Skip to main content
SpectralShifts Blog 
Sunday, March 11 2012

I first started using clouds in my presentations in 1990 to illustrate Metcalfe’s Law and how data would scale and supersede voice.  John McQuillan and his Next Gen Networks (NGN) conferences were my inspiration and source.  In the mid-2000s I used them to illustrate the potential for a world of unlimited demand ecosystems: commercial, consumer, social, financial, etc…  Cloud computing has now become a part of everyday vernacular.  The problem is that for cloud computing to expand the world of networks needs to go flat, or horizontal, as in this complex looking illustration to the left.

This is a static view.  Add some temporality and rapidly shifting supply/demand dynamics and the debate begins as to whether the system should be centralized or decentralized.  Yes and no.  There are 3 main network types:  hierarchical, centralized and fully distributed (aka peer to peer).  None fully accommodate metcalfe’s, moore’s and zipf’s laws.  Network theory needs to capture the dynamic of new service/technology introduction that initially is used by a small group, but then rapidly scales to many.  Processing/intelligence initially must be centralized but then traffic and signaling volumes dictate pushing the intelligence to the edge.  The illustration to the right begins to convey that lateral motion in a flat, layered architecture, driven by the 2-way, synchronous nature of traffic; albeit with the signalling and transactions moving vertically up and down.

But just as solutions begin to scale, a new service is borne superseding the original.  This chaotic view from the outside looks like an organism in constant state of expansion then collapse, expansion then collapse, etc…

A new network theory that controls and accounts for this constant state of creative destruction* is Centralized Hierarchical Networks (CHNs) CC.   A search on google and duckduckgo reveals no known prior attribution, so Information Velocity Partners, LLC (aka IVP Capital, LLC) both lays claim and offers up the term under creative commons (CC).  I actually coined the CHN term in 2004 at a Telcordia symposium; now an Ericsson subsidiary.

CHN theory fully explains the movement from mainframe to PC to cloud.  It explains the growth of switches, routers and data centers in networks over time.  And it should be used as a model to explain how optical computing/storage in the core, fiber and MIMO transmission and cognitive radios at the edge get introduced and scaled.  Mobile broadband and 7x24 access /syncing by smartphones are already beginning to reveal the pressures on a vertically integrated world and the need to evolve business models and strategies to centralized hierarchical networking.

*--interesting to note that creative destruction was original used in far-left Marxist doctrine in the 1840s but was subsumed into and became associated with far-right Austrian School economic theory in the 1950s.  Which underscores my view that often little difference lies between far-left and far-right in a continuous circular political/economic spectrum.

Related Reading:
Decentralizing the Cloud.  Not exactly IMO.

Network resources will always be heterogeneous.

Everything gets pushed to the edge in this perspective

Posted by: Michael Elling AT 08:42 am   |  Permalink   |  0 Comments  |  Email
Sunday, November 13 2011

A humble networking protocol 10 years ago, packet based Ethernet (invented at Xerox in 1973) has now ascended to the top of the carrier networking pyramid over traditional voice circuit (time) protocols due to the growth in data networks (storage and application connectivity) and 3G wireless.  According to AboveNet the top 3 CIO priorities are cloud computing, virtualization and mobile, up from spots 16, 3 and 12, respectively, just 2 years ago!   Ethernet now accounts for 36% of all access, larger than any other single legacy technology, up from nothing 10 years ago when the Metro Ethernet Forum was established.  With Gigabit and Terabit speeds, Ethernet is the only protocol for the future.

The recent Ethernet Expo 2011 in NYC underscored the trends and importance of what is going on in the market.  Just like fiber and high-capacity wireless (MIMO) in the physical layer (aka layer 1), Ethernet has significant price/performance advantages in transport networks (aka layer 2).  This graphic illustrates why it has spread through the landscape so rapidly from LAN to MAN to WAN.   With 75% of US business buildings lacking access to fiber, EoC will be the preferred access solution.  As bandwidth demand increases, Ethernet has a 5-10x price/performance advantage over legacy equipment.

Ethernet is getting smarter via a pejoratively coined term, SPIT (Service Provider Information Technology).  The graphic below shows how the growing horizontalization is supported by vertical integration of information (ie exchanges) that will make Ethernet truly “on-demand”.  This model is critical because of both the variability and dispersion of traffic brought on by both mobility and cloud computing.  Already, the underlying layers are being “re”-developed by companies like AlliedFiber who are building new WAN fiber with interconnection points every 60 miles.  It will all be ethernet.  Ultimately, app providers may centralize intelligence at these points, just like Akamai pushed content storage towards the edge of the network for Web 1.0.  At the core and key boundary points Ethernet Exchanges will begin to develop.  Right now network connections are mostly private and there is significant debate as to whether there will be carrier exchanges.  The reality is that there will be exchanges in the future; and not just horizontal but vertical as well to facilitate new service creation and a far larger range of on-demand bandwidth solutions.

