Why AOL Always Seems To Screw Up

The Internet Gang is Back!

Francis Vale

AOL, Leader of the Pack?

Just when you thought the notorious Internet Gang was gone, or at least getting somewhat civil, another fractious faction has sprung up. But unlike Netscape, Sun, and Microsoft, whose Internet shenanigans were easily visible, this other Net crowd has been, until quite recently, operating mostly sight unseen--which makes this group's increasingly bad behavior all the more insidious. The initials emblazoned on this techno-gang's jackets are "ISP" (Internet Service Provider). Many think this gang's crazed ringleader is the apparently out of control America On Line.

Despite what some panic-struck journals are saying about the falling IP-sky, the real issue is not lack of Internet bandwidth. Indeed, miles upon miles of "dark" fiber are being lit up, and ATM backbones are starting to be deployed in large numbers. However, if your ISP hasn't upgraded its computer hardware, software, and communications systems (most especially, by radically increasing the number of its dial-in modems) to take full advantage of these go-fast nets, then they may as well be two orange juice cans strung together with twine. In this regard, AOL seems to have cornered the market on Minute Maid O.J.

Increasingly, many ISPs, like AOL, have had less and less incentive for spending the big upgrade bucks. Even worse, a number of rival ISPs have been engaged in secretive, internecine warfare, trying to knock each other out of commercial contention.

It wasn't always this way. Formerly, the Net was a government-backed, freewheeling, low speed, egalitarian assemblage of scientists, engineers, and plugged-in users. And the who-carries-what and how issues were mostly non-existent, as spirited cooperation and elegant hacks ruled the Internet day. However, this Internet socialism is fast becoming another quaint, Cold War relic. Now that the Net is Big Business, Inter-ISP acrimony is rapidly racheting up. We are beginning to see the equivalent of digital drive-by shootings. Flame wars; malicious hacking by ISPs into the InterNIC registry to alter domain name information of a rival ISP; deliberately lost bits; anonymous message flooding of a rival's system to bring it to its knees; and a host of other bad ISP behavior are on an alarming rise.

These inflammatory incidents are starting to decay the once friendly and safe Internet neighborhood. Unfortunately, most users are blissfully unaware that their Net community has these nasty, ISP intramural problems. As a consequence, when users see their Web client data slow down to a crawl, the natural response is to "blame the Internet." And their ISP will probably not do much to change that mostly erroneous perception. Remember what AOL's spin doctors were saying about those busy circuits? (Or at least until several state governments hit the company with millions of dollars of fines.)

To see why it's still so easy for many providers to get away with their anti-progress, anti-social behavior, let's consider the differences between the TELCOs and ISPs: 1) Today, ISPs operate in a completely unregulated, performance standards-bereft environment. But the TELCOs work within a set of tightly regulated Federal and State rules, and under the good conduct standards aegis of special oversight agencies. 2) ISPs work in loose, non-binding cooperation to pass through each other's data traffic. However, an ISP doesn't get paid for passing it on. But the TELCOs have long had multilateral compensation agreements for carrying one another's communication's water. 3) Your ISP doesn't have to tell anything to anyone, including you, the customer, about how good--or bad--their systems are operating. But the TELCOs are legally bound to pass on their network performance statistics to powerful governmental agencies, such as the FCC. OK, now let's consider how important the Internet is becoming to the everyday operations of global business, and even the fabric of your daily life. Uh, oh. Houston, we have a problem....

With a lack of uniform commercial system standards, no mandatory performance reporting, Gingrich-style governmental regulation, and $720 million in provider revenues at stake as of early 1996, the ISP's incentives for taking some sort of hostile business action are getting increasingly large (ISP revenue figures from Maloff Company, Dexter, MI). It is therefore now up to users to bring law and order into Internet Tombstone--Either the offending ISPs quickly get back into line, or face the Doc Holiday consequences. But any such Internet Neighborhood Watch--despite the warning message delivered by those massive fines against AOL-- has its work cut out for it. Users just do not have the inside ISP information to catch out the packet busting perps. Nor do users have legitimate access to their ISP's traffic capacity handling data. Moreover, on the far flung, sometimes anonymous Net, the issue of who did what to whom can become quickly obscured. So, unless an ISP is caught red handed in a malicious act, or their system is so obviously, outrageously out of date, users may still be left in the dark. All in all, not a pretty picture, and it is one diametrically at odds with the rosy scenarios continually being painted for the Internet.

