Return-Path: Received: by massis.lcs.mit.edu (8.7.4/NSCS-1.0S) id LAA13965; Mon, 26 Aug 1996 11:46:02 -0400 (EDT) Date: Mon, 26 Aug 1996 11:46:02 -0400 (EDT) From: ptownson@massis.lcs.mit.edu (TELECOM Digest Editor) Message-Id: <199608261546.LAA13965@massis.lcs.mit.edu> To: ptownson@massis.lcs.mit.edu Subject: TELECOM Digest V16 #438 TELECOM Digest Mon, 26 Aug 96 11:46:00 EDT Volume 16 : Issue 438 Inside This Issue: Editor: Patrick A. Townson Re: Why is the Internet So Slow? (Peter Morgan) Re: Why is the Internet So Slow? (Brian Mulvaney) Re: Why is the Internet So Slow? (Henry Baker) Re: Why is the Internet So Slow? (Matt Ackeret) Re: Why is the Internet So Slow? (Dave Read) Re: Why is the Internet So Slow? (Steve Schear) Re: Why is the Internet So Slow? (Charlie Seelig) Re: Why is the Internet So Slow? (Andrew C. Green) Re: Why is the Internet So Slow? (Garrett Wollman) Re: Why is the Internet So Slow? (Tim Shoppa) Is the Internet Slow? (David Richards) Internet Overload (TELECOM Digest Editor) ---------------------------------------------------------------------- From: Peter Morgan Subject: Re: Why is the Internet So Slow? Date: Mon, 26 Aug 1996 10:25:57 +0100 In message Bradley Dunn writes: > The reason a lot of end users think the Internet is slow is because of > their own equipment, though. They may have a computer that only has 4 > or 8 MB of RAM, which is really insufficient for running the latest > and greatest browsers. Most end users are also using a 14.4 or 28.8 > Kbps modem for their access. This "pipe" from the end user to the ISP > is usually the smallest pipe bandwidth-wise. I'm over in the UK. While I mostly agree that individuals will have a number of different things causing them to suffer "slowness", it can often be the case that some software is just more efficient. When I was typing this "offline" (the Zetnet service I use can be used like a BBS and passes "packed" [compressed] mail/news in both directions) I was in the middle of an FTP from "ftp.best.com" which I believe is in California. I'm attempting (fourth time) a transfer of approx 3.5 Mb and found that there was a significant difference between using Netscape's "save file" and WS_FTP. The system I'm running is 2x 540 Mb HD, 16 Mb RAM, 486DX2/66 with a (yes, don't laugh) 14,400 modem (well this has old UARTs anyway). WS_FTP is four to five times faster, and I started this transfer as soon as my dial-up connection failed (after approx 160 minutes :-( ) which, while a local call, is charged for :-< oh yes, I reached 3.4 of 3.5 Mb :-< There's a chance of some problem in the UK, which hampered earlier connections, but I have uploaded 3 Mb in 40 minutes to a UK host and the current estimate for the remaining 70% of this download is some 80 minutes connect time ... ah well ... a loaded FTP site in the middle of a Sunday afternoon is likely to suffer a little, and 5000 miles is worth remembering too (instead of 200 miles to London) :-> Peter ------------------------------ Date: Sun, 25 Aug 1996 17:20:45 -0700 From: Brian Mulvaney Reply-To: brianm@rain.com Subject: Re: Why is the Internet So Slow? > I am performing some research and am interested in hearing from anyone > with an interesting theory of why Internet access is so slow? I would be just as interested in hearing an explanation of how the Internet can be so fast. Haven't you ever marvelled at how you can run a database lookup over the Internet, to a server located who knows where, and get an almost immediate return? Such good response that it's actually faster than querying a database on your local ethernet? Here's an interesting comparison (albeit not apples to apples): I just queried Alta-Vista for "Telecom Digest". I got a full return in under three seconds. The search shows about 300 matches. Now I'll run a query against a customer information database my company uses that should return about 300 records using an indexed field as the search criteria ... The query took about five seconds to run including processing by the server and the client. The query was generated by a Windows application running on a Pentium PC and processed by a high end AIX server running Sybase. The database server is located about 100 feet from me. There should be only one router between my PC and the server. Now look at the network path for the query to Alta-Vista: From my PC to a local router; to another router; onto a T1 across the corporate WAN for a trip from the west coast where I am back to Maryland where corporate HQ is; maybe another router hop at HQ; through the corporate firewall to PSInet over a 56kbs link; a few hops to get to the Alta-Vista site (in Palo Alto?); some processing at the Alta-Vista server, some data returned over the Internet; some processing by Netscape Navigatior to format the page and voila! -- three seconds. If you ask me, the Internet feels pretty darn fast. Brian brianm@rain.com ------------------------------ From: hbaker@netcom.com (Henry Baker) Subject: Re: Why is the Internet So Slow? Date: Mon, 26 Aug 1996 02:02:26 GMT In article , Steve Rice wrote: > In TELECOM Digest V16 #429 Pat said: >> the net is just getting a lot more traffic than it used to get. > Bob Metcalf [sp] wrote a few articles in {Info World} last December > that predicted the collapse of the Internet. His major premise was > that as a collection of independent networks, there is not much > incentive to take responsibility for "quality of service" issues. It > is easy for an ISP to blame their problems on other ISPs, telcos, etc. Bob Metcalfe was on CSPAN2 this morning talking about the 'collapse of the internet'. A key point in his argument is that the internet is designed from the ground up to _share_ traffic. Yet the whole point of _intrAnets_ is to _filter out_ traffic with their firewalls, so that such traffic now has to go _around_ a network, rather than _through_ it. Consider, for example, what would happen if AOL were to suddenly declare that all of the freeways in the Washington, DC, area were _off-limits_ to any but AOL subscribers. All of a sudden, anyone wanting to drive south from Baltimore would have to drive _all the way around_ Washington, DC, and _not_ on the beltway, either! Ditto for the Fortune 1000 companies. So now we have large number of private intrAnets that parallel one another and don't share facilities, and much of the economies of the intErnet then go away. Note that the main impetus for setting up these parallel facilities stems from _fear_ -- the fear of being hacked, and the fear of being swamped. Perhaps if communications service providers provide _bandwidth_ with QOS _guarantees_ and high quality _encryption_, then at least the service provider can take advantage of the statistics of the traffic to reduce prices and still make money instead of selling raw T1's. www/ftp directory: ftp://ftp.netcom.com/pub/hb/hbaker/home.html ------------------------------ From: mattack@eskimo.com (Matt Ackeret) Subject: Re: Why is the Internet So Slow? Organization: Eskimo North (206) For-Ever Date: Sun, 25 Aug 1996 23:35:35 GMT In article , John Agosta wrote: > And although I am not trying to be particularly protective of the 'jerks > who send out thousands of pieces of mail trying to sell something' there > *are* a lot of people using the net who are as equally sincere in their > efforts as you and I. I honestly do not think many of the people who do > junkmail and spam realize they are considered offensive. PAT] Hopefully responding to all of these back to the author and the accounts "abuse", "postmaster", and "root" at the site from which the spam came from (when that can be determined) is helping to teach them that. While this obviously is after the fact, responses to e-mail and Usenet spam complaints is often that the user has had their account cancelled. (No, I don't think *all* of them are from people who get an account, spam with the expectation it will be cancelled, and go on to another account to start the process over again.) Also, most people don't seem to be aware you can cancel Usenet articles you have posted accidentally (or containing an error you want to fix). unknown@old.apple.com ------------------------------ Date: Sun, 25 Aug 96 13:24:00 PDT From: dave@kentrox.com (Dave Read) Subject: Re: Why is the Internet So Slow? Organization: ADC Kentrox Industries, Inc. On 20 Aug 1996 18:00:17 PDT, Mark Friedman <71534.332@CompuServe.COM> said: >> I am performing some research and am interested in hearing from anyone >> with an interesting theory of why Internet access is so slow? >> For instance, >> Is it the data com backbone, the protocol, the routers, the >> Servers, or the browsers? > Pac*Bell has reported serious problems with their ATM switch at their > California NAP. For full details of what's going on and how they hope > to fix it, read > http://www.pacbell.com/products/business/fastrak/networking/nap/nap2-6.html > My understanding is that Pac*Bell is the only NAP operator using ATM > at this time. Pat: The data on the Kentrox ADSU bugs is almost a year out of date. The problems listed were fixed quite awhile ago. TM David Read (dave@kentrox.com) ADC|Kentrox Portland, Oregon, USA ------------------------------ Date: Mon, 26 Aug 1996 11:08:03 GMT From: azur@netcom.com (Steve Schear) Subject: Re: Why is the Internet So Slow? Mark Friedman <71534.332@CompuServe.COM>, wrote: > I am performing some research and am interested in hearing from anyone > with an interesting theory of why Internet access is so slow? > For instance, > Is it the data com backbone, the protocol, the routers, the > Servers, or the browsers? Most people who complain of slow Internet access do so when using Web browsers, since its now the most widespread real-time Web use. A great article on this topic was in the May Boardwatch Magazine, "Bandwidth Arithmetic," pg. 8 (http://www.boardwatch.com). The conclusion of the article is that though some backbone bottlenecks are an overrated source of delays and that much or most are due to inadequate Web server resources (bandwidth and/or server performance). If Internet voice does indeed take off as predicted it will have a major impact on the ISPs and backbone traffic and, if not delt with expedituously, could create other bottlenecks. Steve Schear | Internet: azur@netcom.com Lamarr Labs | Voice: 1-702-658-2654 7075 West Gowan Road | Fax: 1-702-658-2673 Suite 2148 | Las Vegas, NV 89129 | ------------------------------ Date: Sun, 25 Aug 1996 15:03:11 -0400 From: cseelig@tiac.net (Charlie Seelig) Subject: Re: Why is the Internet So Slow? Mark Friedman wrote on 21 Aug 1996: > I am performing some research and am interested in hearing from anyone > with an interesting theory of why Internet access is so slow? > For instance, > Is it the data com backbone, the protocol, the routers, the > Servers, or the browsers? There is a good article on the subject entitled "Web Brownout" by Simson Garfinkel in the September 1996 of Wired. Charlie Seelig ------------------------------ Date: Sun, 25 Aug 1996 14:08:18 -0500 From: Andrew C. Green Subject: Re: Why is the Internet So Slow? > [TELECOM Digest Editor's Note: [...] > We see the same thing on the streets in every large city. Thirty to > forty years ago traffic on the road and highways was not nearly as > heavy as it is now. Other than 'rush hour' roads were relatively quiet > all day with cars going past occassionally. Now the slow times of day > look like 'rush hour' used to look thirty years ago on busy roads, etc. > Prior to the construction of the interstate highway system in the 1950- > 60's era people still traveled; they just did not travel as much or go > as far, and it took longer to get where they did go. The net is the > same thing all over again. Actually, Pat, I think the (extremely overworked) Information Superhighway analogy fits rather well in answering this question: it's not the backbone that's slow and it's not the interstate highways that are slow; it's all the congestion involved in getting in/on and off again that wastes time. My ISP may have a grand T1 connection to the Internet but I may be using my ISP account with a 33mhz 486 and a 14.4 modem. One step up the line from me, my ISP may actually have a bank of modems faster than my own, but be laboring under the load of a huge number of customers all trying to make use of the great deal they got on their new accounts when the ISP decided on the latest cut-rate new-subscriber offer to advertise. Somewhere up the line from there, the backbone of the Internet is happily pushing stuff around at the speed of light, while data arriving at its destinations backs up on the exit ramps. > And although I am not trying to be particularly protective of the 'jerks > who send out thousands of pieces of mail trying to sell something' there > *are* a lot of people using the net who are as equally sincere in their > efforts as you and I. I honestly do not think many of the people who do > junkmail and spam realize they are considered offensive. PAT] Well, I think they do, but they rationalize it by thinking that if only 3% of the recipients are complaining, then they can't be doing anything wrong. And those who get their accounts yanked shortly thereafter are probably embarrassed enough; they're not likely to go publicizing their fate in some effort to ward off the next round of suckers ready to buy Floodgate, Lightning Bolt, Spewmaster, Barfomatic or whatever the next overpriced perl package will be labeled. (Side note: How interesting that the product being advertised via junk email is almost invariably the junk emailing package itself.) And on a local level, yes, the jerks can have a slowing effect on both their system and the target systems. The recipients must store, download and wade through the junk, and the jerk's host must cope with the angry responses sent in reply. Through all this, the Internet backbone goes on its merry way, cheerfully