होम [IEEE 2015 IFIP/IEEE International Symposium on Integrated Network Management (IM) - Ottawa, ON,...

[IEEE 2015 IFIP/IEEE International Symposium on Integrated Network Management (IM) - Ottawa, ON, Canada (2015.5.11-2015.5.15)] 2015 IFIP/IEEE International Symposium on Integrated Network Management (IM) - Measuring enterprise network usage pattern & deploying passive optical LANs

, , , , , ,
यह पुस्तक आपको कितनी अच्छी लगी?
फ़ाइल की गुणवत्ता क्या है?
पुस्तक की गुणवत्ता का मूल्यांकन करने के लिए यह पुस्तक डाउनलोड करें
डाउनलोड की गई फ़ाइलों की गुणवत्ता क्या है?
साल:
2015
भाषा:
english
DOI:
10.1109/inm.2015.7140402
फ़ाइल:
PDF, 1.19 MB
Conversion to is in progress
Conversion to is failed
0 comments
 

अपनी समीक्षा पोस्ट करने के लिए साइन इन करें या साइन अप करें
आप पुस्तक समीक्षा लिख सकते हैं और अपना अनुभव साझा कर सकते हैं. पढ़ूी हुई पुस्तकों के बारे में आपकी राय जानने में अन्य पाठकों को दिलचस्पी होगी. भले ही आपको किताब पसंद हो या न हो, अगर आप इसके बारे में ईमानदारी से और विस्तार से बताएँगे, तो लोग अपने लिए नई रुचिकर पुस्तकें खोज पाएँगे.
Measuring Enterprise Network Usage Pattern
& Deploying Passive Optical LANs
Yaoping Ruan, Nikos Anerousis,
Mudhakar Srivatsa, Jin Xiao

R. Todd Christner, Luis Farrolas, John Short
IBM Global Site & Facilities Servic

IBM T. J. Watson Research Center
LANs in terms of deployment flexibility, easiness of
management, environment friendliness, capital, and operating
costs.

Abstract— Recent advances in the manufacturing and
commercialization of passive optical components are now
extending the capabilities of fiber to edge and campus networks.
This paper presents a comparison between Passive Optical LAN
(POL) and copper-based LAN solution, and demonstrate the
benefits of PON such as reduced infrastructure footprint and
cost, reduced power requirements, future-proof bandwidth,
greener infrastructure, safer, higher security and higher
reliability.

II. ENTERPRISE NETWORK TRAFFIC PATTERNS

2.88% 4.35%

Keywords—component; formatting; style; styling; insert (key
words)

8.45%
42.12%

10.19%

I. INTRODUCTION
A passive optical LAN is an ideal solution for new
infrastructure projects and the upgrade of existing
infrastructure for a number of reasons:

32.01%

1. Guaranteed Bandwidth: Today’s enterprise traffic
patterns fueled by server and data center consolidation,
virtual desktop infrastructure (VDI), bring your own device
(BYOD), mobile, and cloud computing, are better served by
a centralized switch model compared to traditional
workgroup technologies with layered active switches.

Figure 1: Bandwidth Consumption of Enterprise Applications

We analyze network traffic patterns using data captured in
one satellite site of a large enterprise. The site has about 1500
employees with each having an office, and about 30 conference
rooms. Each employee has an IP phone and most have only
one desktop computer or laptop computer. The networking
setup of this site is what can be found typically in large
enterprises. Our measurement and analysis focus on traffic
between the end user and the core router which is typical i; n
enterprise campus network, rather than data center network as
measured in [1] [2].

2. Future Proof: Passive Optical LANs offer a futureproof upgrade path to safer, greener, higher security and
higher bandwidths over the same fiber infrastructure.
3. Capex and Opex Savings: Passive Optical LANs
replace the active Intermediate Distribution Frame (IDF)
equipment (aggregation Ethernet switches) with passive
components, reducing space, energy and cooling
requirements, as well as lower installation costs. Passive
Optical LANs replace traditional copper wiring with fiber
saving space and weight. Passive Optical LANs require
simpler management and offer advanced capabilities that
can be easily integrated with campus-wide provisioning and
management applications.

