Date: Wed, 16 Jan 2002 08:51:28 -0600
Reply-To: "net-people: UofMN Net People"
Sender: "net-people: UofMN Net People"
From: Andrew Odlyzko
Subject: Re: Student use a bandwidth
X-To: net-people@umn.edu
Apropos some of the messages from yesterday:
1. Chris Hertel wrote (in response to a posting from Paul Dokas):
> Looking at yesterday's stats for the all I2 routers shows similar
results.
> I'm willing to bet, without looking, that almost all of this traffic
is
> dorm <-> dorm traffic from/to I2 connected Universities.
Consider what that means. Not now, but in the next five to ten years
as
these people graduate, get jobs, buy homes, have puppies, etc. They
expect to get their entertainment, do their communicatng, build their
worlds and societies all via their computers.
Very, very interesting futures we've dug ourselves. :)
Yes, indeed. Isn't that what happened in other areas as well, though?
In computers, all those students that learned how to program in school
back in the 1960s and 1970s graduated, got jobs, bought homes, had
puppies, and bought PCs, and got their employers to buy computers as
well. :-)
2. On a more serious note, Dave Farmer is right, that the rough estimate
for $42/month at large state research universities includes all sorts of
computing and voice telephony, so that central networking might come to
half of that. (A figure I have heard quoted for UC Berkeley is that they
spend $8 million per year for the campus network, including Internet
access.
Since UCB has around 30,000 students, it probably has 40-50 thousand
people
all told, this comes to something like $20 per month per person. This
provides a rough sanity check.)
It is interesting to compare that figure with estimates for total cost
of ownership. Steve Winckelman cited $1700/machine/year for his (and
my) department. But I assume this counts just the explicit financial
costs. Various industry consultants (IDC, Forrester, etc.) have come
out with estimates for commercial establishments on the order of
$10,000 per machine per year. (This counts the time the owner of the
machine spends waiting for the machine to reboot after a crash, the
time spent looking for help, the time helping coworkers figure out
how to get that #%&! Word file to print, etc.) In comparison to that,
network connectivity is inexpensive still. (That is a key part of my
general argument against most QoS measures, available at
. They would reduce
the small networking expense, but would inflate the much larger expense
that is already born by the users.)
The general trick for us, and also for the networking industry at large,
is how to move ahead on the technology curve without going bankrupt.
There are various measures that can be imposed, many of which have been
enumerated here. Some others that have been tried at some schools
include
limiting bandwidth to the dorms to some fraction of total, or
rate-limiting
outgoing flows. It might be a good idea to undertake a survey of what
other schools have done, and how well it has worked.
Andrew Odlyzko
-----
Date: Wed, 16 Jan 2002 12:42:46 -0600
Reply-To: "net-people: UofMN Net People"
Sender: "net-people: UofMN Net People"
From: Karen A Swanberg
Subject: Re: Student use a bandwidth
Content-Type: TEXT/PLAIN; charset=US-ASCII
on 01/15/02, Steve Siirila wisely declared:
> I guess I'd prefer to see a monthly I/O
> allotment on a per-user basis. The action taken on those going over
> might be to rate-limit that port for the remainder of that month, or
> charge the user for the extra usage. I know at least one ISP which
does
> exactly this (they charge for the extra usage), which probably allows
them
> to keep their base rates lower, while recovering the cost incurred by
the
> heavy hitters.
Do you see this applying solely to resnet, or to the entire U.?
What do you do for cracked machines? Deal with the over-charges on a
case-by-case basis?
And for either ResNet or colleges/departments, how would wireless fit
in ?
I'm not against this scheme, just seeing what the consequences would be.
I actually don't think the current scheme for colleges/departments is all
that bad for E-Jack. But I get angry earfulls regularly from my users at
*all* levels about the recurring expense (and the abiltiy/inability to
charge it to certain grants). The standard response of relating how much
our leases from Onvoy/Qwest (etc.) are and how much the switches cost
doesn't seem to work, most of them think their data jacks should be part
of the maintenance, like the lights. I realize how many problems there
are
with that scheme, and that most of you have heard it before, I'm just
throwing it in as a reminder of the end user attitude.
It is useful at Netpeople to hear just how much it costs to run the
network here, because I can use that information (and do use it, as soon
as I hear it) in conversations such as the above.