pushing TELECOM Digest, Floodgate and Make Money Fast all over the world. Andrew C. Green (312) 266-4431 Datalogics, Inc. 441 W. Huron Internet: acg@dlogics.com Chicago, IL 60610-3498 FAX: (312) 266-4473 [TELECOM Digest Editor's Query: Ummm ... do I hear my name being taken in vain? PAT] ------------------------------ From: wollman@halloran-eldar.lcs.mit.edu (Garrett Wollman) Subject: Re: Why is the Internet So Slow? Date: 26 Aug 1996 11:41:33 GMT Organization: MIT Laboratory for Computer Science In article , Mark Friedman <71534.332@CompuServe.COM> wrote: > I am performing some research and am interested in hearing from anyone > with an interesting theory of why Internet access is so slow? [I deduce from the rest of the question that you really mean "World Wide Web browsing is so slow". "Internet access" in general is pretty speedy.] > Is it the data com backbone, the protocol, the routers, the > Servers, or the browsers? All of the above, plus one you didn't mention which is more important than all the rest: the poorly-designed content. Let's take each of these in turn: 1) The backbone. The principal problem is that there isn't one; rather, there are a fairly large number of competing providers of long-haul IP service, which have engineered their networks with varying degrees of success. Which network or networks you use depends on a large number of factors, including who your local provider is, who the other end's local provider is, where you (and they) are located, and what bilateral agreements exist between various providers and long-haul carriers. One of the principal problems in this area is that many of the attachment points -- where multiple providers exchange traffic -- are underengineered. 2) The protocol. The primary protocol of the World Wide Web, HTTP, runs over the primary protocol of the Internet generally, TCP. In order for standard TCP to work, every connection involves no fewer than three round trips from the origin to the destination. A single round trip can take as much as a second or two depending on how and where each end of the connection is attached to the network. A modified version of TCP, called Transaction TCP or T/TCP, reduces this to two round trips, but it is not widely deployed. None of this would have come into play if HTTP had been designed better to begin with; work is progressing to fix HTTP so that it no longer requires a new connection for each request, but it will be some time before this is widely deployed. 3) The routers. Most routers in the Intenet today use a queueing model which unfairly penalizes bursty connections. It will take a long time for all the routers to be replaced with ones which do not have this property. 4) The servers. A lot of Web servers are underpowered, or have insufficiently large connections to the network. Most Web servers these days are operated by Web service bureaus, which publish a company's content for a fee without the company itself actually being connected to the network; sometimes fifty or a hundred companies may be sharing the same machine and the same net connection for their Web service. The earliest Web server programs were not particularly speedy. 5) The browsers. Netscape Navigator attempts to grab more than its fair share of bandwidth by opening multiple simultaneous connections to the same server, to the detriment of other users. Thankfully, some server operators have recognized this problem and are refusing to allow their servers to be taken advantage of in this way. 6) The content. A lot of Web pages are designed by idiots who flunked out of graphic design school and think that they are qualified to sell their lack of skill for $50 a pop. These people are the ones who tend to play lots of obnoxious games with color, use huge image maps when a simple menu would suffice, use background images and noises, and otherwise interpose themselves between the user and the content. If your browser supports it, disabling support for all of these "features" can make the Web experience much faster and more pleasant. Garrett A. Wollman wollman@lcs.mit.edu [TELECOM Digest Editor's Note: Thank you very much for mentioning your item (6) above. I never have been really enthusiastic about the World Wide Web, feeling that it is stealing important shared resources from the important work on the net. I started using the web for this Digest and the Archives only because FTP has become so damnably *slow* in so many locations, including lcs.mit.edu. I do not know what it is about Info-Mac but people login to do downloads from there and they sit all day, tying up every single connection to this site. When the