rpw3@redwood.UUCP (Rob Warnock) (07/16/85)
+------[from net.lan <740@utcs.UUCP>]--------- | "Never underestimate the bandwidth of a | station wagon full of tapes." | - some wise words from | Dr. Warren Jackson, Director, UTCS +--------------- Reminds me of a calculation done a few years ago, on the cost-effectiveness of 56kb lines for heavy batch transfer cross-country. My numbers may be off a bit, but you'll get the idea: A. One 56kb line (clean, low errors) with fairly large block sizes (say, 90% efficiency) can move about 550 Mbytes per 24 hours, and costs about $1.50/mi/mo (???) or (for 2000 miles) about $3000/mo plus about $1500/end/mo for modems or about $6000/mo or $200/day, which is about $0.37 per megabyte. Average data rate is about 50 kbit/sec; latency for a 1 Mbyte transfer is a little over 2.5 minutes. B. Using 2400 foot magtapes with large blocks (say, 90% efficiency) at 6250 bpi, you can fit about 180 Mbyte/magtape. Three (3) of those is about 540 Mbyte [same as 56kb line per day]. You can send a parcel of three tapes by Federal Express (or equiv.) every day at less than $50/parcel, or $50/day, or $0.09/Mbyte. Average data rate is 50 kbit/sec; latency (1 Mbyte) is 24 hours. [Open question: What is the shape of the weight/cost curve? How many pounds or cubic feet can you ship in one 24-hour parcel? I.e., is there a better ($$$/Mbyte) size than 3 tapes/parcel?] Adding the case of the station wagon [above], we get: C. A medium-small station wagon (say, like my VW dasher) can hold at least 25 cartons of a dozen tapes, or 300 tapes, or 162 Gbyte. With a cross-country travel time of 40-48 hours (55 mph, 2 drivers, minimal stops), running three shifts of station wagons might cost (say) $1600/48hrs (6 salaries @ $30k + $1.50/gal @ 27mpg + depreciation @ $0.20/mi) or about $0.01/Mbyte. Average data rate is 15 Mbit/sec, latency (1 Mbyte) is 48 hours. D. Use cargo vans, 10 times as many tapes, double the expense, and get roughly $0.002/Mbyte. Average data rate is 150 Mbit/sec, latency (1 Mbyte) is 48 hours. So the latency/bandwidth tradeoff holds "in the large", as well. I suspect, however, that using larger packages in a "Federal Express" mode (say, using some of the "Second Day" delivery services) could be as cost-effective as the station wagon, with the same latency. Anybody got any useful (i.e., current) cost numbers on a 2000 mile T-1 circuit, for comparison? It would be interesting to compare both land lines and satellite circuits vis-a-vis the station wagon or parcel delivery. Lest this be considered totally a joke, though it is humorous, has anybody thought about sending USENET traffic via magtape and FedEx? How does it compare to your phone bill? What is the effect on latency? What is the effect on Stargate? ;-} I mean, with traffic still under a meg a day, and with special delivery letters at $2-3 each (depending on weight), we could USnail FLOPPIES and still win! And you don't have to worry about your disks filling up while you're on vacation! Your mailbox will hold WEEKS of floppies! ;-} And you don't have to worry about rnews tying up uuxqt or (alternatively) flooding your system with inews'es... Hmmm... maybe that's how we handle the flood of PCs that are about to hit the net... offer them a cost-effective netnews floppy replication and distribution service! You receive in the mail; you send with a modem or (cheaper) just use mail. Make copies for your friends! Or put the floppies inside the back cover of "NET Magazine: The Weekly Digest of the Highlights of USENET (with commentaries by noted net authors, critics, and flamers)"! I noticed in a bookstore the other day that at least one magazine is already being published in floppy format. 'S truth! Rob Warnock Systems Architecture Consultant UUCP: {ihnp4,ucbvax!dual}!fortune!redwood!rpw3 DDD: (415)572-2607 USPS: 510 Trinidad Lane, Foster City, CA 94404
jrrt@mtuxo.UUCP (r.mitchell) (07/17/85)
Rob Warnock asked:
Anybody got any useful (i.e., current) cost numbers on a 2000 mile T-1
circuit, for comparison? It would be interesting to compare both land lines
and satellite circuits vis-a-vis the station wagon or parcel delivery.
****************************************************************
You asked for it, you got it!
I will assume the 2000-miles are between the two AT&T Communications
Serving Office, with another 3 miles on either end to get from the
SOs to the end location. I further assume that the T-1 will
terminate in host computers, rather than a switch, to avoid the
$36.25 per channel per end surcharge. Without going into details (which
are available on request), that 2000-mile T-1 circuit will run you
2*(452.9+3*34.13+21)+1400+26*2000+2*81 = $54,714.58 per month.
