lml and Thread,
The analogies are running rampant here. Railroads, trucking, MAC operating
systems, super market aisles...
In each instance there are two constants that become obfuscated by the
proprietary views of the service providers, and by the self serving views of
others. Those constants remain the medium and message-supporting services.
Stated another way, transmission paths and services. Above all of this,
we find content.
Those constants are absolute, while all other arguments remain academic,
if not trickful.
The medium in this case is represented by copper twisted-pair wires, or
coxial cable, or fiber or wireless transmission paths. Paths are patently
different from services, which are patently different from the actual
content, which rides over all of the above.
When you can appreciate the imaginary boundary which is made up of
protocols which reside between the medium and the service, then you can
more objectively discuss the other matters which surround this discussion.
And then you can assess whether the specific form factor or physical
appearance of the medium is important, or rendered unimportant.
These delimiters (or shims, as one author called them - see article below),
separate what the facilities based providers have offered, historically,
in the way of wire/air paths, and what the higher orders of services are
that support what we as end users use: dial tone, program TV services,
Internet communications services, etc.
Note, in some cases, facilities based providers offer both classes, but this is
a carry over that can be traced to conditions which preceded competition.
Some regional semantics are at play here, also. Internet Access platforms
in the last mile, for example, are not the same as Internet related delivery of
content from a server.
In the last mile thread we've often discussed the eventual battle of the OSI
Layers, whereby services would become virtualized in the upper strata of
the stack, as in IP voice, multimedia etc., independent of the lower strata
facilities-based providers' provisions for same. In this way, a user could
select their telephony provider at the IP layer, irrespective of who their
local LEC was. The same could be stated for other services, as well.
There is a good article which exemplifies these principles in this month's
X-Change Magazine. In fact, it covers these topics almost word for word,
in parts, the way we once covered them some time ago in the LM Thread.
x-changemag.com
(copied below for posterity)
Regards, Frank Coluccio
-------------------------------
Posted: 06/01/1999
Middle Management
Bridging the Gap Between Services and Access
By Kevin Walsh
Conventional wisdom says offering a bundle of services is a winning
combination for communications companies.
Bundled services over a single-access network can yield higher revenue, higher gross margin, lower customer churn (customer turnover) and higher net margin. Each incremental service provides incremental revenue. Since each new service should require little or no new investment, overall operating margin is substantially improved. Studies have shown that churn is significantly lower when the customer is depending on the access provider for more than one service. And since churn is lower, less marketing expenditure is required to capture and retain customers, resulting in improved net margin.
But there seems to be some confusion over the definition of a service as
opposed to access or transport.
Local dial tone is a service. Long distance is a service. Internet access is a
service. Access to the corporate private branch exchange (PBX) is a
service. Access to a corporate virtual private network (VPN) is a service.
Digital subscriber line (DSL) and T1, meanwhile, are not services, but
rather pipes. But these broadband transport technologies can allow carriers
to deliver multiple--or bundled--services affordably. Carrier investment in
DSL is exploding. Of course, DSL generally delivers greater bandwidth at
a lower cost than prevailing T1 pipes, while still leveraging the enormous
copper-loop infrastructure. But T1 is, and will remain, an important means
of access. According to U.S. Bancorp Piper Jaffray, technology equity
research division, Minneapolis, T1 will be the primary means by which
competitive local exchange carriers (CLECs) reach customers not passed
directly by their fiber rings. T1 should unquestionably be included in the
toolbox of technologies available to access providers, and can be leveraged
significantly by use of inverse multiplexing techniques (i.e., four T1 lines
inverse-multiplexed to provide a 6 megabits per second [mbps] symmetrical
access facility).
All of these advances are beneficial both to the business subscriber and the
access provider. For the business subscriber, the benefit is more bandwidth
at a lower cost over a broader geography. For the access provider, the
benefit is a broadband access network capable of delivering multiple
services, also over a wide geographic region.
While all of these technologies are useful, they are not what the service
provider sells. The service provider sells dial tone, Internet access and the
like.
So how do service providers market a broad set of services, while at the
same time leveraging advances in local-loop transmission technology?
The answer is to separate transmission from services in the access
network using a multiplexing shim. That is a layer placed between the
transmission and services layer to perform the following functions.