Follow The Money

Let's step aside for an Internet minute, and focus on just one nettlesome ISP issue: Where is the provider's financial incentive to do an expensive systems upgrade if it mostly helps another ISP gain business? Why should one ISP potentially cut its own throat by more efficiently passing on a rival's traffic? Not an easy answer in today's highly competitive Internet environment.

For example, consider what happened last year when MCI won the contract for linking together the 13 regional centers that formerly hooked up to the NSFnet. Belatedly, MCI learned that it was also responsible for carrying an enormous, hidden data load. Non-paying users around the world also wanted access to those databases, FTP servers, etc., hosted at the regional centers. And so, a flood of unexpected, unplanned for traffic suddenly showed up on MCI's doorstep. But the MCI facilities tasked for managing the 13 centers weren't up to the global job. The result was woefully predictable: a significant percentage of packets got lost, and never made it across the Net. (AOL, despite what the overheated media may have led you to believe, was not the first big company to get caught with its Internet pants down.)

Finally, MCI, like AOL, said it would eat the massive costs for upgrading its systems to satisfy the data hungry needs of the worldwide Internet community. No doubt, from both company's respective points of view, they see themselves as paying the traffic-carrying freight for all those other users--and their respective ISPs. Much the same type of user/ISP "gimme-gimme, but we won't pay" scenario is also being played at thousands of small ISP shops; except they don't have MCI's or AOL's vast resources. This enormous, lose-the-home-or-else pressure has therefore produced some rather erratic, and sometimes malicious, ISP behavior. Which is why the new RSVP standard has as much chance of Internet success as Frosty the Snowman in hell.

RSVP -- Don't Even Bother Replying

Present day practices of routing IP packets consist of taking a single message, chopping it up into little discrete chunks, sending them off across the Net (many of which don't even travel together) and then assembling them at their final destination. Obviously, this carved up, distributed packet sausage is not the best way to send streaming multimedia data, like video or audio. (This is why your voice over the Internet usually sounds like Charley Tuna swimming in a choppy ocean.) A newly proposed technique, called RSVP, supposedly solves this packet butchery problem by marshaling the multimedia packets into a contiguous, reserved ahead, end to end stream. RSVP thus provides guaranteed quality of service across the IP-based network.

After two years of effort by multiple vendors, the Internet Engineering Task Force (IETF) is about to put its imprimatur on RSVP. But in order to work, RSVP needs a willing partner protocol: RTP (Real-Time Protocol). RTP sits next to TCP, and is the end to end transport protocol. The protocols must also be supported on both clients and routers. A company called Precept (Palo Alto, CA) has developed client-side software that incorporates both RSVP and RTP, and has already done some significant OEM deals. E.g., Netscape's LiveMedia architecture uses Precept's software. With regards routers, Cisco, and Bay Networks have pledged RSVP support on their products.

Via RSVP and RTP, the requesting client tells the network how much bandwidth it needs for a particular application (e.g., 64 Kbs for voice-quality, 1.54 Mbs MPEG video, 128 Kbs to 1 Mbs for videoconferencing, etc.). The request then goes from router to router over the network, clearing a path from the sending client to the receiving client. The RSVP/RTP protocols thereby guarantee that an uninterrupted stream will be available via a pre-allocated network connection.

Unfortunately, the "you scratch my ISP back, I claw yours" game also applies to the use of RSVP. The various ISPs who route the traffic on through for the originating ISP and its clients have to set up the reserved ahead bandwidth, monitor the traffic, and most important, honor their respective chargeback deals. For example, an ISP might charge $5 an hour for an RSVP-audio stream of only 64 Kbs, while charging $40 an hour for high bandwidth videoconferencing. The originating ISP must then compensate the other ISPs out of its RSVP customer revenue stream.

So, for the RSVP scheme to work properly, rival ISPs have to closely cooperate. However, it is reasonable to expect that a competitor ISP may charge different, lower rates. In potential retaliation, the offended ISP may drop a few rival's packets along the way. Even more vexing, how do you accurately charge when several ISPs are involved in the same transaction? Thus, the RSVP pre-allocation scheme, and the associated multi-chargeback deals, can easily get rather messy in the Internet's highly competitive, unregulated environment.

