GMPLS IMPACT ON NETWORK PROVIDER’S OPEX WHEN
COMPARED TO TRADITIONAL APPROACHES
S. Pasqualini
1
, S. Verbrugge
2
, A. Kirstädter
1
, A. Iselt
1
, D. Colle
2
, M. Pickavet
2
, P. Demeester
2
1
Department of Information and Communications, Corporate Technology,Siemens AG,Otto-Hahn-Ring 6,D-81730
Munich,Germany
2
Department of Information Technology (INTEC), Ghent University – IMEC – IBBT, Sint-Pietersnieuwstraat 41, B-9000,
Ghent, Belgium
Keywords: OPEX, network management, network operations, business case, GMPLS
Abstract: This paper provides a detailed analysis and modelling of the Operational Expenditures (OPEX) for a
network provider. The traditional operational processes are elaborated and the expected changes when using
GMPLS are described. GMPLS is promoted as a major technology for the automation of network
operations. It is often claimed to allow the reduction of OPEX. However, detailed analysis and quantitative
evaluation of the changes induced by such technologies is rare. In this paper we quantify the cost reduction
potential of GMPLS. In case of a traditional network, we show an important impact of the used resilience
scheme on the expenses directly related to continuous costs of infrastructure (floorspace, energy,…) and on
the planning and reparation costs. Concerning the service provisioning costs, we show that GMPLS
introduction leads to a reduction in the order of 50% of the OPEX cost compared to the traditional case.
1 INTRODUCTION
This paper presents a quantitative study on the
operational expenditures for a transport network
operator. We evaluate how Generalized
Multiprotocol Label Switching (GMPLS)
technologies impact transport network operators
processes and costs, when compared to traditional
approaches.
During the last years the main focus of transport
network evolution was on increasing transport
capacities and on introducing data networking
technologies and interfaces, e.g. Gigabit Ethernet.
This evolution is complemented by ongoing
initiatives to reduce the operational effort and
accordingly the costs of network operations.
GMPLS together with standardized interfaces like
UNI/NNI automate the operation of telecom
networks (Banerjee, 2001). They allow to efficiently
provide services and to improve the resilience of
networks. For the service provisioning there is the
new paradigm of user initiated service provisioning
(also known as switched connections) where the
client can setup connections without operator
interaction. This does not only speed up the
provisioning process, but also reduces effort for the
network operator.
Currently the approach of using a distributed
control plane for network functions like link
management, failure restoration, or provisioning
services such as leased lines is followed by several
initiatives and standardization bodies including ITU,
OIF and IETF. In this paper we do not distinguish
the details of these approaches but generally assume
a control plane supporting automation of network
operations. We use the term GMPLS to refer to any
kind of control plane according to one or several of
these standards.
2 APPROACH
The total expenditures of a company can be split
in two parts: the capital expenditures (CAPEX) and
the operational expenditures (OPEX). CAPEX
contribute to the fixed infrastructure of the company
and are depreciated over time. They are needed to
expand the services to the customers. OPEX do not
contribute to the infrastructure itself and
consequently are not subject to depreciation (Soto,
2002). They represent the cost to keep the company
185
Pasqualini S., Verbrugge S., Kirstädter A., Iselt A., Colle D., Pickavet M. and Demeester P. (2005).
GMPLS IMPACT ON NETWORK PROVIDER’S OPEX WHEN COMPARED TO TRADITIONAL APPROACHES.
In Proceedings of the Second International Conference on e-Business and Telecommunication Networks, pages 185-192
DOI: 10.5220/0001420001850192
Copyright
c
SciTePress
operational and include technical and commercial
operations, administration, etc. This paper focuses
on the impact of GMPLS on the OPEX in an
operational network, i.e. one that is up and running
(Verbrugge, 2005(1)). We therefore don’t consider
initial installation and network extension costs. All
infrastructure is counted as CAPEX, as suggested in
(Eurescom, 2000). For the traditional network, we
assume that it provides end-to-end services. The
GMPLS network additionally offers dynamic
services.