* Layer separation. The primary function is to separate the transmission
and service layers so each may operate independently. A local dial tone or
Internet access service, for example, shouldn't care (in fact doesn't even
know) whether it is operating over a T1 line or a symmetrical DSL (SDSL)
line. Similarly, an SDSL line shouldn't worry about whether traffic
traversing the local loop is voice traffic or video traffic.
* Multiplexing. Obviously, if multiple services are to share a single access
facility, a multiplexing function is required. The multiplexing function should
use bandwidth as efficiently as possible and supply services that require
them with quality of service (QoS) guarantees.
* Service isolation. Partly as a result of multiplexing itself, services can be
associated with an addressed connection and can therefore be provisioned,
managed and delivered to the appropriate destination (off-ramp). The
service off-ramp aspect of service isolation is perhaps the most important.
In a bundled services environment, each service needs to be delivered to a
different upstream network (e.g., Internet, public switched telephone
network [PSTN] or VPN). The off-ramps to these upstream networks are
likely to be at different locations. By placing services into an addressed
connection, the connection can be delivered to the appropriate upstream
network irrespective of location (see graphic, "Bridging the Gap," below).
Graph: Bridging the Gap
Easing Traffic Flow
What options do equipment manufacturers have in choosing technologies to
implement the multiplexing shim? Basically, all alternatives can be divided
into two categories: time-division multiplexing (TDM) and label
multiplexing. (Frequency-division multiplexing fell out of favor a long time
ago and wavelength-division multiplexing [WDM] is unlikely to hit the local
loop any time soon.)
With TDM technology, which has been around since the invention of the
telephone, services are placed into time slots at the network ingress point
and then extracted from the same time slots at the network egress point.
While TDM techniques are highly reliable and deliver rock-solid QoS
guarantees, they are extremely inefficient in terms of bandwidth utilization,
especially in data-heavy environments.
Furthermore, service providers of all types have been investing billions of
dollars converting circuit-switched core networks to cell/packet
technologies. It makes little sense to invest in access technology that is well
on its way to becoming obsolete and is poorly suited for this new
data-centric world.
Examples of label-multiplexing technologies include frame relay, Internet
protocol (IP) and asynchronous transfer mode (ATM). With label
multiplexing, traffic (such as data bits and voice samples) is placed into
addressed envelopes (cells or packets). Intelligent devices (such as routers
and ATM switches) forward the traffic to the appropriate destination using
the addressing information carried in the packets themselves.
Frame relay has seen widespread deployment as a carrier service.
Primarily targeted at data applications, it is a cost-effective, leased-line
replacement technology used to interconnect routers. While mechanisms
exist in frame relay to differentiate QoS--such as committed information
rate (CIR), and some vendors offer products capable of supporting voice
over frame relay--the technology was never intended to support
isochronous applications. As a result, voice over frame relay has seen very
limited deployment, and what little there is has largely been outside the
United States.
Although a controversial idea, many believe that since the vast majority of
data traffic is IP, it makes sense to use IP as a multiplexing protocol for
nondata applications. It's envisioned that in the future all traffic--voice,
video and data--will be IP packets. Consequently, there would be no
multiplexing function for dissimilar traffic types.
But today, voice and data typically come in different forms and from
different devices. Voice comes from analog and digital telephones, key
systems and PBXs, so multiplexing is required. Then there's the issue of
combining voice and data over certain portions of the network, typically the
wide area network [WAN], to derive cost savings.
IP certainly meets the key requirements for the multiplexing shim: It
consumes bandwidth statistically and services can be placed in "flows" (not
quite virtual circuits [VCs], but good enough). The only challenge is the
ability of IP to make and maintain service quality guarantees on behalf of
individual services--a hotly debated subject.
ATM policing, shaping, traffic management, connection admission control,
queuing, buffering and scheduling mechanisms know these are not present
in most devices that currently forward IP packets, although many
companies are working on it. Also, many products and networks that
support voice over IP (VoIP) use ATM underneath and frequently use
point-to-point protocol (PPP) over Ethernet on the premises. Someday IP
probably will be capable of making the same stringent QoS guarantees that
ATM is capable of making today. But not today.