Internet RSVP is somewhat analogous to American Airlines agreeing to fly some of Delta's passengers, and then billing back Delta. But suppose AA is overbooked; or occasionally refuses to cooperate for competitive reasons; or overcharges; or fails to tell Delta that it flew some of its passengers. Which company's passengers do you think end up getting stuck in East Oshkosh, or price gouged?

Therefore, unless rival ISPs sign on to the plan, and allocate and honestly charge for space on its circuits to route each others' RSVP packets, this new and very needed technology may not even get off the ground.

Rub Em Out, Louie!

For all these ISP-business reasons, and impending ISP-customer revolt (also fueled in large part by AOL's highly publicized problems), we on the cusp of a huge consolidation and industry shake-out, as numerous, small, one note ISPs are either rubbed out, bought out, or merged out of existence. Large, regional, multi-part players will likely be the ISP norm come the year 2000. There is also the added pressure of corporate customers wanting a single point of contact for end to end Intranet service and application development. Consequently, new and powerful Net syndicates are starting to emerge; with one of the most popular being a combination of a large ISP and a systems/network integrator.

For example, BBN Corp. (Cambridge, MA) has hooked up with Andersen Consulting (Chicago, IL) to form a new, standalone, one stop, Intranet subsidiary. And UUNet Technologies (Fairfax, VA) has joined with systems integrator USConnect to offer their customers unified Intranet development and connectivity. UUNet has also begun offering routers to its corporate customers that support frame relay, leased line, and three types of ISDN configurations; the latter going from BRI to T-1. This service, called Pipeline 130, is big league stuff, at one time only offered by the large telecommunications carriers.

Furthermore, thanks to the Telecommunications Deregulation Act of 1996, the Internet gansta' wars are quickly spreading into other, hitherto "safe," TELCO turf. And some of these tough ISP guys are very well armed. The new, readily available communications technologies are becoming the Internet equivalent of gang members' Uzis. In the hands of an aggressive ISP, these techno-guns can quickly give it a marketing advantage; especially if goes up against a big, slow, and formerly over-protected RBOC.

ADSL, Who Ya Gonna Call?

Let's examine one of these new TELCO terror technologies: ADSL (Asymmetric Digital Subscriber Line Service). This twisted copper wire pair, broadband technology provides for 1.544 to 8.448 megabits downstream to the user, and 16 to 640 kbs upstream back to the provider. The extreme range of speeds reflects how close you are to your provider's central office. Do to the fact that an ADSL modem can only overcome a maximum of 1,500 ohms of resistance, the longer the local loop, the less the throughput. If 15,000 feet of 26 gauge wire sits between you and the CO, then 1.544 mbs up, and 16 kbs down are all you get. But if you are only 9,000 or so local loop feet away from the provider (generally considered to be the average distance for U.S. customers), then you might be looking at T1 speeds right into your home PC. Presuming of course, that line noise, the speed of your PC, cross talk between wires, etc., don't parasitically sponge up most of your ADSL throughput.

Originally, ADSL was seen as a way to catapult the phone companies into the lucrative cable TV business. And as an added plus, ADSL gave the TELCOs some CO switch relief; for unlike ISDN, it separated out the voice traffic from the data traffic. But the TELCO video-on-demand race was quickly in the pits due to lack of customer interest. Then enter the Internet. Voila! ADSL is now being touted by many of the TELCO's and RBOCs as the Next Great Thing for Internet users. The First Great Thing, presumably, was ISDN. But given ISDN's rampant tariff gouging, spotty service, and general seeming incompetence on the part of the RBOCs, this might not be saying too much about ADSL's future prospects.

But, fortunately, ADSL, like money, knows no enemies, and has many friends. And here is an example: In September, 1996, Chicago users saw ISP InterAccess, Inc., outgun and outsmart Ameritech Corp., the multi-billion RBOC serving this region of the country. How did it do this? Very shrewdly. InterAccess first got Ameritech to lease its enormously valuable copper links in downtown Chicago for a measly $7 each, and for an equally whopping $22 each in the outlying areas. Next, at the time of the InterAccess copper wire deal, Ameritech was the only one of the big RBOCs not to have announced ADSL (although it has belatedly done so).