admin increases the number of allowable anonymous FTP connections, then the overall performance of this net at MIT suffers. He can allow for 25 connections or he can allow for 100 connections; no matter, five minutes later all allowed connections are in use. The more he allows, the slower things get for me as I sit here and type; the fewer he allows, the more people would complain to me that they were unable to get into the Telecom Archives. As it is right now, I just checked 'uptime' and see a load of 6.58 here on massis and I am the only user logged in as is usually the case. So I took Bill Pfieffer's advice and 'moved to the Web' with the Archives although I don't really know what to think about it at this point in time. According to him, if one is not on the Web these days, one might as well not be on the net at all. Some people write me to say how poorly the web page is put together, but it was all pretty much put together on the spur of the moment one evening in July, and I am not so much interested in a pretty web page as I am in being able to pass a large volume of text in as short a period of time as possible. I do not think the web page looks all that bad, and I certainly do not want to fall into the trap of lots of pictures and sounds. That just isn't my thing, and the pages like that I have seen do seem to take a very long time to load. I just don't know what I am going to do with the Telecom web page at this point in time. PAT] ------------------------------ From: shoppa@alph02.triumf.ca (Tim Shoppa) Subject: Re: Why is the Internet So Slow? Date: 25 Aug 1996 17:48:34 GMT Organization: Tri-University Meson Facility In article , Mark Friedman <71534.332@CompuServe.COM> wrote: > I am performing some research and am interested in hearing from anyone > with an interesting theory of why Internet access is so slow? Whad'ya mean, "slow"? I remember when I downloaded my first "Dilbert Newsletter" over a 6 baud modem connection, filling my 1kbyte hard disk. Fortunately, I was able to rewrite the operating system to make room, creating a new operating system I named "Dogbert's Operating System" (DOS for short). DOS was later adapted from its shareware version into something commercial by a weird guy with glasses who kept hanging around my house to meet my sister -- so you may have heard of it already. Tim. (shoppa@triumf.ca) P.S. My elite status in Dogbert's New Ruling Class ("Pioneer Preference") gives me the right to exagerrate how life was in the early days. [TELECOM Digest Editor's Note: Well, I can tell you when I operated my computer bulletin board system 'Lakeshore Modem Magazine' from 1983 to 1985 and gave the users a choice of 300 baud or 110 baud everyone thought that allowing a choice of the 'higher speed' 300 baud was a wonderful idea. Then when I documented a hardware modification to the Hayes Micro- modem card (which went in a slot in Apple ][ computers) which allowed for just a single cut on a trace and a small jumper which would allow the modem to be 'speeded up' to 450 baud (!) my users just loved it. My Apple ][+ was actually a 'Black Apple'; one in the production run that Apple Computer manufactured for Bell and Howell. I had so many traces cut on the mother board and jumpers added going to a bank of switches on the side that no one but me could figure it out. With Jerry Ablan I had earlier operated a BBS we called 'Think!' based on the old IBM corporate slogan, which we operated from 1979 to 1981 or so. We did that BBS on a Radio Shack Model 1, with all of I think about 8K of memory. My first home computer was an Ohio Scientific, Inc. model C-1-P, with 4K of RAM which I got in 1977. It used Microsoft BASIC as did the Apple ][+, only the Apple version of Microsoft's BASIC and DOS was called 'Applesoft'. The OSI C-1-P came with 4K RAM but one day the company sent out a 'software patch' and a new chip which increased the memory to 8K and to us, that was just like being in heaven! I still like DOS and use it now and then although I prefer Unix stuff. The first BBS anywhere in the world was here in Chicago you know; Randy Seuss and Ward Christianson had "Ward and Randy's BBS" beginning sometime in 1977. PAT] ------------------------------ From: dr@ripco.com (David Richards) Subject: Is the Internet Slow? Organization: Ripco Internet BBS Chicago Date: Sun, 25 Aug 1996 22:51:34 GMT In article , Mark Friedman <71534.332@CompuServe.COM> wrote: > I am performing some research and am interested in hearing from anyone > with an interesting theory of why Internet access is so slow? Perhaps you should first ask "IS Internet access slow, or is it just localized?" > For