|
Impact of Content & Demand on Bandwidth Management
Addressing the source of broadband's network capacity depletion
By: Bruce Bahlmann - Contributing Author (your
feedback
is important to us!)
Created: March 15, 2002
This paper is the product of
Broadband Market Research which is available from Birds-Eye Network Services.
One of the latest trends in the world of delivering broadband data
services involves managing bandwidth (or capacity). Broadband Service
Providers (BSPs) find themselves in this position where the services they
are offering over their CATV, Fiber, DSL, Wireless, or Satellite delivery
media are beginning to consume increasing amounts of their available network
capacity. The "solution" being proposed is that if they manage bandwidth
they can determine where the traffic is coming from, what kind of traffic it
is, and how best to control (i.e. throttle or limit) it. These techniques
are used to help BSPs possibly free up some of this network capacity being
consumed and prolong their investment in network hardware. Only how exactly
can BSPs manage network capacity, really?
Bandwidth management is the method of observing data traffic flowing
through the network, evaluating that data for potential network capacity
concerns, and then making bandwidth-throttling, priority, or possibly even
traffic-filtering decisions based on some set network or usage policies.
There is a terrific white paper (Avoiding Network Capacity Collapse – by
John Kristoff of DePaul University) that goes into to a sampling of a number
of bandwidth management techniques available to BSPs today. There are also a
number of software packages available that monitor network capacity and or
control it by interfacing with various network hardware to change their
operational configurations. Some of these applications even try to manage or
control different types of data (or services) by allocating them dedicated
network capacity to maintain some predefined quality or expectation of
service.
Imagine a traffic cop sitting just beyond your computer making decisions
about what content or data you can and cannot access through the Internet
service you’ve subscribed. While this is perhaps the worst-case scenario of
bandwidth management, it is definitely within the realm of what proponents
of this area are endorsing. Bandwidth management may one day give BSPs
visibility into content or services that are delivered to their subscribers
by third parties. With this visibility in mind, in order for you to play a
game being hosted somewhere out on the Internet you may need to pay a toll
charge to your BSP for the privilege to play that game – even though you
already pay for access to the Internet.
I've always believed that there is no need for such thing as bandwidth
management (at least the kind that manipulates traffic!) Bandwidth is
consumed by content and without content you would not have any need for
bandwidth -- much less the need to manage it. Therefore if you were able to
manage content (either at the source or the destination), there would not be
any need to manage bandwidth -- the network would just do as it is supposed
to do which is take care of it self. Networks do this very well in fact.
Of course managing content is not going to buy you endless freedom from
bandwidth woes (caused by increased traffic on a network), this is where
careful planning comes into play. Besides the need to manage content, BSPs
also need to manage how they size their services that run over their
networks. Quality of Service (QoS) is definitely a step in the right
direction with regards to managing the size of services. The goal of QoS is
to make bandwidth reservations between the source (content) and destination
(subscriber) to ensure a minimum (and maximum) bandwidth allocation of the
service will be maintained. The bandwidth allocated equates to the quality
that the subscriber perceives the service. However, unless overall demand is
kept in check QoS will also suffer from limited bandwidth resources. There
is only so much bandwidth available, and if everyone requests a piece of it
some will ultimately get denied.
Unfortunately, as services have become impaired (slowed by high bandwidth
use), BSPs look to Band-Aids such as bandwidth management (which represents
an additional investment) to determine who their top talkers (so called
“bandwidth abusers”) are and even seek to divide up their networks into
smaller subsets (yet another investment) to sustain some previously
established standard for service speed and quality. The goal of offering
services should not be to seek out and punish people for using it, rather it
should seek out ways to maximize the ability of a service to generate
revenue and expand the subscription base. Many broadband data services are
poorly designed with respect generating revenue in proportion to the amount
of resources invested. If BSP data services currently offered to residential
customers were in line with how long distance carriers value bandwidth the
prices of these connections would either be much more expensive or the
bandwidth offered to subscribers would be significantly less. In addition,
if usage represented revenue would BSPs really care about which subscribers
are taking full advantage of the service they purchased?
