Conference PaperPDF Available

Inter-Operator Dynamic Capacity Sharing for Multi-Tenant Virtualized PON


Abstract and Figures

As the capacity of the optical access networks increases, the case for sharing this capacity amongst multiple operators becomes stronger. In addition to the capital and operating expenditure savings that infrastructure sharing can provide for the operators, providing a higher degree of infrastructure customization will be a strong motivator for operators to participate in the sharing ecosystem. Thanks to the network virtualization technologies, the higher degree of control over the infrastructure can be a motivator for the new virtual operators to join. Given this control, each operator will make decisions for their share of the resources according to their policies. However, when it comes to the infrastructure provider to aggregate all these decisions, ensuring trust becomes vital. It is essential to study the incentives of all the operators and design a sharing mechanism that incentivizes truthfulness. In this paper, we propose such an auction mechanism to monetize the exchange of excess capacity between network operators to increase resource efficiency. The proposed market design is based on a sealed-bid VCG auction for homogeneous multi-item goods with a reserve price. Through market simulations, we show that our proposed market design can achieve all the fundamental economic properties of a market including, truthful value announcing, individual rationality and weak budget balance.
Content may be subject to copyright.
Inter-Operator Dynamic Capacity Sharing for
Multi-Tenant Virtualized PON
Nima Afraz, Amr Elrasad, Hamed Ahmadiand Marco Ruffini
CONNECT Center, Trinity College Dublin, Ireland
Email: {nafraz, elrasada, ruffinim}
School of Electrical and Electronic Engineering, University College Dublin, Ireland
Abstract—As the capacity of the optical access networks
increases, the case for sharing this capacity amongst mul-
tiple operators becomes stronger. In addition to the capital
and operating expenditure savings that infrastructure shar-
ing can provide for the operators, providing a higher degree
of infrastructure customization will be a strong motivator
for operators to participate in the sharing ecosystem.
Thanks to the network virtualization technologies, the
higher degree of control over the infrastructure can be a
motivator for the new virtual operators to join. Given this
control, each operator will make decisions for their share
of the resources according to their policies. However, when
it comes to the infrastructure provider to aggregate all
these decisions, ensuring trust becomes vital. It is essential
to study the incentives of all the operators and design a
sharing mechanism that incentivizes truthfulness. In this
paper, we propose such an auction mechanism to monetize
the exchange of excess capacity between network operators
to increase resource efficiency. The proposed market design
is based on a sealed-bid VCG auction for homogeneous
multi-item goods with a reserve price. Through market
simulations, we show that our proposed market design
can achieve all the fundamental economic properties of
a market including, truthful value announcing, individual
rationality and weak budget balance.
Passive optical network (PON) is one of the most
widely deployed last mile optical access technologies.
A PON is based on a point to multi-point architecture
that enables the service providers to serve the end-points
without a need to provision fiber cables individually for
each end user. The major components of a PON includes:
an optical line terminal (OLT) in service provider’s
central office, multiple optical network units (ONUs)
near the end-users and the fibers and splitters as the
optical distribution network (ODN).
Multi-tenancy in optical access networks is attracting
more research as the high capacity fiber optic networks
are being deployed in bigger scales. The objective is to
utilize this enormous capacity by enabling the infras-
tructure operators (InPs) to lease parts of their owned
infrastructure to secondary network operators. Further-
more, in some cases, such network sharing is enforced
by telecommunications regulators.
The Broadband forum (BBF) standardization body
considers sharing of the access network infrastructures
as the only way for operators to sustainably reduce
network cost and scale the network [1]. Access sharing
addresses the shortcomings of the traditional “bitstream”
approach and the need for virtual network operators
(VNOs) to have more flexibility and better control of
resources. The concept of multi-tenancy would be more
attractive to the network operators when they can have
adequate control over the infrastructure to satisfy their
users’ requirements. Currently, this can be achieved
thanks to the virtualization technologies introduced to
telecommunication networks empowered by software
defined networking, and its correlated features such as
control plane centralization and network programmabil-
ity. Virtualization can provide heterogeneous networks
with on-demand customizability and empower dynamic
management of resources [2].
