Posts Tagged ‘wireless’

“Net Neutrality” does not mean “equally fast, everywhere”

August 22, 2010

“Net Neutrality” means too many different things to different people.

For example, concerns about the potential for discriminatory practices by ISPs or telcos based on content, application, or identities or affiliations of content consumers or producers are usually conflated with simplistic observations about the need for ‘equal speed’ being a necessary condition for net neutrality.

Al Franken discusses issues and concerns related to competition (or sufficient lack thereof), but he too raises the ‘speed equality’ notion as a requirement for ‘free speech’ on the internet.

It is safe to say that 99% or more of the public does not understand how the Internet works, or for that matter how computers attached to the Internet work. Discussing legitimate concerns about non-discriminatory processing of Internet traffic in simple terms of speed only further confuses the public, and create political responses for the wrong reasons.

What we all agree on is NOT the issue

What is not to be tolerated in any scenario, ‘network neutrality’ or not, is discrimination based on type of protocol, or content, or application, or content provider, or consumer.

It is this type of discrimination which has vociferous proponents of ‘net neutrality’ most up in arms, and yet it is not the type of discrimination that is likely to ultimately be the real net neutrality issue, for the simple reason that it is and will be easy for everyone to agree that those forms of discrimination are inappropriate, anti-competitive, and, yes, illegal.

The Need for Speed

Unfortunately, it is “speed” discrimination on which simplistic overtures to net neutrality are based.

For example, the following from the recent op-ed  by Al Franken on the subject:

An e-mail from your mom comes in just as fast as a bill notification from your bank. You’re reading this op-ed online; it’ll load just as fast as a blog post criticizing it. That’s what we mean by net neutrality.”

This definition of ‘net neutrality’ — that every interaction is “just as fast” as any other, is the most dangerously misleading of all attempts to define (and, in Senator Franken’s case, legislate) ‘net neutrality’.

If, by his comments, Senator Franken means to say that “the rate at which bytes are transmitted over the network, for consumers sharing the same level of cost for the same level of quality of service, should be the same”, then yes — (that is, definition #3 of ‘net neutrality’) — and is perfectly reasonable.

But the imprecise language of the appeal to ‘equally fast’ will incorrectly lead people to believe that net neutrality is intended to make the time it takes to download Avatar equal to the amount of time it takes to send a tweet, with the further stipulation that this be via terrestrial or wireless connections, and all for the same cost as the least demanding of levels of service.

Where “Net Neutrality” really applies is in the requirement to not discriminate based on content. This includes, of course, any selective slowing down of traffic based on application or protocol from sources who have paid the same for the same bandwidth. Local ISPs (notably cable companies with their claims of 5MB/sec and more bandwidth) need to be able to support those promises, or else not promise so much — truth in advertising, quite simply.

The “15 Facts” infographic

ReadWriteWeb has posted an ‘infographic’ entitled “15 Facts About Net Neutrality“.

The 15-point summary covers ‘bullet points’, but does not provide sufficient insight into the not-so-obvious distinctions among definitions of ‘net neutrality’.

In particular, the ‘3 definitions of net neutrality’ makes a stab at this, but more attention needs to be focused on specifically and exactly what is being talked about when different folks debate ‘net neutrality’.

To recap, the ‘3 definitions of Net Neutrality’ provided by the Online MBA Programs folks are:

1. Absolute non-discrimination:

No regard for quality of service considerations

2. Limited discrimination without QoS tiering

Quality of service discrimination allowed as long as no special fee is charged for higher quality service

3. Limited discrimination and tiering

Higher fees for quality of service provided there is no exclusivity in contracts

The definitional hole in the above summary points is simply: what is the domain potentially being discriminated? Is it content type ? application ? protocol type ? bandwidth ? consumer ? content provider ?

Definition 1 is, for the most part, the hue and cry of the status-quo, and is bound to run into trouble with consumers at some point.

This will occur when the volume of internet traffic of a (growing) minority of highly active internet consumers reaches thresholds where it taxes the bandwidth and capacity of Internet service to the point that a majority of less-active consumers are noticeably and negatively affected.

