by Sultan Dawood, Solutions Marketing
Manager, SP Marketing Cisco Systems
But that was then. In today’s optical
conversations, talk tends to center on 40/100 Gig links, all the way
up to Terabit advancements. Why? The volume of consumer and business usage of bandwidth is astounding, on fixed and mobile networks. Upwards of 50% CAGR per year, on some portions of the “big” Internet, like last-mile access networks.
If you remember one thing about this article, please make it be this: by converging the optical and IP
layers of the network, capex and opex costs can be trimmed by 25-30%,
according to our ongoing and live research with service providers.
Path identification (traditionally handled within the “transport
silo”) happens much more quickly; apps and services (handled within
the “data services silo”) move more securely.
At least three factors are driving the
world’s explicit and implicit obsession with network capacity:
Device proliferation, video as an app, and the data centers fueling
cloud computing.
No one knows better than the readers of
this publication about the importance of Gigabit networks. The
emergence of 10 Gigabit (10 G) links in big networks began more than
10 years ago, and seemed like enough capacity, for a long time.
If history is any indicator, doomed is the man or woman who publicly wonders why on earth so much capacity is needed. In the 1960s, cable television providers wondered why they’d ever need to build for more than 12 (analog!) channels. Back in the early days of dial-up data connections, some wondered why we’d ever need to go beyond 56 kbps. We’ve seen this “I’ll eat my hat” scenario over and over, in the course of network expansion.
Because the majority of today’s transport networks are conveying data using 10 Gig networks, and at the same time are facing unprecedented volumes of usage, decisions about expansion tend to center on three known options:
1) Add more 10 Gig links
2) Go straight to 100 Gig
3) Find a stepping stone path to 100 Gig via 40 Gig
What is perhaps lesser known are the decision sets and resultant economic impacts of getting to 40/100 GigE, using existing routers, minus the provisioning and maintenance of electrical-to-optical-to-electrical conversion transponders, minus the operational expenses involved with maintaining what is essentially two disparate transport networks.
Consider: What if you could turn up a
link to a customer in minutes -- not months?
IP over DWDM is an innovative option
(we’d argue the option) that economically justifies 40/100
Gig adoption, by reducing additional unnecessary equipment and
associated interfaces including optics – thus lowering
requirements for additional power, cooling and space. It’s been
proven that integrating optical intelligence into a router makes it
cognizant of any optical path degradation. That means routers can
proactively ensure that any apps and services in transit are
protected from degradation or failure.
Why: The forward error correction (FEC)
intelligence as a result of integration of optical into the routers
will provide awareness to automatically switch to a secondary, safer
path, before any optical impairments impact any app or service
performance.
So we won’t venture into questions of
whether 40/100 Gig networks are necessary. Instead we’ll
look at what’s driving the world’s data capacity needs, then
examine the options in getting into “the big GEs,” including the
substantial economic benefits associated with converging the optical
and IP layers.
The Capacity Drivers
Think about the number of
IP-connectable devices in your home or business, 10 years ago,
compared to now. All of them want Internet connectivity – some more
so than others.
Plus, most gadgets in the device
ecosystem are mobile. Not long ago, we connected to the Internet when
we went to the office, or were at home. Fixed connections –
personal computers, tower computers, laptops to some extent. Internet
wasn’t an option when outdoors, or when driving, to navigate via
GPS, or find a restaurant, or locate friends.
Our ongoing VNI (Visual Networking
Index) research indicates that by 2016, there will exist nearly 19
billion global network connections – enough for 2.5 per person, in
the world. (Click here for more VNI information:
http://www.cisco.com/en/US/solutions/collateral/ns341/ns525/ns537/ns705/ns827/white_paper_c11-481360_ns827_Networking_Solutions_White_Paper.html)
Capacity Driver #2: Video
Driver number two dovetails with the
first one: Video. With more and more powerful, HD video-capable
screens fetching and tossing big streams of data in and out of
whomever’s data cloud, the question of how and when to scale the
network is more relevant than ever.