Virtualization of network functions can facilitate
multi-tenant scenarios by providing immediate access
to network functions to VNOs without any intervention
from the infrastructure provider. Thus relieving VNOs
from both owning and operating the physical network
infrastructure while having the luxury of a flexible and
customizable virtual infrastructure.
In the context of Passive optical networks (PON),
which is being considered as one of the most promising
access network options for a wide range of services from
residential fiber to the home to mobile backhaul and
fronthaul [3], virtualization of the OLT and ONU can
bring considerable flexibility to the PON. The OLT is
responsible for management of the ONUs, framing of
the data and scheduling the down/up stream traffic in a
time division multiplexing (TDM) manner.
In a multi-tenant PON scenario, a network operator
may have an incentive to share the infrastructure with
other operators – possibly even its competitors – in return
for lower CAPEX and OPEX. Nevertheless, taking into
account that these network operators may offer different
services with very divergent requirements, the absence of
adequate control over the OLT’s features can be a major
obstacle for the realization of the multi-tenant networks.
As an example, fiber infrastructure is inevitably the only
long-term solution for mobile back/fronthaul [4].
In [5] the authors have studied a market scenario for
shared access network while VNOs are dynamically
selecting the wireless spectrum, antenna sites and optical
transport resources to maximize the resource efficiency.
This kind of harmonic resource allocation scenarios is
essential for smooth fixed-mobile access convergence.
Having this growing interest in using PONs for wireless
data transmission in mind, the access network should
adapt itself to be able to accommodate the new services
in PON.
Previous research [6] has established that using current
PON is not feasible for mobile fronthauling due to its
time consuming dynamic bandwidth allocation (DBA)
schemes and requires new DBA algorithms to be de-
veloped. Consequently, in a scenario that heterogeneous
service providers are supposed to share the infrastructure
one would have to settle for the other’s benefit since
the DBA is hard coded into the OLT’s hardware and
is not easily customizable. Thus, Virtualizing the DBA
mechanism can enable the network operators to use the
infrastructure as they own it. In the rest of this paper,
we will refer to the virtualized DBA as vDBA.
In our previous work [7], we introduced a multi-tenant
PON architecture enabling customized DBA implemen-
tation for the tenant VNOs. A DBA algorithm in a PON
is responsible for generating a bandwidth map dictating
the upstream transmission opportunities for each ONU
per each frame. The DBA has to generate this bandwidth
map for every frame, every 125 microseconds.
In a multi-tenant PON, each VNO will generate a virtual
bandwidth map using the vDBA mechanism and for its
share of the frame. Then the merging and final layout
of the bandwidth map is facilitated by the merging
engine while not imposing any additional scheduling
delay. However, in a scenario where each VNO has
a dedicated share of the bandwidth according to its
service level agreement with the InP, there is a high
chance that some of the VNOs will have excess unused
bandwidth. This is due to the bursty nature of data traffic.
In [7] we studied two cases of capacity sharing from a
technical perspective, to confirm its feasibility. However,
in this work, no incentives for VNOs were considered
for sharing their excess capacity.
In [8] the authors have proposed a dual service-level
agreement to meet the fairness requirements for open
access in a multi-tenant PON, mainly focusing on the
fair allocation of the resources and neglecting the sharing
incentives of the players. In this paper, we will focus
on the incentives of the market players while proposing
an auction mechanism to choose the efficient trades
and determine the pricing. To the best of the authors’
knowledge, this paper is the first to study the sharing
incentives in a multi-tenant PON network.
VNOSeller1 VNOSeller1 VNOBuyer1
Excess Excess
InP (Auctioneer)
Final Bandwidth Map
Fig. 1: Auction Market
In our model, the market consists of multiple VNOs
that can take the role of a seller or a buyer depending
on their traffic in each frame. The InP is responsible
for assigning the excess capacity of the sellers to the
eligible buyers and for book-keeping. We define the
sharing incentives for all of the players as follow:
1) The VNO with excess frame share: Can share its
surplus capacity with other VNOs in return for
a payoff that should be higher than the benefit
received by blindly assigning that frame capacity
to its own users (which have not requested it).