Network operation comprises all the processes
and functions needed to operate a network and
deliver services to customers. That includes the sales
and marketing processes, the various support
functions, as well as provisioning and monitoring of
the network, and the corporate processes in general.
Thus the significance of a reduction in OPEX cannot
be downplayed.
In this study we want to perform a process-based
quantitative analysis of OPEX and the reductions
expected for operators using GMPLS in their
transport network. The study is based on the OPEX
model defined in (Verbrugge, 2005(1)). Starting
from this very comprehensive model, we evaluate
which operations become more or less expensive
when the used technology is GMPLS instead of the
traditional static transport network. Apart from the
operations, particular attention has also to be paid to
the processes’ branches. The probability of each
branch of the processes’ flow has also to be
extrapolated when considering the new technology.
Based on this qualitative modelling, quantitative
results can be calculated. The normal cost of each
operational step is the one assumed in the base
OPEX model, for the traditional approach.
Combining this cost and the qualitative variation, the
new cost can be extrapolated. In this way the
incremental costs/benefits from using GMPLS can
be obtained.
3 TRADITIONAL PROCESS
STRUCTURE
In general, the introduction of GMPLS as well as the
considered resilience scheme will influence the cost
structure of network operators in many ways. The
next sections describe the processes being affected.
A more generic description can be found in
(Pasqualini, 2005). For traditional networks, we
consider 1+1 protection (two connections are setup
simultaneously, one of them being used as backup).
3.1 Continuous and Recurring
Processes
3.1.1 Continuous cost of infrastructure
The cost to keep the network operational in a failure
free situation is the first important cost in this
category. We call this the telco specific continuous
cost of infrastructure. It includes the costs for floor
Figure 1: Typical service provisioning process
ICETE 2005 - GLOBAL COMMUNICATION INFORMATION SYSTEMS AND SERVICES
186
space, power and cooling energy and leasing
network equipment (e.g. fiber rental). The
continuous cost of infrastructure follows the same
trends as CAPEX. Additional network equipment
installed as a backup for failures (1+1 protection)
leads to higher costs for floor space and energy.
3.1.2 Routine Operations
It is the cost to maintain the network or to operate
the network in case a failure can occur. The actions
involved include direct as well as indirect (requested
by an alarm) polling of a component, logging status
information, etc. Also stock management (keeping
track of the available resources and order equipment
if needed), software management (keeping track of
software versions, and install updates), security
management (keeping track of people trying to
violate the system and block resources if needed),
change management (keeping track of changes in the
network, e.g. a certain component goes down) and
preventive replacement are included.
3.1.3 Reparation
Reparation means actually repairing the failure in
the network, if this cannot happen in routine
operation. Reparation may interrupt unprotected
services. The actions involved in the reparation
process are diagnosis and analysis, technicians
travelling to the failure location, actual fixing of the
failure and testing whether it is actually repaired. In
a unprotected network, we expect that reparation is
more expensive because of the additional effort to
reroute the affected traffic.
3.1.4 Operational Network Planning
It is an ongoing task and includes all planning
performed in an existing network which is up and
running, including day-to-day planning, re-
optimization, planning upgrades. The costs for
planning are smaller for the unprotected network,
because backup scenarios do not need to be planned
for. We also assume that the non-GMPLS network is
managed by one standard centralized Newortk
Management System (NMS) per administrative
domain, calculating routes and monitoring alarms
but still requiring human intervention to configure
the equipment.
3.1.5 Marketing
With marketing we mean acquiring new customers
to a specific service of the network operator. The
actions involved are promoting a new service,
provide information concerning pricing, etc.
3.2 Service Management Processes
We investigated the five most technology dependant
processes within the traditional structure of network
operators considering the interactions and operations
of sales department (SD), administration (AM),
project management (PM), network operation (NO)
and external suppliers (ES).
3.2.1 Service Offer
The sales department negotiates the terms and
conditions of the offer with the customer and checks
whether the connection request can be handled by
the standard mechanisms and infrastructure. In case
of non-standard connection inquiries, separate
projecting (PM) is triggered for the various domains
(local, internal, external), and missing equipment
(cards, fibers, etc.) is ordered, causing additional
effort and delay. The projecting results then define
the price calculation (SD), as well as the delay
necessary to set up the service. Then the offer sent is
to the customer.
