You are not logged in. Login Now
 0-24   25-49   50-74   75-99   100-124   125-146     
 
Author Message
janc
Should Grex get a T1 connection? Mark Unseen   Oct 25 16:13 UTC 1996

This issue was raised at the last board meeting.  It was decided that we
should have some public discussion of it.

Through the generousity of various individuals and organizations, Grex now
has a chance of being able to get a T1 network connection in the near future.
A T1 connection is much-much faster than our current connection.  It's
actually rather more than we'd have any use for just now.

We have:

 - An influential person at a local ISP who says they are willing to donate
   a connection to Grex (I understand this has market value of about $1000
   per month).

 - Two anonymous donors that have pledged to pay a significant part of the
   Ameritech fees for bringing the T1 connection from the ISP to Grex for
   the first year.

We'd still have to do more negotiations with the ISP to close the deal, but
it looks pretty promising.  We would have to find the one-time costs of the
T1 interfaces at the Grex end at least and possibly at the ISP end as well.
We need to pay Ameritech for the actual line.

The Ameritech fees work as follows (quoting from board meeting minutes):

  There is a connection fee of $587.50

    $330/month for a month-to-month contract.
    $311/month for a 12 month contract.
    $276/month for a 36 month contract.
    $250/month for a 60 month contract.

  If you need to shorten your contract, they will permit this by paying them
  back in a lump-sum the difference  over the term of the contract so far.

  They will waive the connection fee for 36 or 60 month contracts if ordered
  before 12/31/1996.

For the first year, with the aid of our donors, we could probably do any of
these.  But our donors have pledged only a year, and things could happen that
would make it impossible for them to fulfill that either.  To sustain this
in the long term, we would NEED to find more significantly more income from
more sources.  If we take a 60 month contract, and then try to back out after
two years, then we'd be hit with a bill for $1464 for the difference between
two 12-month contracts and the rate we paid at, plus possibly the $587.50
installation fee we avoided by taking the 60 month contract instead of the
12-month contract.  In other words, we'd get socked really hard.  If we get
into this, we NEED to believe we can sustain it.  Grex would be under
significant pressure to increase its income.

To actually make use of all this bandwidth, we'd need to make some further
upgrades to the Sun so that it can run fast enough.  Upgrading the Sun to a
faster CPU is going to cost us $1000 to $1200, and it's not clear that we
could both this and the T1 interface.  We have a good sum of money in the
bank now, but if our monthly costs increase by $250 a month, then we are
probably going to want to keep a larger cushion in the bank.  (Basically,
we want to always hold two or three months operating costs in the bank.)

So if we got such a connection, we might initially not be able to use it all.
We might not increase the number of telnet connections much, but the ones we
have would be faster.  Backtalk would be faster.  Local people who dial in
and want to use Grex as a cheap ISP would find Grex's faster connection much
more attractive, so our phone lines would be busy much more often, and we
would be under pressure to find money to add more.  We would have to
reconsider restrictions on GIFs on web pages and such, since the existing
rules were based on our limited bandwidth.

This would also be the first time Grex was receiving significant external
support.  In that past, we have been pretty self-contained, getting most of
our money from the people we served.  This is in some ways uncomfortable.

I've kind of emphasized the negatives here because they are a bit easier
to overlook. This would definately be a step up to the big leagues for Grex,
if we can do it.
146 responses total.
steve
response 1 of 146: Mark Unseen   Oct 25 16:40 UTC 1996

   We most certainly can use the extra speed of a T1.  The CPU
as we currently have things set up, is just about at maximum capacity.
But we're waiting on I/O from the net so much that we're having
things stay in memory far longer than they should, which slows
things down.  Here's an example: when the large listservers disgorge
their loads, Grex often gets 20 sendmail SMTP connects running at
once.  Each of these is waiting for data to come over the link;
data that comes very slowly because of the connection we have.
It often takes minutes for a small piece of mail to come through.
Let's compare this current situation to one with a T1, and the
same 20 SMTP connects.  Grex would still be up to its little
eyeballs in processing the connections, *but the data would
arrive much faster*.  And, that data coming in faster means
that copies of sendmail can get out of the way that much faster.
Thats a win.  Just how much of a win that will be I can't say.

   Grex has also has periods of time when significant sums of money
was given to it too--the first 1.5 years of Grex's existence was
largely paid for by a group of people who wanted to see the system fly.
I see the folks who are willing to give money for a T1 as doing
exactly the same thing.
dang
response 2 of 146: Mark Unseen   Oct 25 18:01 UTC 1996

What I would be concerned about, and what I think Jan was pointing to with
his significant outside help comment, wass the $1000+ per month being donated
by the ISP.  If they ever back out, Grex gets hit with the contract costs,
because we can't come close to affording that $1000+ per month.
albaugh
response 3 of 146: Mark Unseen   Oct 25 18:15 UTC 1996

I know it's "old" technology these days, but what about a 56KB leased line?