InterAccess, with Ameritech's cheap copper in hand, and using Westell's FlexCAP ADSL system, then demonstrated Chicago's first ever trial of ADSL service. But this is no longer a trail. InterAccess is now offering the country's first ever ADSL service on a production basis. And Ameritech? Still in trial mode! The new InterAccess service costs $200/month for individual PCs, and $1,000/month for corporate LANs. There is also a $200 to $1,000 one time install fee, and the Westell ADSL modem costs $1,500. This particular InterAccess ADSL service provides 64 kbs uspstream to the ISP, and 1.5 mbs downstream to the user.

Bang! Bang! Bang! Capone's home town just saw another Valentine's Day Massacre. The marketing wounds inflicted on Ameritech in this Internet turf war, while certainly not fatal, probably had some of its executives limping about from the stray PR shrapnel.

If InterAccess could pull this off, admittedly in a very clever, propitious way, then which RBOC is next to be knee-capped by some other hungry, and unregulated ISP? To be sure, not all the other RBOCs are quite so asleep at the ADSL switch. For example, NYNEX Corp., has just announced that it is now in league with Lotus Development Corp., and also Westell, to provide ADSL service. The ADSL trial is intended to link Lotus corporate nets and Lotus employees homes together.

However, also consider this: NYNEX was recently required by Massachusetts to cut its rates by $20 million over a period of two years as a penalty for failing to meet service levels required by state regulators. In addition, residential user complaints against NYNEX for poor service--some people have had to wait for up to two months just to get simple POTS--are up 28% over the previous year. Business complaints for the same period are also up 17%. Many blame this poor service record on the massive layoffs that NYNEX is going through, as it slims down to become "more competitive." If POTS (plain old telephone system) is such a shambles in New England, then how does that augur for NYNEX massively deploying sophisticated new technologies, like ADSL? This RBOC should also start looking over its shoulder. It just may see a lean and hungry ISP sizing it up for ADSL lunch, Chicago-style.

On a more potentially interesting note, GTE Telephone Operations has announced a similar corporate net/employee ADSL hookup. This time, though, it's with Microsoft in Redmond, WA. The trial also uses ADSL technology from Westell, as well as products from Amati Corp. If it stays true to its business ways, Microsoft will likely gain some critical marketing and technical insight into this high speed technology. But perhaps most important, Microsoft will also be running its NT Server platform within the ADSL environment. MS NT is using a key communications technology that Microsoft jointly developed with the help of other vendors, called Point-to-Point Tunneling Protocol (PPTP). This technology encapsulates remote client traffic into secure frames for Internet transmission, which can then only be decoded by the access server. In this ADSL trial, the PPTP decoder is running on the NT Server.

This trial conjures up all kinds of scenarios. Microsoft as a national ADSL ISP? Hey, the regular MSN is already a big bad bear crushing down trees in the ISP woods. Now that's a thought sure to send shivers down an RBOC executive's spine. Or perhaps Microsoft might just bundle NT Server, PPTP, Exchange Server, and ADSL together as a turnkey suite for the thousands of other ISPs out there. If that ever happened, it's no longer a simple frisson-- it's RBOC cardiac arrest time.

ADSL chip set prices are also in the typical semiconductor freefall. Analog Devices has just announced that it has reduced its ADSL chip count from twenty to five, and the new set is expected to sell for about $80 to OEMs. The next generation ADSL sets from Analog, and other vendors will almost certainly have fewer chips, and cost less. Thus, a provider's ADSL line costs will soon be plunging dramatically, to just a couple of hundred dollars (presently, these line costs range as high as $2,000; still prohibitive for most ordinary consumers). Cheap line costs, plus potentially inexpensive NT ADSL servers, add up to a great business opportunity for raptor ISPs. Finally, as it seems that ISPs will continue to follow the Internet model pricing model with ADSL--low fixed monthly fee, and unlimited access-- then the pay as you go RBOCs are in for even more of a rude awakening.

But even with all the RBOC bumbling, and cheap access, the question still begs to be asked: will the combative ISPs cooperate together to provide standardized, reliable service over a large national region? No matter how stumbling, tardy, or otherwise inept some RBOCs might be, they at least have to toe the line, or pay the regulatory piper (see NYNEX, above). So don't assume that this high speed ISP game will automatically go to the swift and voracious (although Microsoft and its MSN will certainly try.)

TV Cable, Look at Me!