3.2.2 Service Provisioning
After the contract has been accepted the service
delivery process starts (fig. 1). The sales department
handles the contract administration, then the order is
split it into work packages according to the network
domains involved (PM). After providing the
connection (NO), an end-to-end test is conducted
and customer care, billing and alarm management
are activated (AM). Finally, a delivery report is
issued by the sales department to the customer. In
case of 1+1 protection, the cost increases
significantly since it is required to setup almost two
connections.
3.2.3 Service Cessation
At the end of a contract or on cessation request by
the customer the cease process triggers (via PM) the
deactivation of the circuits (NO), followed by the
recovery of equipment by field technicians (NO).
SD is informed about the expected cessation and the
final bill is sent out (AM).
3.2.4 Service Move or Change
Moving or changing a connection is the most
complex task since it involves all three previous
processes: Contract update, new connection setup
and release of the previous connection. The
customer’s request for change is handled by the
sales department as a service offer process, checking
again for the availability of resources. The sales
GMPLS IMPACT ON NETWORK PROVIDER’S OPEX WHEN COMPARED TO TRADITIONAL APPROACHES
187
department then generates orders for the service
provisioning and cease process that are implemented
through coordination from the Project Management
department. In the same time the client is receiving
updates on the new installation.
4 IMPACT OF GMPLS ON
OPERATIONAL PROCESSES
From the main operational processes described
above, several are impacted by the use of GMPLS.
We consider that the use of GMPLS impacts the
processes in two ways: the way connections can be
setup and managed, and the wider variety of
resilience schemes it promotes. Regarding the
resilience schemes, we will now assume that in a
GMPLS network shared mesh protection will be
used instead of 1+1 protection. For shared
protection, two connections are also planned, but
only one is actually provisioned, the second one
being provisioned only when the first one has failed.
The advantage is that the resources of the latter can
be shared among several backup connections,
leading to more efficient resource utilisation. This
obviously impacts CAPEX, and as a consequence
the continuous and recurring processes, which tend
to decrease together with CAPEX. Shared protection
could also be used in a non GMPLS network, but we
consider it is more applicable in the GMPLS case
because the backup path can be provisioned and
switched much faster.
4.1 Continuous and Recurring
Processes
4.1.1 Continuous cost of infrastructure
The continuous cost of infrastructure will be
impacted by the amount and the type of the network
components used. With GMPLS the network usually
allows mesh-based restoration, where less backup
capacity is required, which in turn leads to less
network components. The cost to power, cool and
host this equipment will therefore also decrease.
4.1.2 Routine Operations
The cost of the routine operation (maintenance cost)
depends also on whether the network is
automatically switched or not. The use of GMPLS
influences the routine operation costs because
(re)configuration after replacement of equipment can
happen faster. The replacements in the routine
operation process are only those that can happen in
the service window. The service window indicates
the time (e.g. during the night) during which service
interrupts are contractually not considered as
downtime. As GMPLS enables faster
reconfiguration, more operations can happen during
the service window, so that the repair process needs
to be triggered less often. On the other hand,
monitoring the software and the needed software
upgrades becomes more expensive in case of
GMPLS, because its complexity drastically
increases due to the presence of the control plane. In
general, we can expect the routine operation cost to
increase a bit when GMPLS is used.
4.1.3 Reparation
As a result of using GMPLS more failures can be
fixed from the NOC, which could have a beneficial
impact on the reparation cost. On the other hand,
GMPLS leads to more complex network operation,
which might be an additional source of failures.
Rerouting of traffic happens faster: GMPLS allows
for many fast and automated restoration and
protection schemes. Isolating a fault gets cheaper
when Link Management Protocol (LMP)'s fault
management procedure is available (but LMP is
optional in GMPLS, (Banerjee, 2001)). Overall, we
expect the cost for the reparation process to decrease
in case of GMPLS. (Verbrugge, 2005(1)) gives an
overview of the repair process.