T-1s tend to prove in for distances significantly less than 2000 miles.
Who do I send my bill for consultant fees to?
Rob Mitchell
{allegra,ihnp4}!mtuxo!jrrt
Disclaimer: I work for AT&T Information Systems, so AT&T Communications
is a competitor. ;-)henry@utzoo.UUCP (Henry Spencer) (07/18/85)
Netnews via magtape is not a strange notion at all; I believe the
Australians get most of their newsgroups that way.
--
Henry Spencer @ U of Toronto Zoology
{allegra,ihnp4,linus,decvax}!utzoo!henryfrank@aaec.OZ (Frank Crawford) (07/19/85)
> +------[from net.lan <740@utcs.UUCP>]--------- > | "Never underestimate the bandwidth of a > | station wagon full of tapes." > | - some wise words from > | Dr. Warren Jackson, Director, UTCS > +--------------- > > Lest this be considered totally a joke, though it is humorous, has anybody > thought about sending USENET traffic via magtape and FedEx? How does it > compare to your phone bill? What is the effect on latency? What is the > effect on Stargate? ;-} I mean, with traffic still under a meg a day, > and with special delivery letters at $2-3 each (depending on weight), > we could USnail FLOPPIES and still win! And you don't have to worry about > your disks filling up while you're on vacation! Your mailbox will hold WEEKS > of floppies! ;-} And you don't have to worry about rnews tying up uuxqt or > (alternatively) flooding your system with inews'es... > Yes, that's all sounds good, but you still haven't worked out how to stop new being lost. I don't know what the mail system's like in the US but in Aust. no matter how you sent it some would be lost or at least delayed for an extended period of time. Actually from what I have seen that's probably no different to the current situation. Frank Crawford ACSnet: frank@aaec.OZ UUCP: ...{decvax,vax135,eagle,pesnta}!mulga!aaec.OZ!frank (or the ACSnet address above)
bb@wjh12.UUCP (byer) (07/19/85)
> +------[from net.lan <740@utcs.UUCP>]--------- > | "Never underestimate the bandwidth of a > | station wagon full of tapes." > | - some wise words from > | Dr. Warren Jackson, Director, UTCS > +--------------- Rob Warnock then brainstorms a little on other low-tech solutions for multi-Mbyte X-C data transfers. While mention was made of FedEx being one possibility, and they sure do provide good service, it is fairly expensive. Anyone investigating express shipments is surely aware of the other alternatives, including USPS Express Mail. Well, there is another flavor of Express Mail called ``Label C'' which is for Airport-to-Airport service with shipment made on the next flight out, usualy getting point-to-point the same day. The price is $6.15 for two pounds, $10-16 for 10 lbs. and $17-34 for 25 lbs. (variation based on `zones'). This might provide maximum cost-effectiveness in the volume/latency/cost matrix. Brent Byer (harvard!wjh12!bb) ``We are the other people -- you're the other people, too.''
root@bu-cs.UUCP (Barry Shein) (07/20/85)
Re: Fine suggestion to distribute USENET on floppies (or some such.) Re: Re: Problem is Post Office lossage You mention that you are not in the US, the US Mail is not in my experience very bad about reliability (I know, flame flame, but really, about 99.9% gets where it's supposed to w/in 2 or 3 days.) The alternative exists for a dollar or two to special handling which ups reliability considerably (again, favorite horror anecdote to /dev/null, its rare, not rare enough, but rare.) Also, so it got lost once a year, so you get another copy sent, this gets closer to the real problem... WHO EXACTLY IS GOING TO SIT AND MAKE THESE FLOPPIES? The beauty of the current scheme is that while all good children are asleep (or up hacking) the news gets silently delivered, a phone is dialed, a phone is answered, chug chug, done. Yes, there is maintenance involved, but I doubt this will reduce it for the SUPPLIER sites (someone still has to keep it coming in to get it onto floppies.) If someone is willing to copy floppies, god bless 'em, I think tho this is the major stumbling block to the scheme unless a data-duplication service is involved which probably puts the price back to where you started. However, for example, if one site would feed a foreign site who would in turn UUCP to local sites (eg. transatlantic) this might be made to work, but I doubt that is what people are talking about. -Barry Shein, Boston University
john@basser.oz (John Mackin) (07/21/85)
In article <5800@utzoo.UUCP> henry@utzoo.UUCP (Henry Spencer) writes: > Netnews via magtape is not a strange notion at all; I believe the > Australians get most of their newsgroups that way. Ahem. We DID, back in about 1980. Now they come to Melbourne via X.25 between seismo and munnari, the same way this brief note is reaching you. ``We may be down under -- but we're not down and out!'' John Mackin, Basser Department of Computer Science, University of Sydney, Sydney, Australia UUCP: ...!decvax!mulga!basser.oz!john ...!seismo!munnari!basser.oz!john (much faster) ARPA: munnari!basser.oz!john@SEISMO.ARPA (or should that be .CSS.GOV?)