Regardless, TV cable companies strongly believe they have a good shot at getting your Internet access business--and also stealing the RBOC's telephony business, to boot. One cable company, Continental Cablevision, after a beta test of approximately 200 users in the Boston area, has just rolled out its first commercial product offering for Internet users. The service builds on Continental's hybrid coax and fiber plant, and uses ATM (Asynchronous Transfer Mode) as its protocol. In comparison to ADSL, the speed of the Continental system is comparative slow. In the beta tests, data transfer rates for individual users topped out at 350 Kbs. Regardless, this is still twelve times faster than a 28.8 Kbs modem, and almost three times faster than a 128 Kbs ISDN line. The new service will support Netscape's browser and e mail. Significantly, Continental also has its eye on the TELCOs, as it further plans to support telephony service.

A big question immediately comes to mind: If coax cable has a bandwidth of 30 megabits per second, which is far greater than twisted pair, then why is Continental only getting about 350 Kbs per user? This is almost two orders of magnitude slower than cable's maximum throughput. So where did all the speedy bits go? There are many answers to this question, but here are the most essential ones. First, the 30 Mbs cable uses a 10 megabit/sec10Base-T Ethernet port at the user's PC, which immediately cuts its speed down by factor of three. But most bandwidth damaging, TV cable is a party line system. The more users that pile on, the lower the effective throughput. E.g., 100 users simultaneously jumping onto the same 30 Mbs cable will see only 300 Kbs effective throughput. This is in sharp contrast to ADSL, which provides a separate, dedicated line per user. Thus, a rapid increase in ADSL users does not slow down the overall system.

Furthermore, like ADSL, TV cable is also asymmetrical. The TV cable system was not originally designed to act as a bi-directional pipe, nor to provide point-to-point access. The TV cable companies never expected that their subscribers would be sending home videos back to them. So, even though you can get 30/10 Mbs downstream, the user can only ship back 512 Kbs to 10 Mbs, depending upon the type of cable modem used, PC speed, number of users on-line, line noise, etc.

And there are significant, non-performance related issues, as well. The cable modems used by Continental and other cable companies are mostly incompatible with one another. Some modems, like Continental's, use ATM, while others use proprietary formats. Nor does it help compatibility matters much that each company's modems may also use different modulation techniques. Thus, you need the same make modem and access provider at both ends of the circuit. The result: ISP lock-in heaven. This cable speed therefore comes with a big and incompatible price tag. Also, as TV cable doesn't support point to point communications, remote access to a corporate host server is out of the question. Once again, an ADSL provider, especially for business use, seems to have a clear cut advantage over a TV cable-based ISP.

But if the TV cable companies don't yet have the technology to compete with systems like ADSL, they still have one very important ingredient: Money. Lots of it. For example, Time Warner is spending more than $4 billion to upgrade its cable plant. Other cable companies, like TCI, are also spending similarly huge sums of money. So, if the cable service is deficient today, tomorrow may yet bring another story. For example, TV cables use an unswitched tree and branch technology, with the main trunk continually being subdivided as it snakes its I Love Lucy way through the neighborhoods. If they need more bandwidth, then the cable company just drops another line into the neighborhood, thereby decreasing the number of users on any particular party line. This decrease in user numbers directly translates into an increase in individual user speed. Or, the company could dedicate another 6 MHz TV channel to carrying Internet data, which provides for yet another 30 Mbs. The equation is thus quite simple for the cable company executive: how many new users do I have to sign up to compensate for the loss of TV revenue? Finally, in a hybrid fiber/coax system, there are a number of strands of gigabits/sec fiber feeding the cable head. Just by switching on the unused fibers feeding that head, more cable bandwidth can instantly be delivered, either to existing users, or to new subscribers.

Regardless, will the cable companies be spending all these billions on the right stuff? For not to be overlooked is the huge fault line running between the Internet Engineering Task Force (IETF) and the cable companies. The IETF comes from a UNIX culture, with a long tradition of hacking away at something new until it's ready for prime time. The business-oriented cable companies, however, are all about time to market: do it now, or become lunch. This huge difference in implementation styles has lead to a seismometer-busting, techno-culture quake. Furthermore, many in the IETF are fearful that the sudden deluge of millions of new cable-based users, and their high bandwidth demands could overwhelm the Internet, causing massive Net-blackouts.

So, not too surprisingly, the IETF has stated that it will be the one that makes the standards which specify the links between the Internet and the cable TV systems. Of course, the cable people have their own ideas on this interface subject. This intellectual/business schism and its associated fears have produced a potpourri of conflicting interface definition standards for carrying IP traffic between the two systems. (E.g., the IETF's IP over CATV spec; Davic's cable modem Net access spec; the Society of Cable & Telecom Engineers' proposals for security; etc.) Compounding the human communications problem is that the two camps don't seem to have too much formal contact with one another.