We observe most traffic goes through the core switch
which implies very little peer to peer traffic. This is typical in
an enterprise environment since most of the enterprise
applications are client-server based and servers are hosted in
remote data centers.
For bandwidth consumption of different applications, email
and Web traffic consumes more than 74% of the bandwidth,
followed by file transfer since the organization uses a
distributed file system, and online conference which is
commonly used to share screens. A small portion of the file
transfer is induced by Cloud services that appear mostly in the
HTTP traffic since the Cloud service user interface is Webbased. Figure 1 demonstrates the top applications based on
bandwidth usage.

This paper offers a study of the Passive Optical LAN
technology and its implications for cabling infrastructure
projects. We demonstrate enterprise traffic patterns using
network traffic captured in a large enterprise campus, then
discuss traditional LAN architecture, Passive Optical LAN
components, total cost of ownership (TCO) analysis, and
further implications on network management, real estate, and
energy consumption. We demonstrate that the passive optical
architecture is vastly superior to traditional copper cable-based

978-3-901882-76-0 @2015 IFIP

e-mail
Web HTTP
File transfer
Online conference
Instant messenger
Others

890

Percentage of User Population

access layer switch forwards the network packages initiated
from individual computers to the distribution layer switch.
Finally the package gets forwarded to the core switch and
routed to the destination.

46.30%

50.00%

40.30%

40.00%
30.00%
20.00%

13.10%

10.00%
0.20%

0.00%

0.10%

1Mbps – 10Mbps
50Mbps - 80Mbps
0 – 1Mbps
10Mbps - 50Mbps
80Mbps - 200Mbps

Bandwidth Usage

Figure 2: Peak Bandwidth Usage by Individual User

This layered architecture is further complicated by field
deployment. To map the different layers to building or campus
structures, the concepts of Main Distribution Frame (MDF) and
Intermediate Distribution Frame (IDF) are commonly used.
The design of IDF is limited by a few factors including cable
length limit, power consumption, cooling, and density of end
users. Those factors are usually incorporated into building
designs by architects to compete with the maximum usable
square footage of each building.

Figure 2 illustrates the percentage of user population for
each peak bandwidth usage category. The results show that
most users have a peak bandwidth usage less than 50Mbps and
almost all users take less than 80Mbps. A further investigation
discovers that those who reached more than 50Mbps
bandwidth seemed to be file transfer from the enterprise
distributed file system and file download from enterprise
network.

The fundamental limitation in this layered architecture is
mainly due to the characteristics of the copper cable that is
commonly used to connect the workstation and access layer
switches. For example, the maximum length for a copper cable
link between two active devices is 328 feet; complexity in
cable construction to ensure high radio frequencies; amount of
copper and plastic used in copper LAN; and installation rules
of copper cable.

We measure the bandwidth utilization of applications using
HTTP protocol by running the real application and real
workload. The typical enterprise applications and their network
bandwidth consumption are shown in Table 1.

Another main limitation of the traditional LAN architecture
is the complexity of network management. For example,
setting up a Virtual LAN (VLAN) in layered infrastructure
requires changes of multiple switches and creates complex
mapping between the ports and switches. This process is very
labor intensive and prone to human error. Monitoring of
network traffic will need to be deployed across all the layered
switches, if both in-network and out-network packets are to be
captured.

Application
VoIP phone

Configuration
64Kbps setup

Bandwidth utilization
~ 100 Kbps

Video surveillance High Definition MPEG4

~ 6 Mbps

eMail

2 minutes refreshing

50 ~ 500 Kbps

Web Browsing

Non video web sites

50 ~ 300 Kbps

Video conference

720p

~ 2Mbps

Cloud access

enterprise application

50 ~ 200 Kbps

Virtual desktop

1080p full screen display

500 Kbps ~ 2 Mbps

TABLE I.