For example, when a sufficient volume of constant bit torrent traffic — or a sufficient increase in the amount of video-on-demand being streamed — reaches a level where local delivery is saturated or visibly impacted, consumers will notice (and ISPs and telcos will seek to maximize the number of satisfied, paying customers). This is what started the whole issue, after all !

Definition 2 would appear to be the worst of all worlds. “no special fee … charged for higher quality service” would of course never occur in a bandwidth-challenged network. Quite the opposite – that definition is absolutely equivalent to: “same fee charged for poorer quality service” ! It would let ISPs degrade service (or simply let service degrade on its own). This is the least desirable scenario.

Definition 3, which acknowledges that higher quality of service (translation: continuously available higher bandwidth) is something which consumes more of the capital infrastructure resources of the supply chain (ISP and backbone) — and for which a higher cost is appropriate — would seem to be a reasonable starting point for network neutrality that guarantees equal access within a given quality of service.

Unfortunately, this definition is one which is also drawing considerable flak over the concern that it will create two classes of internet users: “rich” and “poor”, or “fast lanes” and “slow lanes”  (See the transcript of Cali Lewis’ interview on CNN, for example).

Until the language and conversation regarding ‘net neutrality’ is cleaned up and made more precise, the controversy will continue to swirl unproductively. And can you imagine the provisions or effect that Congressional legislation — in the absence of such specificity, and in the presence of such fuzzy emotional appeals — will have ? It certainly will not have the desired effect !

If, for example, a clumsy, and poorly thought-out, knee-jerk legislative reaction results — one which is too restrictive or onerous — then the result could be one where growth and competition to provide higher bandwidth service is replaced by a strategy on the part of ISPs and telcos to simply start charging more by the data byte with minimal investment in technical infrastructure. This outcome would ultimately cost all consumers more, and for less service !

The Fuzzy Semantics of “Net Neutrality”

August 14, 2010

For all the legitimate and significant issues regarding the future evolution of the Internet which the term “Net Neutrality” is meant to embody, the phrase would appear to have become a bit of a conceptual chameleon with a life of its own, and whose meaning is not particularly clear — certainly not to many people when they first encounter the notion.

Despite the fact that the events and practices which instigated the initial debate are fairly specific,  the term has become the distilled cornerstone for a veritable uprising against telcos and ISPs, carrying with it a cachet which is nearly on the level of “freedom of speech” and “human rights”.

When combined with the simple lead-in of “those in support of …”, the term “net neutrality” suggests an air of impeccability, its own “self-evident truth” and standard bearer for a cause which all intelligent and righteous beings should without question support.  For that catchy spin alone, whoever coined the term deserves an advertising Cleo award.

But the fuzzy semantics of the term “net neutrality” can be misleading and lead to confusion or distortions in attempting to analyze or discuss the multi-faceted and non-trivial issues and trade-offs which are teeming immediately beneath the self-assured surface of that simple label.

A case in point

For example, in a recent TechCrunch column, MG Siegler writes:

“For net neutrality to truly work, we need things to be black and white. Or really just white. The Internet needs to flow the same no matter what type of data, what company, or what service it involves. End of story.”

Now that is quite a remarkable statement, if you think about it.  With apologies to Mr. Siegler for selecting his statement as the guinea pig in this lab experiment, let’s analyze the statement a bit more closely.

In a very short space of words, this statement contains such phrases as “truly work”, “we need things”,”black and white”,”just white”, and “flow the same”.

Ignoring for a moment the absence of a clear description of just what “net neutrality” is that we should be trying to make “truly work” (the assumption is that we all know, of course), the statement that follows which seeks to define just what is required for that to happen is remarkably vague and imprecise. In fact, it is more emotional than logical, and  — in tone, at least — resembles the kind of pep talk a coach might give to his team before they take the field in a sporting contest.

In fairness, “we need things to be black and white” alludes to the fact, which no one would deny,  that the rules, policies, practices and terms of service of ISPs vis-a-vis their customers should be just that — transparent, clearly documented and understood, i.e., black and white.