Beyond the 50+% compound annual growth
in broadband usage – wired and wireless – new pressure points are
arriving into the consumer and business marketplaces with alarming
regularity.
Consider the spate of recent
announcements from consumer electronics and PC makers about putting
high-resolution screens into handhelds, tablets, laptops and
televisions. High-rez screens means high-rez streams.
Indeed, smart phones and tablets impact
real-time network capacity in a big way, because most include still
and video cameras, capturing images and sound in SD and HD. Video
eats up capacity like nothing else (so far.) Already, and again
according to our ongoing VNI research, more streamed video is done in
HD than SD.
At the high end of the video spectrum,
the 2012 trade show scene is producing regular headlines about the
pursuit of 4K resolution.
Even using the best compression on the
market today (which goes variously by H.264, AVC and MPEG-4), a 4K
stream “weighs” as much as 17 Mbps. Compare that to today’s 1-3
Mbps carrying capacity, for “regular” HD video compressed with
H.264/AVC.
Yes, H.265 compression is on the way,
which will do for H.264 what it did for MPEG-2 – but still. The
point is that network bandwidth is under enormous strain right now,
with no signs of easing up.
Capacity Driver #3: Clouds and Data
Centers
Consider: Networks used to move static
web pages, or haul 64 kbps telephone conversations, or broadcast (in
a one-to-many sense) SD video. These days, they do all that plus
stream HD, (unicast and multicast) video to high resolution displays.
They haul video phone conversations. They carry adaptive bit
rate video, which by its nature behaves like a gas, filling all
available space.
Plus, big networks traditionally were
“silo’ed,” with transport and data departments and people
operating largely independent of one another. Not so anymore. Why:
The fetching of a web page is one thing – simple, isochronous, not
a huge strain on the network.
Today’s emerging applications are
another story entirely, segueing into transport-heavy fare like the
shipping or storing of enormous amounts of digital stuff – think
digital pictures, videos, and cloud-based storage in general .
Transport-heavy networks needs require mutual and simultaneous
attention, from both “data” and “path/transport” departments
in the organization.
“The cloud,” in all its iterations,
then, is capacity driver number 3. Clouds, and the data centers that
enable them, are both sourcing more network traffic -- and
struggling under its weight. Anyone building a cloud designed to
service big geographic areas will need multiple data centers that are
interconnected – preferably intelligently.
Today’s data centers are connected
via a combination of routers and transport networks. Connecting
Router A in Data Center A to Router B in Data Center B, for instance,
requires transport infrastructure. Traditionally and so far, that
transport has been 10 Gig.
However, as bandwidth demands increase,
with routers leveraging 100 Gigabit Ethernet interfaces and data
centers moving large volumes of content, it would make sense to
increase the transport network capacity and transition from 10G to
100G DWDM. To add further to that thought, it is also possible to
integrate optical interfaces directly into routers, thus offering an
innovative and green approach as well as a truly integrated solution
that in turn justifies a faster and more cohesive transition to
40/100 G links.
The Meaning and Importance of
Coherence in Optical Transmission Systems
In optical terms, “coherence”
refers to the ability of a lightwave to produce interference patterns
that work in favor of the intended signal. Coherent optical
communications products arrived in the marketplace roughly coincident
with 40/100 Gig networks, because they are intrinsically suited for
very long-haul networks – upwards of 500 km and higher.
Any time signals are distributed over
very long distances, however, two things can happen that compromise
performance. First is plant anomolies, which cause signal strength to
lag. That necessitates amplification, but amplifiers boost both the
intended signal, and any noise that is present. Dispersion
compensation is then required, to compensate for impairments.
These impairment compensation
activities do not come for free – especially when the distance in
question is measured in thousands of kilometers. That’s why service
providers considering the shift to 40/100G seek ways to do so without
adding additional equipment for signal impairment compensation.