TYPICAL APPLICATION BANDWIDTH CONSUMPTION

The measurement results imply that enterprise traffic is
very much hub-and-spoke-based, with nearly all application
resources residing centrally and being accessed remotely or via
other types of non-local protocols. Gartner Research predicts
that the trend of less local traffic will continue and by 2016 less
than 10% traffic will be local [3]. With understanding of such
traffic flows, there is a strong suggestion that the usage patterns
that spawned decentralized computing and gave birth to LANs
are shifting back to a centralized model and this usage
demands a new architecture and economic justification. Our
observation shares similarity with some prior work [4].
III. RETHINK ACTIVE SWITCH-BASED LAN ARCHITECTURE
Traditional LAN infrastructures are based on layered active
switches commonly referred to as two-tier or three-tier design.
In a typical enterprise LAN setup, a group of individual
computers connect to a hub or an access layer switch. The

IV. PASSIVE OPTICAL LAN AS AN EMERGING ARCHITECTURE
Passive Optical LAN overcomes all the limitations found in
traditional copper-based Ethernet implementations: the optical
fiber cable used in Passive Optical LAN can travel for a
distance of up to 20km ~ 30 km; the fiber cable structure is
much lighter than copper-based cables; the use of bendinsensitive fiber radically diminishes bend radii therefore
diminishing cable tray and pathways requirements; the passive
nature of the intermediate splitter eliminates the need of power
and cooling; the single management console provides
consolidated access to all devices and network ports in the
network.
The main components in Passive Optical LAN architecture
are Optical Network Terminal (ONT), passive splitter, and
Optical Line Terminal (OLT). The ONT connects computer
devices into the Passive Optical LAN network via the Ethernet
ports on the unit. Electrical signals from computer devices get
converted to optical signal in the ONT. Optical splitters simply
split the light signal multiple ways to ONTs and transmit the
multiplexed signal to the OLT. The OLT aggregates all optical
signals from ONTs and convert them back to electrical signals
for the core router. The OLT may also have a range of built-in
functionalities such as integrated Ethernet bridging, VLAN
capability, end-user authentication and security filtering etc.
Figure 3 shows the corresponding layers in traditional LAN
architecture and in Passive Optical LAN architecture. Switches
in the access layer and building aggregation layer are replaced

2015 IFIP/IEEE International Symposium on Integrated Network Management (IM2015): Short Paper

891

by a passive optical splitter and those two laayers do not exist
any longer in Passive Optical LAN architecturre.
An OLT may support 8 ~ 72 fiber ports with each port
connecting a fiber cable to the splitter. The spplitter can support
different splitting ratios with 1:32 or less being the
recommended split ratio. Therefore, each OLT
T port supports 32
ONTs. Different ONT configurations are aavailable ranging
from 2 to 24 Ethernet ports, multiple anaalog voice ports,
coaxial video ports, and even wireless suppport. If only 4
devices are attached to each ONT, an OLT w
with 72 ports will
be able to support 9216 devices.

A. Total Cost of Ownership (TCO)
We use a hypothesis model dev
veloped by multiple parties
in the industry to calculate the total cost of ownership. The cost
of each category is listed in Figurre 4. The solution using a
Passive Optical LAN network hass a capital expenditure of
$465,588 while the cost for a cop
pper network is $736,224,
resulting in 37% savings. The Passive Optical LAN network
also has lower annual operating expense
e
with $100,598 vs.
$167,709, a 40% saving. The net TCO for Passive Optical
LAN technology for one year is abo
out $566,186, and over five
years will be $891,898. On average,, the total cost of ownership
for using Passive Optical LAN technology over 5 years will be
38% less than traditional copper LA
AN networks.
B. Capital Expenditures
The main cost in capital expeenditure is to acquire the
equipment and initial installation. We
W calculate the cost in three
categories: lateral cost, riser closet or
o IDF, and main equipment
room or MDF. Lateral cost includess material and installation of
CAT6 cables from the IDF and walll plates if using a traditional
LAN network, or material and instaallation of fiber cables from
the IDF and ONT units if using
g a Passive Optical LAN
network. There are a few factors in the capital expenditure that
need to be highlighted:

