too specific and can be accomplished by a variety of
> machines. I would think donated hardware could be used to address
> general needs, and the needs which donations failed to cover could be
If somebody donates the hardware that exactly matches our specs, I don't
think we're going to complain, but the servers are both fairly costly
and we often have requirements such as «HP DL460c» (a HP blade server,
since we have a HP blade chassis) where getting a donation of some IBM,
Sun or Dell blade won't work, as we don't have the chassis and even if
we're donated a chassis, getting people to host them takes a bit of
work, since they're big. That example is one of the more generic ones,
often we'll want it to have X amount of memory, Y CPUs of some minimum
number of cores and speed and Z GB of disk space, across some number of
disks or SSDs. Getting donations like that means lots of interaction
with a donor and doesn't often happen.
Also, the kind of «hardware donations» we already get is good pricing on
the equipment we buy, which isn't really a donation as such, but still
means we spend less on hardware.
> Hardware companies from which we accept donations should find it
> interesting to be entitled to claim that Debian uses their machines.
> So I think we should first make it clear that the hardware we need
> most is servers. (And we need bandwidth too)
Hosting and bandwidth is donated today already, and it seems like our
bandwidth requirements are very low, so it's really just about rack
space, power and cooling.
> > > I am curious about what share of our hardware needs come from the need
> > > to support specific architectures. Perhaps even how much money and
> > > time DSA spends on each architecture.
> > We so far only have numbers for our services, where the cost of an extra
> > architecture is fairly small.
> Sorry, what do you mean by "our services"?
ftp-master.debian.org, packages.debian.org, lists.debian.org, etc.
Basically everything but buildds and porter boxes.
> > As for time spent, we don't track that at
> > all.
> Yeah :-|
Why the :-|? I see no point in asking people to track how much time
they spend on DSA tasks. If somebody wants to, they're of course free
to, but this is something we do for fun, not as a job.
> So, I don't know. I'm not much aware of our budget, but it seems
> Wikimedia's income in money donations (around 30 million USD / year)
> can't be compared to ours, and that probably makes its hardware
> procurement a lot easier their Debian's. A less different organization
> might be more interesting for comparison. For reference:
Their budget is multiple magnitudes bigger than our HW budget.
Tollef Fog Heen
UNIX is user friendly, it's just picky about who its friends are