I dunno, given the *on-going* financial commitments getting this T1 line would
require, maybe a safer approach would be to have a separate mail-server
computer, and it & grex could each have a separate line with a very fast modem
steve
response 4 of 146: Mark Unseen   Oct 25 18:42 UTC 1996

   Thinkof the circuit contract this way: a T1 line is $333.00 per
month, onthe month-to-month contract.  The 36 or 60 month lease is
available at the cheaper rate, but if we bow out of it, we'll have 
to make up the difference between what we've paid, and $333 per
month.

    A 56K line would be probably around the area of $150/mo.  Which
means that its 50% (roughly) of the month-to-month price of a T1,
while providing about 1/30th the speed.

   However, the ISP we're talking to does not want to do that.
They think in terms of T1's, and thats what they want to do.
robh
response 5 of 146: Mark Unseen   Oct 25 19:11 UTC 1996

So for those of us not up on bandwidth terms, a T1 would be about
sixty times as fast as our current connection?
ryan1
response 6 of 146: Mark Unseen   Oct 25 19:32 UTC 1996

I think this would be great for Grex.  If Grex can afford this (or somebody
else who is willing to pay, or donate, like thie ISP in question) I see
no reason to object.  
steve
response 7 of 146: Mark Unseen   Oct 25 22:00 UTC 1996

  53.611 times faster, but who's counting?
srw
response 8 of 146: Mark Unseen   Oct 26 00:08 UTC 1996

The ISP in question is looking for a way to support a non-profit community
network with an in-kind contribution. The $1000 retail value of this service
appears not to be a concern. Certainly the possibility exists that they could
change their mind. How could we possibly know what the future would bring?
There is no evidence that they are insincere in offering this net-bandwidth
to us (providing we pay for the connection) on a long-term basis. 

An intermediate upgrade really doesn't make any sense so the other option that
remains is to keep the current configuration. I really could not recommend
keeping our current netlink. It is hopelessly inadequate and makes the system
very hard to use, as well as negatively impacting performance (mail, for 
example). 

Even if we decided we could not afford to upgrade the CPU too, I would favor
the T1 link. Ultimately, I think we'll be able to do both. I doubt that 
the user base will increase much in response to this, but I do think that 
more of the suffering telnetters will become members. As it will be much less
painful to use the system. People will also be able to get on and off faster.

In short, I think this is a long overdue improvement that is sorely needed.
scg
response 9 of 146: Mark Unseen   Oct 26 00:32 UTC 1996

(Actually, I think the retail value on the ISP's end of the T1 is around
$2,000, not $1,000.  If we can get it for free, which looks likely, it will
be an incredible deal)
mdw
response 10 of 146: Mark Unseen   Oct 31 05:44 UTC 1996

It should also be noted that it's a very *special* deal.  In this case,
grex looks like a very good match for the ISP.  I am *very* definitely
in favour of this.  I think it will allow us to provide public access to
a far larger group of people than we might otherwise be able to serve.
Hopefully in that far larger group of people, we will also be able to
find more people who can afford to become members (perhaps having CPU
lag instead of network lag will be more incentive to improve the
system), but we will certainly need to keep up efforts to get more
members, a la PBS.
tsty
response 11 of 146: Mark Unseen   Oct 31 10:52 UTC 1996

donations make grex work. while we will lose the adbantage of lag, which
keeps the loads down, we will gain better access and help all those
ppl who load the system with  .forward files
  
for a while, we will see faster off-system activity. then more ppl will
login becaue

login because the donation will support the added load.  then the
load will increase to meet the demand, and, most likely, the entire
system will slow down.
  
and that will accelerate the absorbtion of available uid numbers so 
whatever was going to happen in 3-4 years will happen in fewer years.
(i don't remember the specific approximate time before we run out
of uids.)
  
i would suggest that if we run into a  cpu  lag problem we are going
to really tick off the system supporters even more.
  
however, for the short run, let;s see what hits the wall first, the
cpu or the supporters
  
if the net lag hits the wall first.... dump it.
  
one way to keep the situation slightly under control would be to 
agree NOT to increase the NUMBER of pseudo tty connections after we
upscale to T1. 
mdw
response 12 of 146: Mark Unseen   Oct 31 11:11 UTC 1996

And turn grex into an ISP?
ajax
response 13 of 146: Mark Unseen   Oct 31 13:58 UTC 1996

  Steve said the ISP probably wouldn't want to do a 56K leased line,
because they think in terms of T1s.  Do you think they also wouldn't
donate service for a 115K ISDN line?  I'd like to know whether or not
that's an option before a decision is made on getting a T1.
 
  I also have a question about the two benefactors who have offered
to shoulder much of the cost of the T1 line for one year.  If their
donation would be spread out over the course of the year, I wonder
if they could be asked to sign an agreement of some sort.  I don't
doubt their generosity at all, but given the consequences of a
pullout, I think they'd understand why that would be desired.
Ideally, it would be nice to obtain the donations prior to signing
a contract with Ameritech, contingent on Grex's acquiring a T1 line
for one year.
steve
response 14 of 146: Mark Unseen   Oct 31 15:37 UTC 1996

   I'm not in favor of asking for an agreement.  Grex has never
done that before, and we've never had problems along those lines.
Sure, we've never dealt with this kind of money before, but that
doesn't bother me--its in the same spirit as things before.