By the way, I found this “old” (circa 2005) chart from the MEF illustrating what and where Ethernet is in the network stack.  It is consistent with my own definition of web 1.0 as a 4 layer stack.  Replace layer 4 with clouds and mobile and you get the sense for how much greater complexity there is today.  When you compare it to the above charts you see how far Ethernet has evolved in a very rapid time and why companies like Telx, Equinix (8.6x cash flow), Neutral Tandem (3.5x cash flow) will be interesting to watch, as well as larger carriers like Megapath and AboveNet (8.2x cash flow).   Certainly the next 3-5 years will see significant growth in ethernet and obsolescence of the PSTN and legacy voice (time-based) technologies.

Related Reading:
CoreSite and other data centers connect directly to Amazon AWS

Equinix and Neutral Tandem provide seamless service

 

Posted by: Michael Elling AT 12:46 pm   |  Permalink   |  0 Comments  |  Email
Sunday, October 30 2011

Without access does the cloud exist?  Not really.

In 2006, cloud computing entered the collective intelligence in the form of Amazon Web Services.  By 2007, over 330,000 developers were registered on the platform.  This rapid uptake was an outgrowth of web 1.0 applications (scale) and growth in high-speed, broadband access from 1998-2005 (ubiquity).  It became apparent to all that new solutions could be developed and efficiencies improved by collapsing to the core a portion of processing and storage that had developed at the edge during the WinTel revolution.  The latter had fundamentally changed the IT landscape between the late 1980s and early 2000s from a mainframe to client server paradigm.

In late 2007 the iPhone was born, just as 3G digital services were introduced by a competitive US wireless industry.  In 2009 “smartphone” penetration was 18% of the market.  By the 3rd quarter of 2011 that number reached 44%.  The way people communicate and consume information is changing dramatically in a very short time. 

The smartphone is driving cloud (aka back to the mainframe) adoption for 3 reasons: 1) it is introducing a new computing device to complement, not replace, existing computing devices at home and work; 2) the small screen limits what information can be shown and processed; 3) it is increasing the sociability, velocity and value of information.   Information knows no bounds at the edge or core.  And we are at the very very early stages of this dramatic new revolution.

Ice Cream Sandwich (just like Windows 2.0 multi-tasking in 1987) heralds a radical new world of information generation and consumption.  Growth in processing and computation at the edge will drive the core and vice versa; just as chip advances from Intel fed software bloat on desktops further necessitating faster chips.   

But the process can only expand if the networks are there (see page 2) to support that.  Unfortunately carriers have responded with data caps and bemoan the lack of new spectrum.  Fortunately, a hidden back door exists in the form of WiFi access.  And if carriers like AT&T and Verizon don’t watch out, it will become the preferred form of access.

As a recent adopter of Google Music I have become very attuned to that.  First, it is truly amazing how seamless content storage and playback has become.  Second, I learned how to program my phone to always hunt for a wifi connection.  Third, when I do not have access to either the 3G wireless network or WiFi and I want something that is stored online a strange feeling of being disconnected overtakes me; akin to leaving one’s cellphone at home in the morning.

With the smartphone we are getting used to choice and instant gratification.  The problem with WiFi is it’s variability and unreliability.  Capital and technology is being applied to solve that problem and it will be interesting to see how service providers react to the potential threat (and/or opportunity).  Where carriers once imagined walled application gardens there are now fertile iOS and Android fields watered by clouds over which carriers exert little control.  Storm clouds loom over their control of and ROI from access networks.

Posted by: Michael Elling AT 09:10 am   |  Permalink   |  0 Comments  |  Email
Sunday, April 24 2011

A couple of themes were prevalent this past week:

  • iPhone/Android location logging,
  • cloud computing (and a big cloud collapse at Amazon),
  • the tech valuation bubble because of Groupon et al,
  • profits at Apple, AT&T vs VZ, Google, most notably,
  • and who wins in social media and what is next.

In my opinion they are all related and the Cloud plays the central role, metaphorically and physically.  Horowitz recently wrote about the new computing paradigm in defense of the supposed technology valuation bubble.  I agree wholeheartedly with his assessment as I got my first taste of this historical computing cycle over 30 years ago when I had to cycle 10 miles to a High School in another district that had a dedicated line to the county mainframe.  A year or two later I was simulating virus growth on an Apple PC.  So when Windows came in 1987 I was already ahead of the curve with respect to distributed computing.  Moreover, as a communications analyst in the early 1990s I also realized what competition in the WAN post-1984 had begat, namely, Web 1.0 (aka the Internet) and the most advanced and cheapest digital paging/messaging services in the world.  Both of these trends would have a significant impact on me personally and professionally and I will write about those evolutions and collapses in future Spectral issues.

The problem, the solution, the problem, the solution, etc….