However, let's assume for the moment that all these standards-setting differences are finally resolved. Even so, we are still left with the same nagging ISP vs. TELCO question: Can the cable companies, who have a terrible record of service, be trusted to provide their users reliable support for critical applications, like voice? Given the deteriorating service track record of some TELCOs, like NYNEX, maybe they will all just meet in the muddled, mediocre middle. This is not a pleasant prospect.

ATM? Isn't That a Money Machine?

So what is a business to do with its mission critical Internet applications? One potential answer: the new Mega-ISPs. As we have seen, several ISPs are fast becoming big, application savvy organizations. Some are already serving large regions of the country via their own end-to-end controlled Intranets. It is quite likely that some of these regional ISPs will further combine to create national Intranets. Even world-wide ISP Intranets are now possible. E.g., AT&T-Unisource Services will gladly supply their new global ATM nets to ISPs on a subcontract basis.

ATM (Asynchronous Transfer Mode) is a technology that will pave the way for this emerging class of big Internet Service Providers . In the OSI stack, ATM, and its fixed size cell of 53 bytes, sit at layer two (datalink). This is below IP, which is at layer three (network). ATM is a switched technology, while IP uses a routing scheme. But an ATM backbone can support IP, as well as frame relay, end points. Inter-networking is also possible, such as ATM to IP, or ATM to frame relay. The main advantage of ATM over IP is that it inherently supports isochronous multimedia applications, such as voice. In contrast, IP has to resort to clever multimedia hacks, like RSVP. But RSVP, as we saw above, requires a level of ISP cooperation that may not be there when you need it. ATM avoids all such uncooperative ISP nonsense. For these important reasons, you can reasonably expect that by the year 2000, most TCP/IP packets will be riding across the globe on top of high speed ATM backbones.

Moreover, ATM nets can also support PBX to PBX communications, the most common means of voice transmission and call control in a company. By using ATM cells that have their own headers, the technology's CBR (constant bit rate) service can be avoided. As ATM/CBR emulates a circuit, like a T-1, it is really no better than a leased line. But with their own headers, the voice carrying ATM cells can be intermixed with other types of data carrying traffic. At the end point, the cells are synchronously reassembled. This makes ATM much more viable for PBX to PBX duties, and thus, a more practical, near universal solution for multiple kinds of communications needs. ATM is therefore still the technology of choice for knitting together a far-flung enterprise.

ATM is not a one size fits all technology. It is a class of services for prioritizing bandwidth allocation. CBR, with its guaranteed fixed rate of service, is suitable for voice, video, and, as we saw, for emulating digital circuits. Then there is VBR (variable bit rate). VBR, as its moniker implies, is intended for bursty, asynchronous traffic. Moreover, VBR can be subdivided into either real time and non-real time VBR. Real time VBR can be used for voice transmission, as the latter uses compression and noise suppression techniques. Non-real time VBR is good for transaction processing. Next, there is UBR (unspecified bit rate), which is good when you want to utilize spare, background bandwidth not being used by other ATM service classes. Finally, if you have low level bandwidth needs, with an occasional requirement for more, there is ABR (available bit rate). ABR tends to get the nod over UBR because it can monitor traffic conditions, thereby allocating extra bandwidth to needy users. ABR also allows bursty traffic to move across heavily used ATM networks. Admittedly, this ATM virtual circuit melange makes for an alphabet soup of services. It can therefore be quite mystifying to a long time manager of clustered, physical circuit LANs. However, it is this very flexibility that gives ATM such power.

For running TCP/IP applications over ATM, either ABR or UBR are used, as these Internet protocols are designed to handle cell loss. Many companies, such as 3COM, Cisco systems, Bay Networks, FORE Systems, and a host of others, sell ATM switches that support IP routing. ATM switches typically provide several bandwidth speeds, and can include support for 25 Mbs, 100 Mbs, 155 Mbs, and 622 Mbs ATM Nets.