I would argue that charting top talkers is a waste of time! Why, because
these subscribers are not doing anything wrong (the service can already
restrict most wrongful things from being possible), they are merely using
the service that they have paid for -- they aren't trying to take advantage
of the BSP. I equate this to pre-paying for fuel at the gas station. When
you pre-pay, you’re allotted some amount of fuel equivalent to that which
you paid. Thing is, if you consume the entire amount you paid for you (as a
consumer) are not wrong to use what you rightfully paid for. In this
example, it’s as if BSPs are expecting people who subscribed (prepaid) for
service to not consume their entitled bandwidth (fuel) -- leave it for the
next guy. I wish someone would leave me some free fuel at my next stop at
the gas pump...
Some day, a smart person is going to figure out what a data service is
really worth (take into account hardware costs, customer care, Internet
access costs, etc.) and those BSPs who have arbitrarily offered data
services to their subscribers in the past without this information are going
to look back and wonder how much money they could have made if only they
would have charged what the thing was worth or only provide the level of
service that the subscriber was willing to pay. As for managing bandwidth,
it will remain a pipe dream as no software really controls what happens
between two points on a network (hardware vendors control this space
transparently).
{ Content } ------- *
Network * -------- [ Demand ] |
expectations (low)
Figure 1.0 Simplified problem of managing bandwidth – (narrowband)
To illustrate this last point, refer to Figure 1.0. This is very simple
illustration of how content is delivered over a network to demand (i.e.
subscribers). There are two things that determine the amount of bandwidth
consumption: content and demand. Figure 1.0 depicts a limited capacity
interconnect (symbolized by the “-----“) between content and demand. The
limited capacity network (a.k.a. narrowband) serves as an interesting
barrier to the content (e.g. Internet) using a typical dialup Internet
service. As a result, the fastest anyone could consume Internet content was
at a rate of 56k bits per second (or the fastest connection speed the
subscriber’s modem could negotiate with the BSP). Both content and demand
adjusted to this limitation. Web sites limited their graphical content (even
provided alternate sites for abbreviated graphics that would load faster
over dialup lines) and demand was limited as each web page took several
seconds to load. Demand was further limited by the number of phone lines
that were available for subscribers to dial into. Since this was the best
that technology could offer, expectations were low. Through managing
content, demand, and expectations the dialup Internet service delivery model
worked for many years – in fact it is still in use even today.
{ Content } ===== * Network * ====== [ Demand ] | expectations (high)
Figure 2.0 Simplified problem of managing bandwidth - (broadband)
When the concept of broadband was introduced, its like everyone forgot
about what made narrowband so successful. Broadband (See Figure 2.0) is a
model that exercises extremes across the board. Networks, Demand, and
Content all have increased in this model. Perhaps the worst consequence that
also increased was subscriber expectations. In fact, some BSPs even
advertised the service as “Boundless”. Interestingly, the race into
broadband is falling back to something well understood in narrowband – the
network eventually becomes limiting. The advantage that narrowband had is
that technology reached a barrier – 56kbs. Subscribers understood this and
their demand and expectations were held in check. Unfortunately, because
it’s “broadband” the idea of something limiting demand or content has fallen
out of line with new subscriber expectations. Therein lies the actual
problem with broadband.
Broadband’s failure to manage demand as well as subscriber expectations
has created a feeding frenzy among hardware and software vendors egger to
offer BSPs solutions to its “problem” of declining bandwidth capacity. Only,
bandwidth capacity is not the problem – it is the result of BSPs refusing to
manage demand as well as subscriber expectations. To be fare, lets not
shoulder the blame entirely on BSPs for this current problem – the Internet
is also to blame. The Internet is a poorly designed animal and does not lend
itself to those companies who spend enormous amount of capital to offer
subscribers a high-speed onramp. Ideally, the best possible model for the
BSPs is one that affords them an opportunity to cash in on all traffic – a
general-purpose type of bill-by-the-bit model. In this way, BSPs can truly
manage demand.
Only by managing either demand or content, will BSPs find a solution to
their bandwidth capacity woes. Since a majority of the content that their
subscribers’ request does not come from the BSP (nor is it owned by the BSP)
it is hard to comprehend BSPs ever becoming a major source of Internet
content. Therefore, the only other avenue available to them is to somehow
manage demand.
In leu of implementing a bill-by-the-bit model (or perhaps in addition to
it), BSPs could manage individual services. In this way, the BSP could
enable individual services for their subscribers based on their needs and
price preferences. Individual services may include web, ftp, email, news,
telnet, vpn, etc. and be managed similar to the way call features are
managed on your local phone service (pay incrementally for each service).