Besides the payoff, each VNO will receive sharing
credits, which will be used as a tie breaker in
situations where there are two VNOs bidding the
same amount.
2) The under-allocated VNO: Gets to clear its users’
buffer from backlogged data in advance for a price
less than or up to its estimation of the true value
of that frame share.
3) The InP: By motivating the VNOs to share their
excess capacity the InP can reduce bandwidth
wastage and achieve higher overall utilization,
while charging a commission fee (the budget sur-
plus) for facilitating the auction.
The dynamics of the multi-tenant PON, with each
tenant potentially providing a different service and us-
ing customized bandwidth allocation algorithms, can
be complex. This complexity, along with the bursty
nature of the traffic, can lead to highly variable capacity
demands, e.g. in one frame, the demand might be very
high from some of the VNOs and not in others. This
will make the traditional service level agreement with
pricing models based on average peak and committed
information rates inefficient, requiring a new valuation
method accounting for this dynamicity. Auctions, if well
designed are efficient economic systems based on com-
petition for achieving a good performance through value
reporting or bidding which leads to a trade happening for
a price that reflects the valuation of all of the traders. The
standard auction for multiple items takes place by sorting
the bids from high to low and allocating the item to
the highest bids. All the standard auction formats follow
this mechanism; however, the pricing principle differs
from one mechanism to other. Fig. 1 depicts the excess
frame sharing model of the multi-tenant PON. The InP
(auctioneer) facilitates the auction and communicates the
bids between the VNOs. The InP will achieve the highest
utility when the frame share is sold for the maximum
of the buyers’ true value, and that is only possible
in the case of truthful bidding. Whenever one of the
VNOs detects excess bandwidth in its pre-defined frame
share –agreed upon in the contract– will announce the
amount of excess frame share to the InP. To minimize
the communication delay overhead between the VNOs
and the InP we are to choose a sealed bid auction. In
sealed bid auctions all the traders simultaneously submit
their valuations without any knowledge of the others’
valuation, and the auctioneer holds the auction based on
the received valuation on a single round without back
and forth communications.
Different auction formats are designed to address the
truthful bidding. Vickery-Clarke-Groves (VCG) mecha-
nism [9] is one of such mechanisms which is designed
in a way that provides incentives for the buyers to
bid truthfully. The VCG mechanism satisfies all the
three essential auction properties i.e. dominant-strategy
truthful bidding, weak budget balance, and individual ra-
tionality. The truthfulness property simplifies the optimal
bidding strategy of the buyers and consequently leads
to the lowering of expenditure on resources learning
about competitor buyers’ strategies [10]. VCG tries to
exclude the traders’ announced value from the trade
price determination process. This technique eliminates
the chance of untruthful reporting from traders with the
hope of increasing their utility by paying less (as a
buyer) or receiving more (as a seller). VCG auctions
have been widely employed in telecommunication net-
works ranging from spectrum sharing [11] to resource
allocation in cloud computing [12]. In this paper, we
propose an auction mechanism that complies with the
stringent time limitations faced in our problem (e.g., as
bandwidth allocations are granted every 125 µs). The
problem we addressed is the design of a mechanism
that facilitates sharing of the excess parts of the frame
between VNOs operating a multi-tenant PON. Other
researchers [13] have proved the optimality of truthful
reporting for traders in VCG auction. The desired auction
properties in our system are:
1) Truthfulness (Strategy-proofness): An auction
mechanism is considered strategy-proof if report-
ing the true valuation for the supplied or demanded
item is a dominant strategy for all the traders, i.e.
none of the traders can increase their utility by
reporting an untruthful valuation. This is possible
through eliminating the role of the traders in the
determination of the final price so they cannot
manipulate the market through strategizing.
2) Individual Rationality: Each VNO will have a non-
negative utility i.e. no seller/buyer VNO will regret
participating in the auction.
