The New Carrier Core: Make It Truly
BY ESMERALDA SWARTZ
Due to declines in legacy business and a perceived market slowdown,
carriers are under heavy competitive pressure to grow their networks in a
way that squeezes maximum margin out of their existing legacy services while
readying for the inevitable growth of more profitable IP-based services. But
the mismatch is widening between current-generation router performance and
the demand for more capacity to support IP traffic, new services, and to
keep pace with advances in the optical transport network. Emerging
next-generation routers bring truly scalable routing to the network core in
order to support these new requirements.
While awaiting those revenue-driving new IP services, we've already seen
Internet traffic mushroom. Fueled by swarms of new users and faster access
speeds, it's doubling every few months, expected to be a full 90 percent of
network traffic by 2003. That translates to a 150-million terabyte per month
load by 2003 -- up from a mere 350,000 terabytes per month in 1999,
according to industry estimates from analyst firm RHK. So as carriers offer
best-guess forecasts of future bandwidth requirements and which services
they'll need to support, competitive carriers now should be considering
shifting to this new routing paradigm.
The solution must deal with a hydra-headed dilemma facing carriers.
First, it must accommodate the unpredictable nature of the Internet itself.
It must be able to increase the capacity of network links to carry swelling
data traffic. It has to let carriers provision new services quickly to take
advantage of fast-surfacing opportunities to increase profitability. And,
bottom line, it needs to translate into minimized operational (and capital)
costs needed to run networks supporting these new requirements. To
appreciate the potential, compare the traditional method of scaling to what
is possible with a legitimately scalable, carrier-class, multiservice core
THE OLD WAY
In the past, service providers met bandwidth demand by increasing the
interface speeds of existing routers or adding more routers. This model is
increasingly cost prohibitive and impractical for several reasons. Using
multiple routers and/or displacing a core router with the next bigger box
disrupts the network and negatively impacts the provider's ability to
deliver reliable service.
For example, a customer has a fixed-configuration router with eight
slots, each with OC-48 connections. To add another OC-48 interface, the
customer must buy another such router. Once installed, it must be
interconnected to the first by two OC-48s. To get one more OC-48's worth of
capacity, the customer has to purchase a new router and three OC-48's worth
This method also uses valuable router interfaces just to interconnect
routers. And as the router cluster grows, more ports are needed, only
compounding the problem. Clustering increases the cost per port since the
ports used for interconnection are not serving customer links (and therefore
not generating revenue). This is an expensive proposition, and causes
disruption to the network, as well as the need to dispatch scarce IP talent
to install and configure new routers.
Current fixed-configuration routers are not well suited for the core of
large service provider networks, and are increasingly being pushed to the
edge of large networks, or employed only by smaller providers. Carriers no
longer can afford to run their networks constrained by such underlying
network topology and are seeking equipment footprints that can scale along
with requirements while the network remains in service.
A BETTER WAY
Carriers have no choice but to deploy scalable routers to future-proof
their networks and grow bandwidth on demand, and it was only a matter of
time before they began adopting a non-disruptive and incremental method of
scaling. New core routers incorporate a distributed architecture with
backplane interconnects. This allows for continual in-service upgrades so a
carrier can easily grow from OC-3 to OC-192 -- or even OC-768 -- in the same
platform. The undeniable reliability of such a system, combined with
advanced quality of service (QoS) and multi-protocol label switching (MPLS)
capabilities, lets carriers support legacy services and sell new
revenue-generating services such as VPNs, VoIP, and video-on-demand (VoD).
A key differentiator of next-generation routers is the ability to perform
at wire rate, regardless of route prefix length, route table size, MPLS
label swapping, QoS, or multicast. This means QoS needs to be hardware-based
and buffers must be dynamically configured to support both real-time, delay
sensitive and non-real-time, delay-insensitive applications. For competitive
differentiation, service providers obviously need to possess advanced QoS
features in order to offer enhanced service level agreements (SLAs).
Using inherently scalable technologies at the heart of the network will
evoke multiple benefits with tangible reductions in both capital and
operational costs, as well as increased operational efficiency, performance
and reliability. This will become more apparent as the solution grows in
size, since capacity grows on a linear basis. In large and high-speed
networks, where the rate of growth is high or likely to be high, the use of
scalable core routers will prove a compelling core-network value
A scalable router approaches the issue of growth in a fundamentally
different way than traditional architectures and offers an array of
advantages by affording the ability to scale over time as a single router.
Such specific operational cost savings include:
Administration of Route Tables
The burden and operational cost of making changes to route tables
and having to send out scarce IP experts is eliminated. A single route table
is maintained even when the router itself has scaled to many hundreds of
interfaces and multiple routers' worth of capacity.
Having a single router with a single route table dramatically simplifies the
network for upgrades, design, management, and operations. Power, cooling,
maintenance, and spare requirements are static. These are unknown quantities
in a multi-router architecture. The design process can be reduced to minor
revisions each year, rather than an ongoing planning and product evaluation
process involving lab and architecture groups.
A network designed with a scalable router will make it easier to back up
configuration files, easier to apply new system-wide policies, and easier to
cable manage. Day-to-day operations will be simple and familiar to
operations staff. As the POP scales in size, the router will still look the
same and have the same field support requirements.
Networks are constantly changing to keep the service offerings competitive
and feature rich. Software upgrades therefore will be a necessity and
something that operations will be asked to perform regularly. By using a
secure software upgrade solution with fast fall back (dual servers, dual
images), a single, scalable router can provide simple and reliable upgrades.
The multiple-router approach is more complex due to node count and multiple
hardware and software revisions.
Time To Provision
New service offerings can be rolled out quickly. Time-to-market
advantage will be a critical success factor for both incumbent carriers and
new service providers hoping to compete with the market leaders. The
increased abundance of inexpensive bandwidth will provide nimble carriers
with the opportunity to quickly introduce new services and fill up network
pipes with attractive offerings. We are rapidly migrating from bandwidth
expansion to service creation.
The new carrier core will be comprised of routers specifically designed
for carrier and ISP backbones. These routers will offer the scalability,
resiliency and port density to enable just-in-time bandwidth provisioning,
high reliability, and network availability -- all with the quality of
service needed for carriers and ISPs to support mission-critical
applications of the future.
Esmeralda Swartz is the director of strategic marketing at Avici
Systems, Inc., a developer of core routing solutions providing scalability,
resiliency, and port density; and enabling just-in-time bandwidth
provisioning, high reliability, and the quality of service needed for
carriers and ISPs to support mission-critical applications of the future
such as video streaming, VoIP, and Multi-protocol Label Switching VPNs.
To The April 2001 Table Of Contents