- -
Karen Swanberg | Sys Admin | Dept. of Geology and Geophysics
206 Pillsbury Hall | 310 Pillsbury Ave. SE | University of Minnesota
Minneapolis, MN 55455 (612) 624-6541 (612) 625-3819 (f)
* <---- Tribble . <--- Tribble.tgz
-----
Date: Tue, 15 Jan 2002 21:05:01 -0600
Reply-To: thecubic@bigfoot.com
Sender: "net-people: UofMN Net People"
From: David Carlson
Subject: Re: Student use a bandwidth
X-To: net-people@nts.umn.edu
In-Reply-To:
Content-Type: text/plain; charset="us-ascii"
Personally I think this is the best idea of the aforementioned (and not
just
because Steve's my boss...), and I could see it being better received in
the
dorms than other scenarios (I happen to live in the dorms).
Perhaps some way of having self-monitoring (like is possible with the
modem
pool) of your traffic, if you go over that amount then you will be
charged.
I used the dial-up connection to the U for years, and a lot of times I
went
over 50 hours, and paid it, because the connectivity was more important.
Who knows, potentially it could end up paying for itself.
While this is not directly predatory to the 'top talkers', it would be an
implementation that would make more people monitor their own bandwidth
usage, which is something that I think is lacking.
Now the shortfall seems to be, and correct me if I am wrong, is that our
equipment can't handle the monitoring? I see that as only a temporary
obstacle.
Oh yeah, as an aside, the current dorm setup is contrary to inter-dorm
file
sharing (SMB). AFAIK there is no WINS server (which serves all dorms),
the
result being that I can only see less than ten computers from my subnet,
four or five of which are my neighbors and friends which have the same
stuff
I have anyway. I can see where some people would then turn to kazaa or
any
other internet service for their content, which is what we're trying to
prevent. Similar to the idea of the central fileserver, but without the
liability.
-dave
-----Original Message-----
From: Steve Winckelman [mailto:steve@math.umn.edu]
Sent: Tuesday, January 15, 2002 4:09 PM
To: Fred Dulles
Cc: farmer@nts.umn.edu; Kevin KleinOsowski; net-people@nts.umn.edu
Subject: Re: Student use a bandwidth
One of my ISPs does exactly this. I contract for n Gbytes of data
movement per month. If I go over I get a higher bill (plus the option to
increase my contract).
Another thought, we had at one point considered a separate VLAN for each
portable computer the faculty and their RAs were connecting to our
networks. This would give them network isolation as well as accounting
info. Probably doesn't scale well. In the end we didn't use this plan
either.
Steve Winckelman
On Tue, 15 Jan 2002, Fred Dulles wrote:
>
>
> I kind of like the U modem charging model, though I am not sure of
how
> practical it would be to implement. The idea that the 90 or 95% who
use
> under whatever is analogous to the first 50 hour limit go for free and
those
> who want to go nuts can, for a fee.
>
> Now I had heard that metering usage by traffic volume was not
doable
> though with snmp switches (and ignoring mobile usage which currently
seems
> relevant for the dorms at least) I am not sure why. Is there a one
> paragraph explanation for this. Also if per port usage metering is too
> hard, what about per subnet? (Which would be useless for students but
> useful for departments?)
>
>
> -Fred Dulles
> ---
> Fred Dulles College of Biological
> Sciences dulles@umn.edu
> Coordinator of Information Technology
>
>
> David Farmer wrote:
>
> > I'd like to hear more.
> >
> > I agree that it is part of the University mission to encourage
> > students to explore and use the Internet. On the other side I don't
> > think the people that wrote that mission statement envisioned the
> > University spending hundreds of thousands of dollars a year for
> > what seems to be students downloading music and movies.
> >
> > I've see two basic ways of limiting usage, one way is to limit each
> > student (IP address) to some amount of access, and let them
> > decide how to use their bandwidth. The other way is for the
> > University to decide what is good use and bad use and enforce
> > that, this means blocking or rate-limiting protocols.
> >
> > Another solution is for students to pay more for access, but this
> > has limits as well. Is it fair for all students to pay for the %10
or
so
> > that are causing problems? Don't quote me on that stat, it's just a
> > guess.
> >
> > The goal is to bring the budget under control but still allow growth,
> > we just need to manage it down to something reasonable. Even
> > with regular price reductions we would need anual budget
> > increases in the %50-%75 range to keep up with unrestrained
> > growth. Given the fact that the University is looking at budget
cuts,
> > those kind of increase don't go over well.
> >
> > Do you have other ideas.