instance, is it the data com backbone, the protocol, the > routers, the Servers, or the browsers? The constantly predicted 'meltdown' of the Internet has not yet happened. But, some providers (Netcom comes to mind) may have localized traffic problems, and occasionally difficulties with Nationwide connectivity providers and specific NAPs lead to local congestion. Your question is like asking the US Dept. of Transportation in D.C. "Why is travel so slow?" Some cities (providers) have major traffic jams, and there may be construction and delays affecting some interstates, but overall the system is very healthy. Just because travelling from Atlanta to Chicago is slow, or because there's high packet loss when you use a web Browser on Compuserve to access Netscape, just means the route you're using has problems. David Richards Ripco, since Nineteen-Eighty-Three My opinions are my own, Public Access in Chicago But they are available for rental Shell/SLIP/PPP/UUCP/ISDN/Leased dr@ripco.com (312) 665-0065 !Free Usenet/E-Mail! [TELECOM Digest Editor's Note: As long as I mentioned my own experiences with BBS's from fifteen years ago, I should point out that Dave Richards and his associate Bruce Esquivel are no spring chickens either. Their system Ripco has been running now since 1983, although like all of us back then it was a single server, one caller at a time BBS; nothing like the fine ISP operation they have today on the north side of Chicago. Everything is relative I guess; indeed, is the Internet 'slow'? PAT] ------------------------------ Date: Mon, 26 Aug 1996 09:57:23 EDT From: ptownson@massis.lcs.mit.edu (TELECOM Digest Editor) Subject: Internet Overload An article in the {Chicago Sun-Times} for Sunday, August 25 discusses in some detail (a full page article) the topic of the huge volume of traffic on the net and the belief of many that sometime during 1997 the net will begin to experience 'a series of brownouts' in which portions of the network fail for a short period of time. Beth Gaston of the National Science Foundation which helped develop the Internet backbone (NSFNET) is quoted discussing plans for a so-called 'Very High Speed Backbone Network Service'. Various other network adminis- trators are quoted in the article as well. You may wish to review the article in detail at http://www.suntimes.com. PAT ------------------------------ TELECOM Digest is an electronic journal devoted mostly but not exclusively to telecommunications topics. It is circulated anywhere there is email, in addition to various telecom forums on a variety of public service systems and networks including Compuserve and America On Line. It is also gatewayed to Usenet where it appears as the moderated newsgroup 'comp.dcom.telecom'. Subscriptions are available to qualified organizations and individual readers. Write and tell us how you qualify: * ptownson@massis.lcs.mit.edu * The Digest is edited, published and compilation-copyrighted by Patrick Townson of Skokie, Illinois USA. You can reach us by postal mail, fax or phone at: Post Office Box 4621 Skokie, IL USA 60076 Phone: 847-329-0571 Fax: 847-329-0572 ** Article submission address: ptownson@massis.lcs.mit.edu Our archives are located at mirror.lcs.mit.edu. The URL is: http://mirror.lcs.mit.edu/telecom-archives They can also be accessed using anonymous ftp: ftp mirror.lcs.mit.edu/telecom-archives/archives A third method is the Telecom Email Information Service: Send a note to tel-archives@mirror.lcs.mit.edu to receive a help file for using this method or write me and ask for a copy of the help file for the Telecom Archives. ************************************************************************* * TELECOM Digest is partially funded by a grant from the * * International Telecommunication Union (ITU) in Geneva, Switzerland * * under the aegis of its Telecom Information Exchange Services (TIES) * * project. Views expressed herein should not be construed as represent-* * ing views of the ITU. * ************************************************************************* Finally, the Digest is funded by gifts from generous readers such as yourself who provide funding in amounts deemed appropriate. Your help is important and appreciated. A suggested donation of twenty dollars per year per reader is considered appropriate. See our address above. All opinions expressed herein are deemed to be those of the author. Any organizations listed are for identification purposes only and messages should not be considered any official expression by the organization. ------------------------------ End of TELECOM Digest V16 #438 ******************************