4.1.4 Operational Network Planning
Indirectly the used network technology will also
influence the cost of planning, as more complex
systems require a higher planning effort.
4.1.5 Marketing
As GMPLS technology allows to offer new services,
which are initially unknown to the customers,
additional marketing will be needed to inform the
customers. This will lead to higher marketing costs.
On the other hand, of course, it may also lead to
higher revenues.
4.2 Service Management Processes
Finally, technologies automating some of the
network operation allow to significantly reduce the
cost for service provisioning, because the signalling
can be done via standardized interfaces (User
Network Interface UNI and Network to Network
Interface NNI), without requiring manual
intervention. This means that the cost for setting up
a new connection decreases strongly. In this case,
the service offer process and provisioning process
ICETE 2005 - GLOBAL COMMUNICATION INFORMATION SYSTEMS AND SERVICES
188
will be changed fundamentally (Chahine, 2005).
Since the service delivery will now be automated
and executed on the pure machine level, correct
agreements and regulations have to be negotiated by
the sales department, and implemented well before
in the form of Service Level Agreements (SLAs).
The use of GMPLS technologies and the possibility
to offer dynamic services are strongly
interconnected issues. The strongest impact of the
dynamic services is on the pricing and billing
process. Fixed price services, e.g. leased lines, will
definitely be cheaper in pricing and billing than
dynamic services. For dynamic services it is much
more difficult (and thus more expensive) to correctly
assign costs to customer accounts. Calculating a new
price for a new service is more expensive than just
applying a traditional pricing scheme. This is
elaborated below as “negotiations” in the service
provisioning process.
4.2.1 SLA Negotiations
The process chain therefore starts with the SLA
negotiation process. Before the single services are
ordered and delivered, a contract framework
specifies in detail all sections of a generic service
template. Technical aspects like bandwidth
(minimum, burst) and its granularity, service
availability, quality of service are specified as well
as legal and organizational questions (penalties for
requirements not met, compensation, tracking and
reporting, etc.). Within the network operator this is
accompanied by forecasts (SD), Planning (PM), and
adaptation of the infrastructure (NO). For new
customers, it also involves connecting the
customer’s location with the network (which is
carried out in the service provisioning process for
the non-GMPLS case).
4.2.2 Service Provisioning
After this framework has been set up the, service
delivery process can be simplified due to the
introduction of standardized interfaces (fig. 2).
External signalling at the UNI is directly forwarded
to the call control (PM) that splits it into RSVP
signalling for each domain (NO). Manual
intervention is necessary to set up the connection
completely only if no positive responses were
received. After database update (AM), customer care
is informed, and billing and alarm management are
activated. At the end of this process, the client
receives the delivery report.
4.2.3 Service cessation
In the GMPLS case, the cessation request is also
received via the UNI. The cease process then
triggers the sales department to assess the cessation
request and trigger the billing and confirmation of
cessation to the client. On the physical side, the
network operation centre is requested to cease the
physical connection. Once the connection is
Figure 2: Service provisoning with GMPLS.
GMPLS IMPACT ON NETWORK PROVIDER’S OPEX WHEN COMPARED TO TRADITIONAL APPROACHES
189
released, this is confirmed to the project
management and the order is closed.
4.2.4 Service Move or Change
The GMPLS-modified move and change process is
initiated by the customer requesting a move and
change. The sales department transforms this request
directly to check for possibility of request and
availability of resources within the framework of the
SLA. Once the check of SLA is done, the customer
is sent the offer to accept or refuse it. If the customer
accepts the sales department generates orders for the
service delivery process and cease process that are
implemented with coordination from the project
management department. At the same time the
customer is receiving updates on the new
installation.
5 QUANTITATIVE RESULTS
5.1 Service management
For each of the processes described in (Verbrugge,
2005(1)), costs have been assigned to the process
steps (boxes in the figure) and a probability to the
branches. We focused on labour costs, expressed in
terms of hours required to carry out the task
described in the box. Then we calculated the hourly
fully accounted cost of each kind of employee, and
multiplied it by the number of hours. As suggested
in (Van Steen, 2004), we distinguish several
personnel categories: sales, administration,
engineers and technicians (in the NOC or field
technicians). Each department displayed in fig 1 and
fig. 2 is composed of one type of employee, except
the NOC where engineers, technicians and field
technicians have been considered. Summing up costs
for all steps gives then an upper bound estimate of
the overall cost of a given process. Cost and effort
figures for the current network operations were
collected based on surveys and interviews with
several carriers. From these figures we extrapolated
the figures for the new GMPLS process model.