Service providers seek the most
economical, yet best performing signals that can exist on their
network as a way to control total cost of ownership, by adding 40/100
Gig capabilities to existing routers and over existing infrastructure
– even if the fiber plant is marginal in places.
This is where coherent optical systems
really shine. We’ve seen (because we built) 100 Gbps connectivity
over a 3,000 km link, on top of existing 10 Gbps fiber
infrastructure, that is vendor-independent. (More here:
http://bit.ly/PndKol)
Here’s a real-world example: You’re
running a video connection to a customer. It’s an MPLS tunnel
mapped onto an optical wavelength. Let’s say that fiber degrades.
With the forward error correction (FEC) techniques within IP-DWDM,
thresholds can be set ahead of time, to default to another optical
path.
Maybe the pre-FEC
value is 10-17 , but at 10-19 the router knows
to switch the video connection to a cleaner path – proactively.
Having ways to set thresholds and interact between layers ensures
that the video connection stays solid – and your customer has no
idea that a problem almost occurred.
Now What? Harmonizing Optical and
Packet Transport
We’ve talked about capacity drivers,
and the benefits of IP-DWDM as a way to get to 40/100 Gig- without
stranding prior investments in routers or optics. The fact is that
the dominant type of traffic on broadband networks today is
packet-based; existing optical networks aren’t as well suited for
packet-based delivery than other types of traffic.
This is how IP-DWDM began, for what
it’s worth. Service providers asked how to bulk up capacity without
disrupting capital or operational spending. How to save money in a
packet environment led to the need to do certain things differently,
which led to the development of IP-DWDM.
Because of that, the drumbeat toward
converging the optical and IP domains began, as a way to reduce
capital and operational costs, as well as to have a better handle on
network controls. Equally relevant: The ability to launch new
services/apps more quickly, and more securely. IP over DWDM is one
example of this convergence, which provides between 25-35% in capital
and opex spending, combined.
Perhaps some day it will seem quaint,
that at one time network architects were debating the convergence of
the optical and IP layers in long-haul transport. But for now, the
decision to go with IP-DWDM is still a bit maverick, for those going
through it.
Why? Because getting there involves
cutting across people and organizational domains. Never easy to do.
Despite ongoing proof that a) 100 Gig gear exists that works over
distances of 3,000 km, without need for signal compensation, and b)
IP-DWDM is a cleaner solution, because it eliminates excess optical
equipment and interfaces, and c) its pre-FEC can proactively re-route
mission-critical data before signal paths become impaired, IP-DWDM is
still in the “crawl” portion of any “crawl-walk-run”
technological evolution.
We’ll end with this: Service
providers will continue to sprint to keep up with capacity, and to
compete with new, over-the-top providers. A more integrated and
converged network lends itself better to the packet-based traffic
loads of today. It scales for the future, and it saves capital and
operational costs. Because one thing is certain: Data loads are not
letting up anytime soon.
Sultan Dawood holds the position of Senior Marketing Manager for Core Routing and Transport Solutions at Cisco Systems. He has spent the last 18 years of his career focused on data networking and telecommunication systems working closely with both Enterprise and Service Provider customers. Prior to Cisco, Sultan held senior marketing and engineering positions at Hammerhead Systems, Motorola, 3COM, ADC Telecommunications and Litton Systems.
Sultan has a Bachelor of Science degree in Electrical Engineering from Old Dominion University in Norfolk, Virginia. He is also a Board member and the Vice President of Marketing for the Broadband Forum.
Sultan Dawood holds the position of Senior Marketing Manager for Core Routing and Transport Solutions at Cisco Systems. He has spent the last 18 years of his career focused on data networking and telecommunication systems working closely with both Enterprise and Service Provider customers. Prior to Cisco, Sultan held senior marketing and engineering positions at Hammerhead Systems, Motorola, 3COM, ADC Telecommunications and Litton Systems.Sultan has a Bachelor of Science degree in Electrical Engineering from Old Dominion University in Norfolk, Virginia. He is also a Board member and the Vice President of Marketing for the Broadband Forum.