But does all this flexibility also mean that your mega-ISP's ATM cells will soon be going straight into the desktop PC, as has long been predicted? The portents are certainly there. Start-up Avidia Systems is now offering desktop ATM for only $199 a port. Its new 25 megabit/sec switches are one fourth the cost of FORE Systems, and one third less than its other competitor, Madge Systems. This price pressure will likely cause these other two companies to make a market counter attack. So it looks like ATM is finally about to fulfill its desktop high speed LAN promise. But maybe not. ATM LAN lab test reports in various journals, e.g., LANTimes, March 3, 1997, are saying that using 155Mbs ATM on servers and high performance workstations is currently a disaster, with not one ATM NIC (network interface card) vendor offering a satisfactory solution. It appears that 100Mbs Fast Ethernet is still far and away the best and most cost effective solution for high performance LANs. The best current advice is to use ATM on the backbone, and forget about using it on a desktop/server LAN.

But even on the backbone, ATM is in for a fight, as Gigabit Ethernet is now here. This new technology takes a familiar, well know connectivity system and makes it go hypersonic. Companies such as 3Com, IBM, Cisco, Bay Networks, IBM, and even FORE, have all pitched their hats into the switched Gigabit Ethernet ring. But although the vendor hype may make it appear that is an ATM vs. Gigabit Ethernet war, it needn't be. Already, companies like 3Com have announced Fast Ethernet to ATM OC 12 (622 Mbit/sec) switches. The reality is that ATM will probably be on the big league ISP's backbone, and you will just have to shop around for the best deal for the fastest desktop connection, which will likey be Fast Ethernet.

The Mega-ISP; The ISP Doomsday Machine

But whether it be ATM or Gigabit Ethernet, strap on your network seatbelts, for the day of the multi-faceted, Mega-ISP has finally arrived. Many of these new Intranet service organizations will successfully challenge both the TELCOs and the TV Cable companies for your communications heavy lifting. And the 1996 Telecommunications Reform Act seals the Mega-ISP deal.

But to make this Mega-ISP gambit work, the user corporation has to act like its own FCC, State Dept. of Public Utilities, etc., all rolled into one. It must draw up a rigid set of guidelines for monitoring its new Mega-ISP's behavior. The user-Provider contracts should include mandatory reporting on ISP traffic statistics, ISP data traffic capabilities, type of ISP equipment installed, software used, ISP systems maintenance and upgrade schedules, etc. Finally, the user must put in the appropriate financial penalties if the ISP drops the ball. In effect and in reality, the user corporation must make their new Mega-ISP behave just as if it were a government regulated, closely watched TELCO or RBOC.

One good example of such a tightly regulated ISP-relationship is the upcoming Automotive Network Exchange (ANX), a joint effort by General Motors, Ford, and Chrysler, to establish a user-managed Intranet. The goals of the ANX are to let the respective companies' business partners efficiently, securely, and predictably share sensitive data, such as competitive bids, delivery dates, E mail, CAD/CAM files, invoices, etc. ANX security will be based on the IETF's IP security standard. A number of tactics are also under consideration to allow smaller companies who do not have firewalled Intranets to participate in the ANX. Also under review are ways to enable the companies' off-shore partners to participate in the ANX security procedures (export of security algorithms, like DES, are still a bone of government/industry contention).

The ANX will utilize a select number of ISPs that have been rigorously screened, and all must meet the Big Three's high standards for IP performance and reliability. In addition, all the participating ISPs have to submit to being supervised by the ANX Overseer, a separate company (and said likely to be Bellcore.) Like the FCC, or a state regulator, the ANX Overseer has the authority to monitor, troubleshoot, and publish performance reports on the participating ISPs. Metrics such as packet loss, latency, link utilization, network congestion, and other such performance issues--like the ISP's procedures for rapid disaster recovery--will all be carefully scrutinized by the Overseer. Presumably, the ANX Overseer will also have the means to enforce its will on the contracted ISPs, should they not measure up to the contracted standard.

Such ISP Intranet contracts, like the ANX, will not be easy to negotiate. And there will be lots of growing pains and difficulties. So what else is new? Nothing much, except that now you will never look at your friendly ISP the same way again--for good, or for bad. Of course, in the upcoming network salad days, the RBOCs, TELCOs, and cable companies will only see the bad.

And probably, that's all to the good. Unless the Microsoft Monopoly ends up taking over the Internet from all of them. Which is probably bad.

Good? Bad? You decide, the next time you get your ISP's busy signal.

Copyright 1997, Francis Vale, All Rights Reserved

21st, The VXM Network,