The transition to “or really just white” is to say that these rules and policies should not create discriminatory distinctions among customers or types of service, bifurcating them and their respective internet access into two classes: “legacy” and “future”, rich or poor, first class or coach, i.e., black or white.  That is a clever segue and turn of phrase.

Caution: thin ice ahead

Where Mr. Siegler runs into some significant thin ice is when he follows with “The Internet needs to flow the same no matter what type of data, what company, or what service it involves.”

This statement may well appear to most people to be an adequate, if not altogether necessary and sufficient, definition of “net neutrality” — but perhaps therein lies the rub with the truly fuzzy semantics of the term.

To be clear, there is no quibble with the “no matter … what company” part of the statement. The most compelling part of the criticisms leveled at ISPs has to do with their documented use of ‘discretionary policies’ to ‘manage’ Internet traffic and the natural concerns over the potential abuse by ISPs of the inherent power and leverage which comes with being in the position of being the gatekeepers of the Internet.

Discriminatory practices by ISPs, including preferential treatment of customers or partners, whether for commercial or even political purposes, would clearly be improper, and have a negative effect on the innovation and economic growth which the Internet fosters and enables, particularly when the ISPs involved are “size huge” and affect millions and millions of customers.

Go with the flow … but fasten your seatbelts

Rather, it is in the declaration that the “Internet needs to flow the same no matter what type of data … or what service it involves” where major turbulence in the “net neutrality” debate is encountered, requiring that we fasten our seatbelts for the duration of the flight, as it were.

It is at this point in our journey down the rabbit hole where the matter of ‘net neutrality’ becomes  more turgid, yet multi-faceted, and without clear solutions — not ‘black and white’ ones, and certainly not ‘just white’ ones.

The phrase “flow the same” is where the devil indeed lurks in the details, cloaked in that larger statement which, for all intents and purposes, could well have been crafted by tireless opponents of discrimination in the U.S. Civil Rights movement.

What exactly does “flow the same” mean, and what are the implications of any of its possible interpretations ?

Picky, Picky

Before proceeding, you may be asking “why so picky”. Rest assured, I am not defending, nor am I a shill, for telcos and ISPs in this debate ! And I am just as concerned as the next person about establishing any precedents or slippery slopes that would lead to the bifurcation and balkanization of the Internet.

But to demand that the “Internet flow the same” regardless of data or service is not only completely ambiguous, it is also a notion with inherent contradictions along multiple dimensions.

In an effort to explore the problems that the fuzzy semantics around this particular notion entail, let’s do some “thought experiments”.  Lab coats and protective eye goggles ready ?

Thought Experiment

In this “thought experiment” , let’s take two scenarios, or snapshots, if you will, of two possible alternative internet universes.

Alternate Internet universe 1.A

Alternate Internet universe 1.A (AIU 1.A) consists of one thousand servers and one billion clients. The principle service in this alternate universe consists of sending and receiving short messages, called “Burps”.

  • Clients poll all 1000 servers for Burps addressed to them, evenly spread out over the course of  repeated 60-second intervals.
  • When a client contacts a server in the course of this polling, a client can retrieve all Burps in their Burp inboxes, respond to or acknowledge Burps retrieved in previous polls,  and also post, or deposit, one new Burp of their choosing at each server.
  • Finally, Burps are limited to 100 bytes.

Let’s analyze the total “Internet flow”, in terms of both bytes and bandwidth, of this alternate Internet universe across the entire network, shall we ?

The maximum utilization of Internet 1.A occurs when every client posts one new Burp at each of the 1000 servers every 60 seconds, and responds (with a reply Burp) to all Burps retrieved from any and every server at the time of the poll of each server 60 seconds previously.

Remember, the rules of this universe specify that clients can only deposit one Burp per server when they check that server for queued Burps every 60 seconds.

Even though some clients may retrieve many Burps from their inBurp queue, others clients may have none to retrieve.