Figure 3: Traditional LAN Architecture vs. Passivee Optical LAN
Architecture
DEPLOYMENT
V. EXPERIENCE IN PASSIVE OPTICAL LAN D

In the past years, the IBM Site and Facilitiies Services team
has successfully deployed a number of Passiive Optical LAN
projects including both new installationn and existing
infrastructure upgrades, yielding millions off dollars in total
cost of ownership savings for customers. The m
main benefits our
customers have realized include lower capital expenditures and
operational expenditures, easier network maanagement, more
usable floor space, less building designn steps, power
consumption and cooling cost.
We share a model of a mid-sized with thhree floors in the
building; each floor has 2 IDF/riser closets. Eaach IDF supports
100 cubicles or 8 office/conference room
ms. Each cubicle
requires 2 Power over Ethernet (POE) portss and each office
needs 4 POE ports. There are 15 wireless aaccess points per

Material costs
ble is significantly less than
Material used in fiber optic cab
material used in copper cable. If
I we only calculate the
horizontal distribution cables, one half
h or even one third of the
cables are needed to provide the same number of Ethernet
outlets. The fiber cable itself is much
h thinner than the CAT 5 or
CAT 6/6A cables. In this installatio
on case, the Passive Optical
LAN solution resulted in a reduction of 3,000 pounds less
plastic than CAT 6 cables and 10,5
500 pounds less than CAT
6A cables, and a reduction of 3,000
0 pounds copper. The glass
used in fiber only weights about 15 pounds
p
in this solution.
Construction costs
The fiber cable infrastructure costs substantially less to
install than a copper-based LAN sy
ystem, since there are fewer
cables to install. Improved terminatiion tools and the possibility
of using pre-connected fiber also hav
ve significantly reduced the
cost of fiber installation. Fiber caables are much lighter and
require fewer cables per Ethernett port, making the wiring
structure simpler that may result in needing a J holder instead
of a traditional ladder channel.
The impact of capital expenditu
ure can be more sensitive in
existing infrastructure upgrades wh
here old cables need to be
removed before installing the neew ones. Copper Ethernet
cabling has experienced a few gen
nerations with new ones on
the horizon already. This has an imp
pact on all enterprises but is
extremely significant in businessees where each upgrade is
mandatory or commonly practiced
d such as in the healthcare
industry.

Figure 4: Total cost of ownership comparrison

floor. The total number of cable drops is 1440..

892

C. Operational Expenditures
Operational cost for LAN inffrastructure is one of the
biggest expense sources for all en
nterprises. We discuss and
compare the cost of the two solu
utions in terms of network
management, floor space requiremen
nt, power and cooling cost.

2015 IFIP/IEEE International Symposium on Integrated Network Management (IM2015): Short Paper

Network Management
The typical network maintenance tasks include:
•

Capacity management such as provisioning a new
workstation / port, removing a disposed workstation / port,
creation and modification of IP addresses and virtual LAN
setup, configuration of any L2 services like quality of
service etc.

•

Upgrades and patches to keep all the hardware, firmware
and management software up to date and replacement of
defective devices.

•

Regular care such as monitoring and fixing any alerts or
defects, checking and fixing any problems within the
chassis.

•

Testing and certification of all devices, cables and
connections.

•

Management equipment and software.