In the case of a typical incumbent operator (fig. 3),
the service offer process involves expensive sales
and availability checks operations. In the end it is
nearly as expensive as the service provisioning
itself. The cease process involves nearly no work
from project management and network operations
center, which explains why it is much cheaper. The
move and change process is the combination of
service offer, provisioning and cease (in principle, it
is a little more expensive since it requires some more
coordination). Looking at the GMPLS-modified
processes, we first notice that SLA negotiations are
more expensive than the typical service offer. This is
normal since the former includes some operations
that are usually carried out in the service
provisioning process (plan, install and configure
equipment boxes). For a fair comparison, one needs
to compare the combination of service offer and
provisioning. In the case where GMPLS is used,
project management and sales are involved only
once - when the SLA is setup - leading to substantial
savings. Another advantage is that the same SLA
can serve for several services. So once the SLA is in
place, provisioning several services with GMPLS
costs much less.
5.2 Overview of all Operational
Processes
In order to compare the costs of all processes, a
specific case needs to be studied. We consider an
optical (WDM) network carrying 2.5 Gbps leased
lines and calculate the costs over one year.
Although, in a realistic network, leased lines would
probably be offered via SDH or OTN over WDM,
we focus on this architecture for the sake of
simplicity. The topology is the reference German
network (Hülsermann, 2004) with 17 nodes and 26
links and the associated traffic for 2004. This traffic
leads to 1214 services for one year, 80% of which
we assume to be standard services. We also assume
that there are no service cessations or move.
5.2.1 Equipment
For each type of equipment (WDM line systems,
optical amplifiers, transponders, unequipped OXC
Figure 3: Normalized costs for a non standard service.
0,00
20,00
40,00
60,00
80,00
100,00
t
ra
d
i
ti
o
na
l
se
rv
i
c
e
o
ff
e
r
SLA negociations
traditional provisioning
G
MPL
S se
rv
ice
p
ro
v
i
s
i
o
ning
t
radition
a
l
s
ervice c
e
ase
GM
PLS
s
e
rvice ce
a
s
e
Cost
Sales Administration Project management NOC
ICETE 2005 - GLOBAL COMMUNICATION INFORMATION SYSTEMS AND SERVICES
190
itself) we take the following parameters into
account: price, and Mean Time Between Failure
(MTBF). Detailed values can be found in
(Verbrugge, 2005(2)). The equipment life time is set
to 10 years. In order to be able to calculate the
yearly capital equipment costs, the total equipment
cost obtained from dimensioning is divided by this
life time.
5.2.2 Failure probabilities
We assume two types of alarms in the network:
preventive alarms and failure alarms. Based on
(FCC, 2004), we assume 39% of all alarms to be
preventive alarms. Considering failure alarms, we
differentiate between CPE problems, external
problems (power disruption, etc.), misconfigurations
or software failures, and hardware failures (incl.
cable cuts). The probabilities of these problems are
specific for an optical network (Verbrugge,
2005(3)). Based on this and the MTBF values, we
determined a total of 1171 failures and 749
preventive alarms per year over the entire network.
5.2.3 Estimated yearly OPEX
With the above assumptions, we have been able to
estimate yearly costs for this network scenario, as
shown in fig. 4. In this respect, we need to point out
that our study only considered OPEX for a network
which is up and running. Also, including the costs
for first-time installation in the process-based costs
would have probably changed the picture.
This case study on a specific network scenario
allows to compare the process-based costs over one
year. We see a significant decrease for service
management processes in the GMPLS case. The
other processes’ costs remain of the same order in
both scenarios, and could be investigated further.
But this case study already allows to have an
evaluation of cost reduction that could be awaited
from the automation of some processes, and
compare it to other OPEX categories.
