Why Ethernet switches with WAN PHY and EoS have an economic advantage
Steve,
One of the traps that many of us have sometimes fallen into, is not
understanding the distinction between infrastructure and
services. Infrastructure is inherently layer one of the OSI
model. Services are what rides on the infrastructure providing OSI layer
one and above functional processes that generate billing records of one
sort or another. Because of their layer three switching functionality,
routers are used for service enabling facilities, not transmission
infrastructure. There is some layer one type of emulation tunneled over
layer three systems, this is most commonly how Frame Relay services are
implemented. In spite of the lower cost, Frame Relay still does not have
the market that plain old "Private Line" service does. The ironic part of
this is that the same inter-machine trunk point to point infrastructure is
used for all services, including Voice, ATM, Frame Relay, Private Line, and
the Internet.
There is a lot of effort underway to create the "God Box", a cost effective
multi-service edge interconnect system. The key to this is "cost
effective". The forays into attempts at using ATM as the common switching
fabric for such a "God Service" have not proven cost effective without
something called statistical gain. Anyone that has ever operated a shared
service network (a LAN is the most common example) within a common
demographic environment (a multi department network within a large company)
knows that busy hour traffic (normally at 8:30-9:30; 12:00-13:30;
16:00-18:00) will saturate the network such that the probability
"statistical gain" actually degrades the service. For companies that are
paying high monthly charges for services, degraded services, just when they
need it most, is unacceptable, and they will go to someone else that does
not attempt to use statistical gain as part of the service model. I
personally believe that the actual market for such a "God Box" is not very
large compared to the overall market. I am skeptical that using MPLS will
reduce the cost and complexity enough to make "God Services" cost effective
long term.
Besides, by the time that such "God Boxes" get into the market, 10GbE, and
Ethernet over SONET/SDH will start having an impact on the services that
customers are requiring from their service providers. The IT customers
will be looking for the ability to use very simple services that greatly
reduce the cost of ownership of their data networks. That will put such a
massive amount of competitive economic pressure on the "God Services" that
they may not be able to pay for the facilities that it took to implement
them.
The WAN PHY on 10GbE will make use of very simple optical point to point
services. The lower cost of the Ethernet switches with upper layer
capabilities, able to use the optical infrastructure, instead of expensive
routers, will be part of competitive economic pressure.
Except for commodity Internet type services, I don't believe that routers,
even with EoS interfaces will have a massive market. Those routers may
individually have massive bandwidth capabilities but they will be very
expensive and subject to major commodity economic pressure on the services
provided on them. I don't believe that they will be able to become an
actual part of the simple transmission infrastructure that is in place
today. (The perceived complexity of the existing infrastructure is actually
a factor of the scale of the infrastructure not that of individual systems
or functions.) Routers will continue to be service facilities systems.
Thank you,
Roy Bynum
At 08:59 AM 1/24/01 -0500, Steve Augusta wrote:
>Roy,
>
>Assume that a router had a WAN PHY which provided all of the SONET
>bytes (not just the subset being utilized by 802.3ae) and automatic protection
>switching. Why could this router not be utilized as part of the core
>transmission
>network?
>
>Thanks,
>Steve Augusta
>