Individual services could even be implemented without expensive upgrades.
Additionally it would allow the BSP to charge by the bit at different rates
based on the types of services the subscriber decides to purchase. Today,
these services are pretty much all available – again BSPs place little (if
any) restrictions on demand. Failure to manage this demand forces the BSP to
keep up with it -- so BSPs must invest in pricey upgrades in equipment,
network links, and software to sustain increasing subscriber demand and
expectations. However, unless BSPs manage demand and expectations they will
ultimately face this fact – the network will become limiting.
Of course, implementing bill-by-the-bit and or individual services will
also impact bandwidth capacity and there is really no avoiding the
inevitable – that ones network will one day become limiting. I like to think
of this as a good problem to have! If you’ve implemented measures to manage
demand, receive a fare value for the services you deliver, and you still
have bandwidth capacity problems that means your running a tight ship and
extracting the most out of your capital investments. It also means that
upgrading (increasing) the size of network links is less of a burden of
increasing costs but rather an opportunity for increasing revenues.
I want to make an important distinction in regards to this idea of
managing bandwidth. Earlier I explained that bandwidth management involves
observing traffic (monitoring), evaluating (analyzing) it, and then acting
upon it by changing some aspect of the existing network. I cannot condone
this process in essence because ultimately all networks must be changed in
response to increased traffic utilization. However I am in violent
disagreement with a number of positions held by supporters of bandwidth
management in its current context.
- Bandwidth management can taint, filter, or otherwise manipulate traffic.
I have a problem with the proposed manipulation of traffic to restrict
legitimate subscriber usage. Legitimate use should be all that is permitted
via the services the subscriber has purchased. If that service is access to
the Internet, than so be it. For example, if someone wants to download songs
from Napster the BSP should not in any way limit or alter this capability –
even if Napster refuses to make a deal with the BSP for revenue sharing.
- Bandwidth management woes are the fault of a few bad subscribers. There
is no such thing as a subscriber who is exploiting their Internet service –
that is completely absurd! BSPs needn’t shift the blame for not placing any
restraints on subscriber demand while elevating subscriber expectations
through marketing a service that claims up to a T1 speeds for some
unbelievably low price. With few exceptions, this feeling of subscribers
exploiting a service is caused by poor service sizing and pricing and the
absence of any restraints on demand. BSPs can pretty much restrict
subscribers from doing most bad things.
Content will be the only thing that can be managed but its control is
only possible at the edge of the network (source or destination). The
destination appears to be the first logical place to start as it is closest
to the subscriber as well as one of the bottlenecks of the system (the last
mile -- the other bottlenecks being the Internet and the content source).
One can throttle this endpoint to prohibit individual users from consuming
too much bandwidth (goes back to my point about designing services that the
operator's technology can sustain and will permit linear growth in bandwidth
demand), bill-by-the-bit, break down services into individual components, or
subscribers can better manage the content they already have (there are
several applications out there today that for example make browsing more
efficient by working more like a proxy). The source of the content can also
be managed but this often requires significant changes, as the goal here is
to optimize distribution of content. Very few content services actually do
this, thus they create a host of new problems downstream for others (like
BSPs) to solve. One example of managing content at the source would be to
only send unique content or only the changes (deltas). Teleconferencing
software has done a terrific job of this however few other applications have
taken this approach.
As a result you have companies trying to make up for lack of
sophistication and optimization in content delivery and consumption (demand)
by tweaking the network it traverses. Making changes to the network does not
fix this problem; rather it complicates many aspects of service creation and
delivery while only temporarily delaying the inevitable (consumption of the
available network capacity). Again, if you don't regulate demand or the
distribution of content you cannot fix this problem between the endpoints
(by managing the network) -- that is the pipe dream (not to mention a very
expensive endeavor). The network is merely the scapegoat, taking the blame
for BSPs failure to properly manage demand and subscriber expectations along
with years of content's lack of innovation and optimization in its delivery
methods.
Check out these other Birds-Eye.Net papers/products regarding
this article:
Can Birds-Eye.Net help you or your Company?
Receive your Birds-Eye.Net articles and white
papers hot off
the presses by adding our RSS feed to your reader.
|