The problem back in the 1970s and early 1980s was the telephone monopoly.  Moore’s law bypassed the analog access bottleneck with cheap processing and local transport.  Consumers and then enterprises and institutions began to buy and link the PCs together to communicate, share files and resources.   Things got exciting when we began to multitask in 1987, and then by 1994 any PC provided access to information pretty much anywhere.  During the 1990s and well into the next decade, Web 1.0 was just a 1.2-way store and forward database lookup platform.  It was early cloud computing, sort of, but no-one had high-speed access.  It was so bad in 1998 when I went independent, that I had 25x more dedicated bandwidth than my former colleagues at bulge-bracket Wall Street firms.  That’s why we had the bust.  Web 1.0 was narrow-band, not broadband, and certainly not 2-way.  Wireless was just beginning to wake up to data, even though Jeff Bezos had everyone believing they would be ordering books through their phones in 2000.

Two things happened in the 2000s.  First, high speed bandwidth became ubiquitous.  I remember raising capital for The Feedroom, a leading video ASP, in 2003 and we were still watching high-speed access penetration reaching the 40% “tipping point.”.  Second the IP stack grew from being a 4 layer model to something more robust.  We built CDNs.  We built border controllers that enabled Skype VoIP traffic to transit foreign networks “for free.”  We built security.  HTML, browsers and web frontends grew to support multimedia.  By the second half of the decade, Web 2.0 became 1.7-way and true “cloud” services began to develop.  Web 2.0 is still not fully developed as there are still a lot of technical and pricing controls and “lubricants” missing for true 2-way synchronous high-definition communications; more about that in future Spectrals.

The New “Hidden Problem”

Unfortunately, over that time the underlying service provider market of 5-6 competitive service providers (wired, wireless, cable) consolidated down to an oligopoly in most markets.  Wherever competition dropped to 3 or fewer providers bandwidth pricing stopped falling 40-70% like it should have and only fell 5-15% per annum.  Yet technology prices at the edge and core (Moore’s Law) kept on falling 50%+ every 12-18 months.  Today, the price differential between “retail” and “underlying economic” cost per bit is the widest it has been since 1984.

That wouldn’t be a problem except for two recent developments:  the advent of the smartphone and the attendant application ecosystems.  So what does this have to do with cloud computing, especially when that was “an enterprise phenomenon” begun by Salesforce.com with its Force.com and Amazon Web Services.  A lot of the new consumer wireless applications run on the cloud.  There are entire developer ecosystems building new companies.  IDC estimates that the total amount of information accessible is going to grow 44x by 2020 to 35 zetabytes.  And the average number of unique files is going to grow 65x.  That means that while a lot of the applications and information is going to be high-bandwidth (video and multimedia), there are also going to be many smaller files and transactions (bits of information); ie telemetry or personal information or sensory inputs.  And this information will be constantly accessed by 3-5 billion wireless smartphones and devices.  The math of networks is (N*(N-1))/2.  That’s an awful lot of IP session pathways.

Why is That A Problem?

The problem is that the current wireless networks can’t handle this onslaught.  Carriers have already been announcing datacaps over the past 2 years.  While they are falling over themselves to announce 4G networks, the reality is that they are only designed to be a 2-3x faster, and far from being ubiquitous, either geographically (wide-area) or inbuilding.  That’s a problem if the new applications and information sets require networks that are 20-50x faster and many factors more reliable and ubiquitous.  The smartphones and their wireless tether are becoming single points of access.  Add to that the fact that carriers derive increasingly less direct benefit from these application ecosystems, so they’ll have less and less incentive to upgrade and reprice their network services along true technology-driven marginal cost.  Neustar is already warning carriers they are being bypassed in the process.

Does The Bubble Have to Burst?

Just as in the late 1990s, the upper and middle layer guys really don’t know what is going on at the lower layers.  And if they don’t then surely the current bubble will burst as expectations will get ahead of reality.  That may take another 2-3 years, but it will likely happen.  In the meantime, alternative access players are beginning to rise up.  Even the carriers themselves are talking about offloading traffic onto femto and wifi cells.  Wifi alliances are springing up again and middle layer software/application controls are developing to make it easier for end-users to offload traffic themselves.  Having lived through and analyzed the advent of competitive wired and wireless networks in the 1990s, my sense is that nothing, even LightSquared or Clearwire in their current forms, will be significant enough to precipitate the dramatic restructuring that is necessary to service this coming tidal wave of demand.

What we need is something that I call centralized hierarchical networking (CHN)™.  Essentially we will see three major layers with the bottom access/transport layer being controlled by 3-4 hybrid networks.  The growth and dynamic from edge to core and vice versa will wax and wane in rather rapid fashion.  Until then, while I totally get and support the cloud and believe most applications are going that route, let the Cloud Players be forewarned of coming turbulence unless something is done to (re)solve the bandwidth bottleneck!

Posted by: Michael Elling AT 09:34 am   |  Permalink   |  0 Comments  |  Email
Email
Twitter
Facebook
Digg
LinkedIn
Delicious
StumbleUpon
Add to favorites

Information Velocity Partners, LLC
88 East Main Street, Suite 209
Mendham, NJ 07930
Phone: 973-222-0759
Email:
contact@ivpcapital.com

Mastodon

Design Your Own Website, Today!
iBuilt Design Software
Give it a try for Free