400 Gb Layer-1 connectivity would
cost $2.40. If the same 100 Gb-
CWDM4 infrastructure is changed to
serve 400 Gb-DR4 (PSM4) optics by
changing from Duplex-LC to MPO
based infrastructure, the additional
cost would be $1.47, which is about
5% more than staying with the
Duplex-LC architecture. When
switching from a Duplex-LC to
MPO-based infrastructure, the fiber
count is increased by 4x, but the
lower cost of PSM4 based optics
will offset the total cost.
In the second case, it is assumed
that the initial infrastructure is
based on 100 Gb-PSM4 migrating to
400 Gb-DR4 (PSM4). The additional
cost incurred (when compared to
a reference $1 for 100 Gb-CWDM4
link) is $0.68. When migrating
from a PSM4 at 100 Gb to 400
Gb-FR4 (CWDM4) architecture,
the additional cost is $1.64. For the
specific structured cabling design,
as per Table 2 and per the reasonable
assumptions laid out in Table 1, the
total cost analysis indicates that it is
cost-effective to choose the PSM4-
based optical fiber architecture and
MPO-based structured cabling for
a future-proofed infrastructure.
CONCLUSION
The choice of optical fiber transceiver
has a big impact on the total cost
of the link. Decisions made solely
based on the cost of optics may not
result in optimized cost effectiveness.
Rather, a total link cost, which
considers all the components of the
link, will result in cost optimization.
If a data center infrastructure is
designed and built only for a single
56 I ICT TODAY
speed, then a CWDM4 transceiver
with duplex-LC based cabling and
connectivity infrastructure is the
most cost-effective option at 100 Gb;
at 400 Gb, there is almost no cost
differential between the CWDM4
and PSM4-based infrastructure.
If the same data center
infrastructure is designed for 100
Gb today with the intention of
migrating in the future to 400
Gb, while maintaining the cable
infrastructure as much as possible,
a different conclusion arises.
When migration is considered, it
is observed that choosing a PSM4
architecture with the MPO-based
cabling infrastructure provides
the best value, despite paying a bit
more for parallel fibers initially.
The relatively higher cost of
CWDM4 at 100 Gb and 400 Gb is
disadvantageous when total link
cost is considered. These results and
conclusions are based on a particular
data center structured cabling
configuration, so results may differ if
different variables are considered or
a unique design is deployed.
AUTHOR BIOGRAPHIES: Mike
Connaughton, RCDD, is the market
segment manager at Nexans Data Center
Solutions. He is responsible for strategy
related to the data center segment. Mike
has been involved in fiber optic cable
manufacturing and the ICT industry for
over 20 years with experience in
engineering and marketing. He is a
graduate of Wentworth Institute of
Technology. Mike can be reached at
mike.connaughton@nexans.com.
Rakesh Sambaraju is the applications
engineer and Nexans TESLA Sr. expert
for Nexans Data Center Solutions where
he is responsible for the technical road map
by staying current with standards activities,
as well as customer demands. Rakesh
received his MSc from Technical
University of Denmark and a PhD from
the Universitat Politecnica de Valencia.
For the past 10 years, Rakesh has worked
on different aspects of fiber optic and data
center communications technology.
He can be reached at
rakesh.sambaraju@nexans.com.
REFERENCES:
1. Sambaraju, Rakesh, Cloud Technology
and the 400 Gb/s Ethernet Landscape, ICT
Today, pgs. 33-40, 2018.
2. http://psm4.org/.
3. CWDM4 MSA, http://www.cwdm4-
msa.org/.
4. Ethernet Transceivers Forecast, Lightcounting,
2018.
Devices based on
CWDM4 technology
for the 4-lane coarse
wavelength division
multiplexing architecture
have lasers operating at
different wavelengths.
These modulated optical
signals with wavelength
diversity are multiplexed
into a single optical fiber
using a multiplexer.
link
link