Re: Re: [Micronet] effect of music sharing on shared 10Mb network

Well, it's definitely not helping. As a practical recourse, I would
simply email folks using the net and ask them to turn music sharing off
while on that network, due to congestion issues, and also see about a
network upgrade on general principles.

10Mb shared with a full /24 subnet just isn't a viable service model
these days, with the gains in workstation capability, and file sizes
ever-increasing.

FYI, anything over 30% utilization sustained is considered over-utilized
for shared-10Mb. It's a rule of thumb obviously, but there is
official-ish documentation to that effect. I've seen usability decline
noticeably after 40%.

And before you ask, no, we really don't have much we can do about it
(other than an upgrade) on the technical side. The gear is too old and
too dumb to attempt blocking data, and we don't do that without serious
provocation anyway. This probably doesn't qualify as serious, unless
it's demonstrably killing the network usability on a regular basis.

So I would attempt a social engineering approach, were I you, to this
problem. Once educated, most users are reasonable, especially when it's
shown to effect other folks they know and/or have to deal with. Um, I
did say "most"...

-Jay

gleno wrote:
> In this particular case, it's a shared 10 Mb network with over 200
> assigned IPs which on occasion becomes virtually unusable during
> periods of high utilization (85-90% and beyond). I don't know if that
> falls outside the technical definition of "properly-functioning
> network," but it purely describes "unhappy network campers." ;-)
>
> The most recent occurrence was not due (at least not explicitly) to
> music streaming, but to a large file transfer during the busy
> mid-morning hours. Once the user was located and asked to reschedule
> their large (legitimate) file transfers for off-hours, things returned
> to normal.
>
> So, if the constant network traffic is increased by some (unknown)
> multiple of 128 Kbps, the potential effect is rather significant (for
> this network), is it not?
>
>
> At 1:20 PM -0800 2/24/06, Mike Hunter wrote:
>> I was off by a factor of 10 in my own head (although I wrote down the
>> correct figure), so I should revise what I wrote and say that a person
>> streaming audio at 128Kbps is not going to cause a problem on a
>> properly-functioning network. If they are trying to share the network
>> with 200 of their closest friends and colleagues who are also trying to
>> use a crowded network, then they're using more than their fair
>> share. But
>> that's not for CNS to adjudicate.
>>
>> On Feb 24 at 13:04, "John D. MacDonald" wrote:
>>
>>> Well, thanks to the delay in Micronet (and/or my mail client checking
>>> once every 10 minutes) I did actually write my post before I got Mr.
>>> Hunter's, so my apologies for any toe stepping or repetition. But
>>> even
>>> talking about 5-6 Mbps throughput, each stream is still less than
>>> 1/40th
>>> of the available bandwidth...I'm not saying this isn't or can't be an
>>> issue/problem, especially if Glen is talking about a shared (hub)
>>> situation. Just trying to give some figures for perspective...and
>>> yes,
>>> YMMV indeed.
>>>
>>> John
>>>
>>>
>>>
>>> Jay Bryon wrote:
>>> >Just one correction, it's not a 10Mbps pipe, in practice it's more
>>> >like 5-6Mbps.
>>> >So in deference to Mr. Hunter, who actually is the authority on
>>> >traffic flow around here, it's a problem. How great of one?
>>> >Well, what else do you want to use the network for?
>>> >YMMV.
>>> >
>>> >-J
>>> >
>>> >John D. MacDonald wrote:
>>> >>I certainly won't speak of any larger policy issues about mp3
>>> storage
>>> >>on campus owned machines, but to the original question...
>>> >>
>>> >>some simple math...128kbps streams with say 10% packet overhead
>>> gives
>>> >>about 140kbps, which goes about 73 times into a 10mb pipe. Now, the
>>> >>scale and scope of the effect will depend on both the number of
>>> users
>>> >>streaming, and the network infrastructure itself. If you've got 48
>>> >>people on a 10mb hub (single collision domain), yeah, you've got
>>> >>problems, but the same 48 users on 48 10mb switch ports...almost no
>>> >>problem.
>>> >>
>>> >>And as for the legality or policy issues (RIAA and such) of this
>>> >>sharing, there is no issue. Apple negotiated with the recording
>>> >>companies and RIAA to ensure that the "sharing" feature of iTunes
>>> was
>>> >>'kosher' with all. If anyone remembers, the first version allowed
>>> >>you to share (stream, not distribute files) to your closest 6.49
>>> >>billion friends over the internet. The record companies exerted a
>>> >>little pressure, and the iTunes sharing was changed to allow only
>>> >>local network sharing. So, you can be pretty sure that the RIAA
>>> >>won't be knocking down your door for using iTunes sharing,
>>> whether at
>>> >>home, or on campus.
>>> >>
>>> >>That said, the individual user that downloads 10,000 mp3s from
>>> kazaa,
>>> >>emule, whatever, onto a campus machine and shares them exclusively
>>> >>through iTunes isn't immune to RIAA (or UC) smackdown, since the
>>> >>files are still illegal.
>>
>> [snip]
>>
>> ------------------------------------------------------------------------
>> The following was automatically added to this message by the list
>> server:
>>
>> For information about MAGNet, its meetings and events, and its
>> mailing list, including information on subscribing and unsubscribing,
>> see the MAGNet Web site at <http://magnet.berkeley.edu/>.
>
>

-- 
-Jay Bryon
Senior Network Engineer, CNS
U.C. Berkeley
jay@berkeley.edu
2-5636
------------------------------------------------------------------------
The following was automatically added to this message by the list server:
For information about MAGNet, its meetings and events, and its
mailing list, including information on subscribing and unsubscribing,
see the MAGNet Web site at <http://magnet.berkeley.edu/>.