kre@ucbvax.ARPA (Robert Elz) (07/22/85)
In article <5800@utzoo.UUCP>, henry@utzoo.UUCP (Henry Spencer) writes: > Netnews via magtape is not a strange notion at all; I believe the > Australians get most of their newsgroups that way. No they (we) don't. Finding someone willing to mount a magtape, every (day/week/..) forever, then post it, ... isn't an easy task. Robert Elz ps: Henry: I've been waiting for you to say something utterly and demonstrably wrong for ages now ... finally caught one :-)
henry@utzoo.UUCP (Henry Spencer) (07/22/85)
> WHO EXACTLY IS GOING TO SIT AND MAKE THESE FLOPPIES? The beauty of the > current scheme is that while all good children are asleep (or up > hacking) the news gets silently delivered, a phone is dialed, a phone is > answered, chug chug, done. ... Speaking as somebody whose original network connection was *manually* dialed for about a year until we got an autodialer, Barry's got a point. Our manual dialing ended three years ago, and I *still* break out in smiles every time I hear the autodialer clicking away... -- Henry Spencer @ U of Toronto Zoology {allegra,ihnp4,linus,decvax}!utzoo!henry
dave@pyramids.OZ (Dave Horsfall) (07/24/85)
> Netnews via magtape is not a strange notion at all; I believe the > Australians get most of their newsgroups that way. > -- > Henry Spencer @ U of Toronto Zoology > {allegra,ihnp4,linus,decvax}!utzoo!henry Not for over a year, mate ... -- Dave Horsfall ISD: +61 2 438-1266 Lionel Singer Group STD: (02) 438-1266 20 Waltham St ARPA: munnari!pyramids.oz!dave@SEISMO.ARPA Artarmon NSW 2064 UUCP: seismo!munnari!pyramids.oz!dave AUSTRALIA ACS: dave@pyramids, dave@elecvax, dave@runx --
dcm@mimir.OZ (Dennis Mills) (07/24/85)
> Netnews via magtape is not a strange notion at all; I believe the > Australians get most of their newsgroups that way. We received this message today strapped to the back of a platypus. Worried that some people may get a false impression of our communications systems, we dashed off this reply. This was sent via our high-speed link, which goes by camel between the major bark hut villages. Unfortunately our previous message was routed via the Simpson Desert and the carrier died. We apologise for any delay. :-) :-) :-) :-) :-) :-) :-) :-) :-) :-) :-) :-) :-) :-) :-) :-) :-) :-) -- IM Group, CSIRO Div of Manufacturing Technology, Melbourne, Australia. UUCP: mulga!mimir.dmt.oz!iem@decvax.uucp PHONE: +61 (03)418-0211 ARPA: iem%mimir.dmt.oz@seismo.arpa CSNET: iem@mimir.dmt.oz ACSNET: iem@camel.platypus.oz :-)
throopw@rtp47.UUCP (Wayne Throop) (07/25/85)
> > Netnews via magtape is not a strange notion at all; I believe the > > Australians get most of their newsgroups that way. > No they (we) don't. [...] > ps: Henry: I've been waiting for you to say something utterly > and demonstrably wrong for ages now ... finally caught one :-) Wait a minuite... how did you find out that Henry doesn't believe the statement you quoted? I hope you haven't been consorting with telepathic Space Aliens! :-) -- Wayne Throop at Data General, RTP, NC <the-known-world>!mcnc!rti-sel!rtp47!throopw
henry@utzoo.UUCP (Henry Spencer) (07/26/85)
> In article <5800@utzoo.UUCP>, henry@utzoo.UUCP (Henry Spencer) writes: > > Netnews via magtape is not a strange notion at all; I believe the > > Australians get most of their newsgroups that way. > > No they (we) don't... > ps: Henry: I've been waiting for you to say something utterly > and demonstrably wrong for ages now ... finally caught one :-) "The Lord has delivered him into my hands!" -- Huxley Sorry, you don't get the prize for that one. I didn't say that Australia did get its news that way, just that I thought it did. :-) Which is 100% true, since I wasn't aware things had changed. Better luck next time :-). -- Henry Spencer @ U of Toronto Zoology {allegra,ihnp4,linus,decvax}!utzoo!henry