For a number of years now, industry watchers (including myself) have tracked, commented on, and promoted the “IP-Over-Everything and Everything-Over-IP” theory of technology evolution.  That is, the proposition of IP becoming the unifying technology for all communications over all kinds of media.  And it’s hard to argue the “IP-Over-Everything” part of the mantra.  IP does run over almost everything.  We have IP-over-Ethernet, IP-over-FDDI, IP-over-ATM, Packet-over -SONET, Resilient Packet Rings (RPR), IP-over 802.11, IP-over-cellular of various and sundry forms, IP-over-optics, and in the government sector, IP-over-budget.  IP has proven to be technically flexible.  It’s hard to find something you can’t run IP reasonably well over.  IP has also always been an open and ever-evolving protocol, which gave it the market edge it needed over competing Layer 3 protocols and eventually resulted in enough critical mass to totally squash them.  As a result, connectivity of all kinds of devices for non time-sensitive packet-based applications have soared. 

In the case of IP-Over-Everything, it was inevitable one common set of underlying protocols would be adopted to enable all-packet networking innovation to focus on a single common environment.   However, the investment and technology communities have been approaching Everything-Over-IP as if there were a similar inevitability.  For the past few years, we’ve all heard a lot about Voice-over-IP and Video/Audio streaming over-IP.  More recently we’ve been hearing a lot about storage protocols like SCSI and Fiber channel over IP.   Now we’re starting to hear about Remote DMA over IP (iWARP).  But how much of these Everything-Over-IP examples are really being deployed on a production basis these days?  In the case of Voice-over-IP, some, but certainly not critical mass.  In the case of streaming video-over-IP, very little.  In the case of iSCSI and FC-IP, almost none.  Even many applications inherently bursty and packet-based haven’t been converted to IP yet.  Somewhat surprisingly to us, we just ran across a huge project, just now in the planning stages, to convert a large DECNET environment to TCP/IP.   I had thought migrations like that had been completed years ago. 

Why is this? Certainly there are technical reasons.   IP is at its best when the applications it supports don’t care all that much when packets are delivered and how they get there. E-mail, file downloads, and simple content publishing all share these characteristics.  Indeed, applications of this nature dominate the traffic over packet networks today.   But when it absolutely, positively has to get there at a certain, precise time, some characteristics of IP and its sidekick TCP turn from advantages to handicaps.  Thus much innovation and development has to be executed to bring IP kicking and screaming into these traditionally non-packet oriented environments.  This takes time.  And any time a lot of innovation is needed, different smart minds will approach it in different smart ways, and it takes the world even more time to sort through the options and figure out what the best combination of approaches are. 

But more than the technical reasons, I think the reason why Everything-Over-IP hasn’t generally taken off yet is the lack of a compelling reason.  Everything-Over-IP solutions can be fairly inexpensive, but not relative to equipment you’ve already bought and have been using for years.   As with any technology migration, much of the expense is the up-front cost related to changing directions, both in capital and human resources.  Often, it’s hard to justify taking on that up-front cost to achieve a return that is mostly far in the future.  You can always find a reason to put it off until next year…or next…or next. 

And let’s face it, none of the Everything-Over-IP initiatives defined to date actually bring any significant new services to end-users.  I can make phone calls without VoIP.  I can even coordinate applications on my computer screen with my voice call without VoIP.  I can watch any movies I want to at home with just a little planning without Video-over-IP.  I can store data and move big chunks of it from one site to another without iSCSI.  I may need four wires instead of two to do many of these things, but guess what – I already have eight. 

So, those start-ups that want to make money in an Everything-Over-IP space, and those established companies who need to convince Wall Street their future market opportunity is much larger than their past market opportunity because of the proliferation of Everything-Over-IP, need to face up to the fact that it’s not inevitable.  IP has to earn its right to transport each and every new application conceived for it. And if the resultant application implementation doesn’t provide some significant new way of doing business or personal interaction, then it better have a super compelling economic advantage - now, not in some hypothetical future time.   One or more of these characteristics is required to accelerate adoption. 

Right now, I don’t see any of the Everything-Over-IP initiatives meeting the requirements for accelerated adoption. So I think it’s going to be a long hard road for all of them.  With this in mind, I have decided to devote my next few Musings to market trends that are actually happening – VPNs and variations on the theme, XML proliferation, new web site acceleration techniques, and (surprisingly) Managed Services.  These trends may not be as sexy, but real people are spending real money and companies positioned well are making real revenue. In the second half of the year the Musings will look at some of the Everything-Over-IP initiatives to separate the ones I believe have a future as long as they have enough stamina, vs. the ones I feel are doomed.

(volume 5, number 1)

Send email to info@acuitive.com with questions or comments about this web site.
Copyright ©1997-2002 Acuitive, Inc.  All Rights Reserved