> >
> > On 14 Jan 02, Kevin KleinOsowski wrote:
> >
> > > OK, you said you wanted to hear all opinions on this so how about
the
> > > Universities mission statment:
> > >
> > > http://www1.umn.edu/systemwide/factsmission.html
> > >
> > > "... The University's mission, carried out on multiple campuses and
> > > throught the sate, is threefold:
> > >
> > > Research and Discovery
> > > Teaching and Learning
> > > Outread and Public Service"
> > >
> > >
> > > You can read the rest of the mission statement, but I know I dont
read
> > > anything that states "limit students bandwith, they are using too
> > > much" etc. Maybe we are loosing sight. It seems all I am reading
in
> > > these messages is how can we limit students from learning,
> > > discovering, and accessing. The real question is how do we
encourage
> > > correct usage. Putting in limits punishes the bad and the good.
Yes
> > > you realize I am not pointing out a solution, it alludes me too.
> > > However, this is the other side of this debate.
> > >
> > > Kevin
> > > --
> > > Kevin KleinOsowski
> > > Department of Electrical and Computer Engineering
> > >
> > > University of Minnesota - Twin Cities | Phone: (612) 625-4585
> > > Electrical and Computer Engineering | Fax: (612) 625-4583
> > > 200 Union Street SE, Room 4-174 | kevinko@mail.ece.umn.edu
> > > Minneapolis, MN 55455 | Office: EE/CSsci 1-183
> >
> > ==================================================================
> > David Farmer Email: farmer@oit.umn.edu
> > Office of Information Technology farmer@umn.edu
> > University of Minnesota Phone: 612-626-0815
> > 2218 University Ave SE Cell: 612-812-9952
> > Minneapolis, MN 55414-3029 FAX: 612-624-4035
> > ==================================================================
>
>
---
Date: Wed, 16 Jan 2002 22:22:41 -0600
Reply-To: "net-people: UofMN Net People"
Sender: "net-people: UofMN Net People"
From: Ben Kochie
Subject: Re: Student use a bandwidth
In-Reply-To:
Content-Type: TEXT/PLAIN; charset=US-ASCII
haha.. more responses from the math department.
On Tue, 15 Jan 2002, David Carlson wrote:
> Perhaps some way of having self-monitoring (like is possible with the
modem
> pool) of your traffic, if you go over that amount then you will be
charged.
> I used the dial-up connection to the U for years, and a lot of times I
went
> over 50 hours, and paid it, because the connectivity was more
important.
> Who knows, potentially it could end up paying for itself.
*snip*
this is the key thing here.. students, especialy the new ones in the
dorms
may not yet understand that bandwidth costs money, to them it's just a
free thing. and since they are in school to learn, learning cost
accounting on a network may make them better users in the real world..
providing some way for them to monitor their use is probably the best
thing for both admins and users. and if there is a price attached
(political or monitary) to use, it would teach them that "more isn't
better".
-ben
----
Date: Tue, 15 Jan 2002 23:17:31 -0600
Reply-To: thecubic@bigfoot.com
Sender: "net-people: UofMN Net People"
From: David Carlson
Subject: Re: Student use a bandwidth
In-Reply-To:
Content-Type: text/plain; charset="us-ascii"
> -----Original Message-----
> From: Karen A Swanberg
> Do you see this applying solely to resnet, or to the entire U.?
It seems only appropriate for resnet. I'm sure that although departments
are big, they may, only take,
say, 60KB/sec? Maybe not CS/EE/ITLabs...
Plus departmental users aren't going to use the file-sharing programs,
which
is really the issue.
Anyone have any numbers as to the ratio of departmental to dorm internet
usage?
> What do you do for cracked machines? Deal with the over-charges on a
> case-by-case basis?
The charge is automatic, perhaps with an e-mail notification, so people
should suspect shennanigans if they get one and don't warrant it; if the
machines are indeed cracked it's an issue we (and I do mean we) need to
deal
with. It would probably help the cracking situation a bit to have a
little
more knowledge of when it's happening.
If the machine is cracked, and there is some proof that it is, then the
user
can appeal the charge. I don't see this happening much. Then again, I
am
using Outlook :)
> And for either ResNet or colleges/departments, how would wireless fit
> in ?
Colleges/departments are not applicable (in case of math anyway), the
traffic goes through the same path as wired.
ResNet wireless should be free*. No one is going to go wireless to get
around the fees unless they're really, really exorbatant, or stupid.
ADCS's
wireless is free*, right?
-dave