The steady-state bandwidth per second of universe 1.A in maximum utilization (calculating the total amount of information processed by the network in a 60-second cycle with all clients performing to maximum capacity, and then dividing by 60) in bytes/sec is therefore

[ #clients X  #servers X  bytes/burp X  3 ]  / 60

The factor ‘3’ above is due to the fact the number of burps per client per cycle can be analyzed thusly:

  • each client can respond to each retrieved burp (1) with a response burp (2) as well as post a new burp (3), and
  • because every client is constrained to posting only one burp per server per polling cycle, the statistical aggregate amount of burps which flow through the Internet of alterrnate universe 1.A is steady-state bounded and enumerable, regardless of the distribution of different number of burps in the queues of different clients.

Plugging in the numbers in the expression above, we get:

[ 1 billion X  1 thousand X  one hundred X  3 ]  / 60, or

5  x  10 ^ 12 bytes/sec   =   40 x 10 ^ 12 bits/sec

Alternate Internet universe 1.B

Alternate Internet universe 1.B has the same number of clients and servers as AIU 1.A.  However, the type of data and service provided in AIU 1.B is different than that of AIU 1.A. Alternate Internet universe 1.B provides realtime Channel-Multiplexed Multimedia Stream (CMMS) sharing among all clients.

  • Each channel-multiplexed multimedia stream consists of any number of different individual media streams on different virtual channels of the single multiplexed stream, as long as the total allowed bandwidth of the  a channel-multiplexed media stream is not exceeded.
  • The individual media channel streams are multiplexed together into a single CMMS solely in order to simplify the transport-level management  and routing of media channel stream sharing between client and server nodes across the network.
  • Each client can stream exactly one CMMS to each server, and can choose to subscribe to up to 1000 multimedia channel streams across the population of servers.
  • Clients can choose to subscribe to more than one CMMS to a single server.
  • Furthermore, the up to 1000 CMMS which clients can stream to each server can all be unique.
  • The maximum bandwidth of a single CMMS, fully compressed after being multiplexed and ready for transmission, is 1 Gigabit/sec

What is the information bandwidth of AIU 1.B at maximum utilization?

This is fairly easy to calculate, because, when each client node operates at capacity, 1000 CMMS streams are originated, and another 1000 are subscribed to and received.  If we furthermore assume that every CMMS is unique and remove systemic opportunities for multicasting or redundancy optimization, we arrive at

#clients X  CMMS-maxbandwidth X ( #maxstreams-out + #maxstreams-in )


1 billion X 1 Gigabit/sec X  2000 =   2 X 10 ^ 27  bits/sec

Kind of Blue (i.e., “So What”)

Each of these alternate Internet universes have the exact same configuration of clients and servers — and one could have been instantaneously cloned from the other (theoretically speaking) by some  sleep-deprived string-theorist-in-the-sky who fell asleep and whose head lurched forward and accidentally hit the ‘Big Universe Morph Process (BUMP)’ that-was-easy button in the middle of a DNS root node upgrade …

If we compare the total system bandwidth of Alternate Internet universe 1.A with that of Alternate Internate universe 1.B, the system bandwidth of  AIU 1.B is  5  x  10 ^ 13 times greater than that of AIU 1.A.

That is not merely 10 ^ 13 more data in the network (a pittance).  No, it is a 13 orders of magnitude (powers of 10) difference in BANDWIDTH .  AIU 1.B is has over 10 Billion times the bandwidth of AIU 1.A, or, put another way, it would take 10 Billion  AIU 1.A entire networks to equal a single AIU 1.B network.

Miles and Miles to go before we sleep

The two Alternate Internets have, yes, different data and services. Yet according to the definition of  the fundamental principles of ‘Net Neutrality’ so succinctly stated above, “The Internet needs to flow the same no matter what type of data …or what service it involves.”

Alternate Internet universe 1.B would of course have no trouble ‘flowing the data’ of AIU 1.A.  But obviously, if the tables were turned that would not be the case at all.

At this point in the infancy of the Internet, we are certainly farther along than 1.A, but nowhere near 1.B — perhaps more like 1.A.2.0.  And we are already mixing in a bit of 1.B ‘beta’ into the 1.A.2.0 ‘release’.