3) Budget Balance: The expected utility of the InP
(auctioneer) is non-negative. For our model, since
the InP, as the auctioneer needs to be paid for
facilitating the auction, our desired auction needs
to be weak budget balanced i.e. the auction results
in a positive payoff for the auctioneer.
4) Efficiency: In the context of auction, perfect effi-
ciency means the ability of the auction to success-
fully conduct all the efficient trades until all the
demands are satisfied. Trade between a seller and
a buyer is efficient if the seller’s value is lower
than the buyer’s.
The auction consists of a set of M=
{m1, m2, ..., mi}sellers (VNOs with excess capacity)
and a set of N={n1, n2, ..., nj}buyers (VNOs with
excess demand) and one auctioneer (the InP). We
consider an XGS-PON [14] with 10-Gbps symmetrical
capacity in both upstream and downstream. We define
the item to be traded in the auction as the most granular
frame unit FU in an XGS-PON. This means for an
upstream line rate of 9.953,28-Gbps each FU is one
block (16 bytes) according to the XGS-PON standard.
Thus, each XGS-PON frame (125 microseconds) can at
most contain 9720 allocation structures. i.e. 9,720 FUs
per frame.
The seller V NOiSand the buyer V NOB
j’s true value
is given by vS
i,F U and vB
j,F U respectively. The natural
objective of the auction is to buy the items from the
sellers with the lowest vS
i,F U and sell it to the buyers
with the highest vB
j,F U . This way the revenue generated
by the frame will be maximized i.e. the FUs will be taken
from the VNOs with the lowest probability of utilizing
them and will be allocated to the VNOs with the highest.
The auction starts by all the VNOs announcing their per
unit valuation for the frame along with the quantity of
the excess or demanded FUs. The number of excess FUs
of each seller VNO is shown by qS
iand the number of
demanded FUs of the buyer VNOs is shown by qB
After receiving the information from the VNOs, first
the InP checks the validity of the auction i.e. if there is
at least one seller and one buyer eligible for an efficient
trade. If the auction is valid, the InP will remove the
buyers with a value lower than the lowest value of the
sellers and likewise for the sellers with a valuation higher
than the highest of the buyers’ values. This is due to
the fact that the main objective of the auction is to sell
the items to the VNOs which value it the most and
selling the item to a buyer that values it less than all
of the sellers will not be acceptable. The same applies
for removing the sellers since it won’t be acceptable to
sell an item to a buyer who values it less than the seller.
We assume the auctioneer (InP) faithfully executes its
duty and all participants unconditionally trust him. We
also consider that all the items to be sold (FUs) are
identical and indistinguishable, and no buyer prefers one
sellers item over that of another. The InP sets a fixed
BaseP r ice for each FU and admits the FUs from the
seller with vS
i,F U lower than the BaseP rice up to the
value QB
j=1 qB
jsince the InP does not want to
admit more FUs than the total demand of the buyers. The
Process is shown in Algorithm 1. The InP acquires the
items from the sellers through holding a simple uniform
price reverse auction, i.e. the InP starts admitting the
items from the sellers with the lowest vS
i,F U and moves
on to the next until it has enough items to satisfy all
the buyers while updating the qS
i,sold for each seller.
After calculating the qInP and B aseP rice the InP holds
Algorithm 1: FU Admission From the Sellers
1while vS
i,F U < BaseP rice and qI nP < QB
2qInP + = PM
i=1 qS
a VCG auction based on these values. We implement
the VCG auction in two phases: the first phase is the
winner determination and the second phase the price
A. Phase One: Winner Determination
In this Phase the InP simply sells the minimum of
qInP and qB ,j FUs to the buyer with the highest value
until qInP = 0 or no buyer requires more FU. The
Process is shown in Algorithm 2. In Algorithm 2 λj
Algorithm 2: Winner Determination
1while qInP ! = 0 and V NOB
2if vB
j,F U > B aseP rice then
jwins min(qInP , qB
j)of FUs
j,won+ = min(qInP , qB
5qInP = min(qI nP , qB
InP =vB
j,F U ×min(qInP , qB
j=0 λj
is the value brought to the auctioneer from the winner
buyer V NOB
jand ΛN
InP is the total value brought to the
Auctioneer from all the winner buyers.