   Getting an ISDN line isn't likely either.  Again, they think
in terms of T1's.
slynne
response 15 of 146: Mark Unseen   Oct 31 16:37 UTC 1996

I generally like the idea of getting a T1. I think grex is stable enough
right now to support that kind of financial commitment. However, things
can change rather quickly so if grex were to do something like this, I
would suggest that people look long and hard at the problems Mnet
currently faces. 

scg
response 16 of 146: Mark Unseen   Oct 31 17:49 UTC 1996

Actually, they've offered 56K or T1.  ISDN is the only thing they've said they
don't do.  56K priving and T1 pricing aren't really different enough, as far
as the Ameritech cost goes, to make it an attractive alternative to a T1. 
Our current load would saturate a 56K line; it would just be somewhat less
saturated.
void
response 17 of 146: Mark Unseen   Oct 31 18:53 UTC 1996

   slynne, a major difference here is that grex is looking at a donated
service, rather than going out and purchasing one about which there is
considerable doubt.
chelsea
response 18 of 146: Mark Unseen   Oct 31 22:05 UTC 1996

I hope no decision is made to proceed until we are very sure we can pay
the bills.  This means: 

One: we are very sure those who are generously offering to contribute are
a good risk.  Because, it is a risk when Grex signs a contract obligating
us to new expenses.  We have had folks give gifts before but never, that I
recall, have we been vulnerable to a debt load if the gift was withdrawn.
The names of the donors should be disclosed.  The only reason I don't
think a contract should be signed is that, in the event the donors
backed out for whatever reason, I can't see us trying to enforce
any such document.  So why bother.

Two: Can we afford this upgrade or are we hoping that we'll find support
once we've spent the money?  I really, really, really don't want to see
Grex get into the position where we expand without a realistic financial
plan and end up begging for funding we thought would be there.  You want
to turn off potential members one of the best techniques going is look
desperate, fragile, and broke. 
ajax
response 19 of 146: Mark Unseen   Nov 1 08:50 UTC 1996

  Written agreements, even if not legally binding, can help avoid
misunderstandings or resolve disputes.  Six months down the road,
if the donors say "we donated based on Grex retaining its 1996
access policy," and Grex says "you never said that," it would be
nice to have something to refer to.  It could also clarify when
checks are expected, when they will cease, and so on.  Of course,
all these concerns are moot if the donors can provide an up-front
donation.  That would remove a lot of risk to Grex.  I also feel
kind of like if a person isn't fairly easily able to make it a
lump-sum donation, then despite their desire to continue donating,
their circumstances may change to make it impossible (for example,
if they lose their job).
adbarr
response 20 of 146: Mark Unseen   Nov 1 11:57 UTC 1996

Grex does not need the contract, only the benefits. The donors could make the
contract to pay for the access with the benefits going to Grex as a
third-party beneficiary of the deal. Grex signs nothing. Donors get the
benefit of helping Grex, Grex gets the benefits of the help, and the provider
gets the $$. Has been done lots of times in many contexts.
ajax
response 21 of 146: Mark Unseen   Nov 1 18:19 UTC 1996

  There's not a single person who would be paying the phone bill; as
proposed, Grex and two individuals would share the expense.
janc
response 22 of 146: Mark Unseen   Nov 1 19:05 UTC 1996

I would like to see Grex raise a larger portion of the price of this from
pledges from other usess.  I'd much rather have 20 people pledging $10 a
month than two pledging $100.

I also think we should right *now* be talking to the ISP to try to work out
the formal details of the plan.  I think we are pretty sure we *can* do it,
though maybe not 100% sure if we want to do it or how exactly we want to
finance it.  That's a reasonable point to be at to start serious negotiations.
We can't make a final decision until all the details of the agreement are
mapped out anyway.
kaplan
response 23 of 146: Mark Unseen   Nov 1 19:11 UTC 1996

What is the cost of the CPU upgrade that grex would need to support a T1?
Will this new CPU have a significant impact on the electric bill?
dang
response 24 of 146: Mark Unseen   Nov 1 19:33 UTC 1996

Does Grex need a CPU upgrade to support a T1?  I don't think so, as long as
we don't raise the ptys.
The advantage of a T1 is that it will take a long time to saturate it.  No
longer will we have several hardware bottlenecks, with the cpu upgrades not
showing any improvement because the link is saturated.  If we upgrade the CPU,
then we will see the improvement.  (Granted, there's still disk and memory,
as well as modems, but they're not quite as important to perceived slowness)
 0-24   25-49   50-74   75-99   100-124   125-146     
Response Not Possible: You are Not Logged In
 

- Backtalk version 1.3.30 - Copyright 1996-2006, Jan Wolter and Steve Weiss