In fact, the T3 backbones look a lot like bundles of the ‘CMMS’ multiplexed media channel streams conjured up above.  And we have billions of Internet clients ‘burping’ in various flavors of social networks at the same time that businesses are offloading proprietary IT shops to “clouds”, consumer e-commerce is looking for that “next big thang” ahead of Web 3.0.

Meanwhile, the wireless data revolution is, well, kind of bringing the wireless network to its knees just as the first generation of hand-held “pre-CMMS” streamers are being introduced and fanning the flames and whetting consumer appetites and expectations with dreams of sugar plums and voice-controlled robot candy canes.

Here comes the Q.E.D.

It is perhaps the not-exactly-the-same zone of mobile wireless Internet that is the most incontrovertible evidence that the fuzzy semantic “same flow no matter what” notion of ‘Net Neutrality’ is inappropriate, and an overly simplistic star to hitch your Internet wagon to.  As a ‘thought experiment’ exercise at the end of the chapter, consider doing an analysis of pushing out the envelope of the Internet Universe in which the terrestrial zone of the Internet grows exponentially in bandwidth because of the essentially unlimited ability to simply lay more and more pipe to do so.  Terrestrial pipe has the following magical property: it is truly scalable and bandwidth-accretive, because adding another pipe does not interfere or detract from the pipes already in place.

This just in, in case you did not get the memo:  the wireless spectrum, regardless of how cleverly it is compressed, has some theoretical bandwidth limits. Why ? because, by its very nature, it is just one big “pipe in the sky” which is shared, and ultimately contested, by everyone in the geographic area within the cone or zone of the immediate local wireless cell.

When the analog voice cell network upgraded itself from 3-watt phones to 1-watt phones, way back in the oh-so-nineties, the wireless network transitioned from larger area cells to much smaller cells.  The main objective was simply to be able to increase the density of mobile subscribers per area and thereby allow a large increase in the total number of wireless subscribers system-wide.  The auction and re-purposing of electromagnetic spectrum by the FCC, making more spectrum available to the wireless industry, is a welcome, but ultimately finite, addition of bandwidth and headroom to the wireless space.  And, as we all know, for every headroom, there is a max.

All you need to do is extend out the timeline sufficiently, and you will (likely sooner than we think) reach the point where the growth in wireless bandwidth is again constrained, and we ultimately run out of road.

The only strategy which will mitigate the trend to saturation of bandwidth within a given size cell and a given fixed spectrum is the same one described above regarding the first mobile analog voice cell network.  Namely: reducing the effective size of the cell so that the exponentially increasing bandwidth of the terrestrial Internet can, by extending its tentacles and hubs into a larger number of lower-energy spaces, make more bandwidth available to a smaller number of users but in a much larger number of micro-cells.

This strategy, or one very similar to it,  will be required to continue expanding and improving the performance, reliability, availability, and bandwidth of the wireless domain for a simultaneously growing population of consumers, types of data, and services.

But to think, maintain, or imagine that the amalgamation of terrestrial and wireless interconnections to the Internet will “flow the same”, no matter what type of data or what service, is merely a recipe for ongoing angst, disappointment, and internet class warfare, in my opinion.

Chocolate or Vanilla ? or … Coppa Mista ?

Yes, we must collectively find our way, with consumers  holding the feet of both content-providers and ISPs to the fire by voting with their feet in a robust and competitive market economy.

It is the factors which impede and interfere with the power of consumers – individuals and businesses alike — to collectively and competitively shape the outcome of the next wave of the Internet’s evolution which are the real issues and dangers that  threaten the prospects on continued survival and prosperity of an innovative and open Internet.

It is not quite as ‘black and white’  (or chocolate and vanilla) as saying that the ISPs (and their purported Vichy Regime collaborators) are wearing the black hats while the rest, actively championing ‘Net Neutrality’ (whatever that really means ultimately), are wearing the white hats.

The fuzzy semantics of just what “Net Neutrality” is thought to be are not necessarily helping, either, despite the critical importance of the issue, the debate, and what happens next.