6 CONCLUSION
In this paper we show that most network operators'
processes are similar and can be modeled quite
generically. When introducing GMPLS, OPEX cost
reductions in the order of 50% compared to
traditional operations can be identified for service
offer and provisioning together, as shown in fig.3.
Based on these results the introduction of GMPLS
can generally be recommended to significantly
reduce OPEX. This advantage can even be
improved, if all network domains and all network
layers support interworking GMPLS control planes
and hereby also reduce the operational cost for end-
to-end connections across multiple operators'
domains. When comparing the use of GMPLS with
shared protection capacity to the traditional network
without control plane, using 1+1 protection, some
other cost effects become clear. The amount of
backup capacity in the network has a direct effect on
the CapEx costs and the OpEx part which is directly
related to continuous costs of infrastructure
(floorspace, energy…) A similar trend is seen for the
costs of service provisioning, which is most
expensive in case of 1+1 protection (more
connections need to be set up). Also planning costs
grow with the amount and complexity of failure
scenarios that need to be planned for. Finally, also
the reparation process is impacted by the
dimensioning, because more equipment leads to
more possible failures. On the other hand, the
availability of backup capacity strongly reduces the
time to get the network operational after the
occurrence of a failure and therefore reduces this
cost.
This paper reports preliminary results, and future
work will focus on refining the model and the input
data by further questioning network operators.
ACKNOWLEDGEMENTS
This work was supported by the European
Commission through IST-NOBEL and ePhoton/One.
And by the Institute for the Promotion of Innovation
through Science and Technology in Flanders (IWT-
Vlaanderen) through a PhD grant for the second
author, and the GBOU-project IWT 010058.
REFERENCES
Banerjee, A., 2001. Generalized multiprotocol label
switching: an overview of routing and management
0,00E+00
2,00E+05
4,00E+05
6,00E+05
8,00E+05
1,00E+06
1,20E+06
1,40E+06
1,60E+06
1,80E+06
2,00E+06
ro
uti
ne operation
repair process
service
offer
s
er
vi
ce
p
r
o
v
isi
o
ning
mar
ke
ting
planning
Traditional (1+1 protection) GMPLS (shared path protection)
Figure 4: Yearly OPEX (€) for all processes
GMPLS IMPACT ON NETWORK PROVIDER’S OPEX WHEN COMPARED TO TRADITIONAL APPROACHES
191
enhancements. In IEEE Communications Magazine,
pp. 144-150, Volume 39, Issue 1.
Soto, O., 2002. Network planning – business planning. In
ITU/BDT-COE workshop.
Verbrugge, S., 2005 (1). Modeling Operational
Expenditures for Telecom Operators. In ONDM 2005,
9th Conference on Optical Network Design &
Modelling.
Eurescom, 2000. Extended investment analysis of
telecommunication operator strategies. In Project
P901-PF, Deliverable 2, vol. 2 of 4, annex A.
Pasqualini, S., 2005. Influence of GMPLS on Network
Providers’ Operational expenditures – A Quantitative
Study. In IEEE Communications Magazine, Volume
43, Issue 7.
Chahine, R., 2005. Operational Cost Reduction using
ASON/ASTN. In OFC 2005, Optical Fiber
Communication Conference and Exposition.
Van Steen, B., 2004. Service Provider Study: Operating
Expenditures. In MEF, Metro Ethernet Forum.
Hülsermann, R., 2004. A Set of Typical Transport
Network Scenarios for Network Modelling. In 5
th
ITG
Workshop on Photonic Networks.
Verbrugge, S., 2005 (2). Impact of Resilience Strategies
on Capital and Operational Expenditures. In 6
th
ITG
Workshop on Photonic Networks.
FCC, 2004. In http://www.fcc.gov/wcb/iatd/stats.html .
Verbrugge, S., 2005 (3). General Availability Model for
Multilayer Transport Networks. In DRCN2005, 5th
International Workshop on Design of Reliable
Communication Networks.
ICETE 2005 - GLOBAL COMMUNICATION INFORMATION SYSTEMS AND SERVICES
192