Skip site navigation (1)Skip section navigation (2)
Date:      Mon, 31 Jul 2006 21:13:39 -0500
From:      "Nikolas Britton" <nikolas.britton@gmail.com>
To:        "User Freebsd" <freebsd@hub.org>
Cc:        freebsd-questions@freebsd.org, Ted Mittelstaedt <tedm@toybox.placo.com>
Subject:   Re: Gotta start somewhere ... how many of us are really out there?
Message-ID:  <ef10de9a0607311913h6bc62e85ob5103ad7296e3a05@mail.gmail.com>
In-Reply-To: <ef10de9a0607291023v10008ab1yd2ad0b399311b8d9@mail.gmail.com>
References:  <20060728164526.E27679@ganymede.hub.org> <df9ac37c0607281319s5da0f64ese5fa57df1ef11a4d@mail.gmail.com> <ef10de9a0607282139i51fdde5ch58525fa3347364d2@mail.gmail.com> <003f01c6b2cb$c42d5bd0$3c01a8c0@coolf89ea26645> <20060729021705.F27679@ganymede.hub.org> <ef10de9a0607291023v10008ab1yd2ad0b399311b8d9@mail.gmail.com>

next in thread | previous in thread | raw e-mail | index | archive | help
My calculations are off, I though the monthly periodic was relative to
the system install date. Here are the new numbers:

Lets say each client sends 20 bytes and their are 10^7 clients for a
total of 190.7MB per month. Now... Lets say 50% (10^6.7) of those
clients are set to UTC and all of them trigger on the first of the
month within 5 minutes of each other (10^6 per minute). This equates
to 16706 clients per second. We would need 326KB/s or 2610Kbit/s to
handle this load. This is a problem, even half of that is a problem.


On 7/29/06, Nikolas Britton <nikolas.britton@gmail.com> wrote:
> On 7/29/06, User Freebsd <freebsd@hub.org> wrote:
> > On Fri, 28 Jul 2006, Ted Mittelstaedt wrote:
> >
> > > You might think this sounds harmless but folks have done this kind of
> > > thing in the past with other products and wreaked havoc on the Internet.
> > > You can start by referencing "dlink ntp fiasco" in google to get an idea
> > > of what can happen to these kinds of well meaning attempts.  Let
> > > sleeping dogs lie.
> >
> > 'k, you lost me on how this relates to the fiasco ... I did a quick search
> > on Google for it, and, unless I didn't find the right reference, the
> > 'fiasco' had to do with DLink setting up their software to ping PHKs NTP
> > Server, without getting permissions first, and, thereby, flooding him with
> > NTP requests ...
> >
> > > People just don't realize just how very big the Internet is.
> >
> > That is the problem, yes ... nobody knows how big the FreeBSD community is
> > ... :)
> >
>
> I have to agree with Marc on this one. The extra load required to send
> all of this data is not much:
>
> Lets say each client sends 20 bytes and their are 10^7 clients for a
> total of 190.7MB per month or 6.25MB per day . Now...
> Lets say 50% (10^6.7) of those clients are set to UTC and 50% of those
> clients (10^6.4) trigger the monthly periodic over a 5 day period
> (10^5.7 each day) and all of them phone home within 5 minutes of each
> other (10^5 per minute) for a total of 1666.67 clients per second. We
> would need 32.6KB/s or 260.4Kbit/s to handle this load spike... I did
> the calculations for 10 million clients, but I highly doubt FreeBSD
> has 5 million so this is a non issue.
>


-- 
BSD Podcasts @:
http://bsdtalk.blogspot.com/
http://freebsdforall.blogspot.com/



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?ef10de9a0607311913h6bc62e85ob5103ad7296e3a05>