Interestingly, ATM seems to have finally found a role for itself. It is not,
and will never be, the ubiquitous unifying protocol the industry thought it
would be many years ago (other protocols are now held up as the grand
unifying protocol). Nevertheless, there are certain functions within the
network for which ATM is well suited. The multiplexing shim in the access
portion of the new public network is one.
ATM consumes bandwidth statistically; it makes use of VCs and can make
highly granular QoS guarantees. Also, ATM is--after years of politically
charged standards development, interoperability testing and actual
deployment--a reliable, mature technology.
So most service providers and equipment manufacturers are leaning toward
ATM as the local-loop multiplexing protocol of choice. This even applies to
many that bill themselves as pure IP networks. At the same time, however,
most service providers and equipment manufacturers are integrating
advanced IP capabilities and leaving the door open to using IP as the
local-loop multiplexing protocol. This does not lessen the viability of key
technologies such as VoIP, provided these technologies are driven by
application needs.
The obvious benefits of label-multiplexing protocol in the local loop are
bandwidth efficiency and transmission independence.
By using statistical techniques, it long has been known that IP and ATM
accommodate data traffic far more efficiently than TDM. However, as the
data component drops, and the voice component increases, TDM
efficiency improves (see graphic, "A Balancing Act," below).
Graph: A Balancing Act
The use of a label-multiplexing protocol in the local loop allows services to
be deployed independent of underlying transmission technology. This
transmission independence allows network planners complete flexibility in
choosing transmission technologies.
Probably the most important, yet least understood, benefit associated with
label multiplexing in the local loop is service virtualization. Think of it as a
"service extension cord." The service, such as local dial tone, Internet
access or VPN, actually is provided by someone other than the local
access provider. VCs within the access provider's network deliver services
from the nearest upstream network point of presence (POP) all the way to
the customer premises.
The advantages of service virtualization primarily have to do with service
provisioning. Since services are delivered in VCs, they can quickly, easily
and cheaply be turned on and off from element management systems,
legacy operations support system (OSS) infrastructure or even by the
subscribers themselves.
The Span of Provisioning
One of the most challenging aspects of bundled services networks is the
process of turning services on and off, or service provisioning. In the old
days (and still today in far too many cases), turning a service on required
weeks or months of time, the involvement of many people and, in many
cases, trucks had to be dispatched. In short, it takes far too much time and
costs far too much money. Often, as much as 70 percent of an access
provider's cost is tied up in the people, systems and trucks required to
provision services. But new technologies and procedures can reduce those
costs.
Element management systems (EMS) with northbound common
object request broker architecture (CORBA) interfaces to
provisioning orders from the OSS to be accepted.
Dynamic service selection (DSS) allows subscribers to dynamically
select upstream network providers without involving the access
provider.
Directory-enabled provisioning (DEP) allows subscribers to
purchase new services such as VoIP or VPNs without involving the
access provider.
However, none of these technologies work unless the services are carried
in virtual connections.
For an EMS to flow through provisioning orders from the OSS, it must be
able to dynamically make connections between the subscriber and the
ultimate service provider, such as ISP or long distance carrier. With VCs it
is possible to do this; with physical circuits it is not.
For a subscriber to dynamically change upstream providers, the access
network must be able to move the connection from one service provider to
another, as well as perform the necessary address translation. With VCs it
is possible to do this; with physical circuits it is not.
Finally, for subscribers or third-party service providers to set up new
services, the access network must be able to create new connections
between existing subscribers and new providers (as well as query directory
infrastructure for security and accounting purposes). With VCs it is
possible to do this; with physical circuits it is not.
It is increasingly clear that access providers must offer a bundle of services
to survive in the highly competitive LEC market. Furthermore, access
providers must have adequate infrastructure flexibility to provide bundled
services over any type of access technology: DSL, T1/E1, inverse
multiplexing over ATM (IMA), T3/E3, etc. Doing so requires the use of a
label-multiplexing shim across the local loop. Today, that's ATM.
But the greatest benefit in using ATM, and label-multiplexing protocols in
general, over the local loop is that services become virtualized. This not
only dramatically reduces provisioning costs, it also enables entirely new
applications, such as subscriber-driven provisioning.
Kevin Walsh is vice president of marketing for Accelerated Networks
Inc., Moorpark, Calif. He can be reached at
kevinw@acceleratednetworks.com.
|