In addition to maintenance costs, expenses on service
contract, training courses, and sparing need to be included,
which are usually offered by network solution providers and/or
device vendors as a certain percentage of the entire contract
value.
Costs for capacity management, upgrades and patches, and
management equipment are significantly lower with a Passive
Optical LAN solution than with a traditional solution, since the
Passive Optical LAN network eliminates all the active switches
in the access layer and distribution layer. The only active
device in the Passive Optical LAN solution that requires
maintenance and provides management interface is the OLT.
Using the built-in provisioning features provided in the OLT, it
provides a single interface for well-defined control and
monitoring of the quality of service offered to individual users
of the shared infrastructure, including dedicated bandwidth and
bandwidth restrictions.
The remaining expense sources, floor space, power and
cooling, are essentially the major contributors of the cost
saving in total operational expenditures. The Passive Optical
LAN solution reduces floor space used for networking by
approximately 69% and reduces the cooling energy cost by
approximately 74% since all the splitters are passive and
require no cooling.

building floor is about 20,000 square feet, which traditionally
requires 100 – 200 square feet to hold the two IDF closets.
Such floor space can be easily converted to usable rooms
contributing to extra revenue generation.
The saving for large campus with multiple floors, or
multiple buildings is bigger than a small campus. Since each
fiber cable can reach up to 12 miles from the main switch
closets to the user outlets, it is feasible to have only one full
size MDF in one building to serve the entire campus.
Power and cooling consumption reduction
There are many aspects of power consumption reduction in
the Passive Optical LAN solution. Power savings resulting
from the reduction of cooling and electronic devices in IDF
closets is quite straightforward. This reflects a reduction of
power circuits, HVAC equipment provided by the building
infrastructure, and operational savings with reduced cooling
loads. We have observed an approximate 74% cost reduction
from the elimination of cooling in IDF closets.
Besides the energy savings from minimal cooling
requirements, most Passive Optical LAN equipment is
inherently energy efficient. Because a large number of Ethernet
endpoints can be supported from one single OLT, ranging from
a few hundred to a few thousands depending on the ports the
OLT has, power consumption of the OLT is much lower than a
comparable traditional distribution switch. Similarly ONTs
also consume less power per Ethernet port than a comparable
intermediate workgroup switch. In this deployment, we have
observed about 26% less power consumption in the Passive
Optical LAN network.
VI. CONCLUSION
With the commercialization of Passive Optical LAN
technology, it has quickly demonstrated the advantages as one
of the most revolutionary technologies in the networking era. It
adequately accommodates all the demands required by modern
enterprise applications with much lower cost than traditional
LAN implementation. The energy-efficient nature of the
solution inherently qualifies as a green technology. The rich
built-in advanced capabilities provide a seamless enablement
for smart buildings and campuses.
REFERENCES
[1]

Floor space saving
The Passive Optical LAN solution eliminates the need for a
dedicated IDF because the passive nature of the splitters and
the long distance capability of fiber cable. Splitters do not
require any cooling so they can be put in a very small closet on
the floors, in enclosures behind walls, shared with the electrical
closets, in raised floor architecture, or even in the ceiling space.
The only communication closet needed for Passive Optical
LAN is the main distribution frame. In this deployment, each

[2]

[3]
[4]

Benson, Theophilus, Aditya Akella, and David A. Maltz. "Network
traffic characteristics of data centers in the wild." Proceedings of the
10th ACM SIGCOMM conference on Internet measurement. ACM,
2010.
Karagiannis, Thomas, and Richard Mortier. "Address and traffic
dynamics in a large enterprise network." Local and Metropolitan Area
Networks, 2008. LANMAN 2008. 16th IEEE Workshop on. IEEE, 2008.
Gartner Research, Rethinking LAN Switching Architectures, 25
February 2011, G00210808
Pang, Ruoming, et al. "A first look at modern enterprise traffic."
Proceedings of the 5th ACM SIGCOMM conference on Internet
Measurement. USENIX Association, 2005.

2015 IFIP/IEEE International Symposium on Integrated Network Management (IM2015): Short Paper

893