InP =vB
j,F U ×min(qInP , qB
InP =
InP (2)
B. Phase Two: Winners Pay Determination (VCG)
In VCG the payment of each buyer equals the amount
of harm they cause to other buyers by winning the items.
The difference of the total generated value with the
winner VNO present in the auction and not present in the
auction is the amount that the VNO pays. The Pseudo
code is shown in Algorithm 3. In Algorithm 3 ΦB
jis the
difference between the total value brought to the market
if V NOB
jwas absent and λj
InP minus the value brought
by V NOB
j= ΛNj
InP λj
InP )(3)
After phase one and two, the winner buyers will be
Algorithm 3: VCG Winners’ Pay Determination
1while V NOB
jIs A Winner do
jpays max (ΦB
j, B aseP rice )
assigned the amount of FUs that they have won and
will pay the price ΦB
jor the BaseP r ice to the auc-
tioneer whichever is the highest. The last step of the
auction is the uniform payment to all the seller with the
BaseP r ice.
The Utility of the Buyer represents the difference
between its true valuation for all the acquired items and
the total price paid to the auctioneer:
UB= (vB
j,F U ×qB
The Utility of the seller is the difference between the
total amount paid by the auctioneer in return for the
sold items and its true valuation:
i,sold ×(BaseP r ice vS
i,F U )(5)
The Utility of the InP is the budget surplus which is the
difference between the amount paid by the buyers and
the amount to be paid to the sellers:
UInP =
BaseP r ice ×qS
i,sold (6)
(a) Buyers’ Utility (b) Sellers’ Utility (c) Inp’s Utility
Fig. 2: Market simulation results in various (Supply-Demand) differences. a) Buyers’ Utility, b) Sellers’ Utility, c)
Inp’s Utility.
In this section, we report the results of the market
simulation proving the achievement of the two important
properties of auction namely individual rationality and
weak budget balance.
We simulate a market with ten VNOs each with an
equal share of the upstream bandwidth i.e. 9,953.28-
Mbps1-Gbps. This translates to 972 blocks (16 bytes)
or FUs per frame (125 µs) per VNO. Each VNO is
allowed to ask for up to twice its share of the frame
i.e. in the range [0 1944]. VNOs will generate an
optimistic bandwidth map considering being allocated all
their demanded extra FUs in case that they are buyers,
and selling all their excess FUs in case that they are
sellers. Then, after holding the auction if the outcome is
different the merging engine (operated by the InP) will
adjust the issued bandwidth map according to the items
that each VNO successfully has acquired or sold. This
is important in order to eliminate the need for an extra
round of communication between the InP and the VNOs
to announce the outcome of the auction.
The valuation for each FU reflects the probability of
the VNO to have a demand from its users for a FU.
This valuation is done by the VNO and it is either
based on the methods specified in the standard i.e. traffic
monitoring and the queue occupancy reports from the
ONUs or new methods such as traffic forecasting [15].
The valuation (demand probability) is between [0,1] for
both the seller and buyer VNOs.
A VNO is a seller if it has excess frame share and has
a valuation less than BaseP rice for each frame unit,
i.e. sellers’ valuation is in the range [0, BaseP r ice).
A VNO is an eligible buyer if its valuation is higher
than BaseP r ice, i.e. in the range (BaseP rice, 1]. Note
that we interpret the high valuation for an item as a
sign that the buyer with the highest valuation is the least
likely of the buyers to waste the frame unit. We set the
BaseP r ice to 0.5 in our simulation meaning that if a
VNO has a strictly smaller probability of utilizing the
frame share it will be willing to trade it for a higher
value. The choice of 0.5 as the BaseP rice is to achieve
a trade-off between the number of eligible sellers and
buyers. A higher BaseP rice will lead to a smaller
population of eligible buyers and a smaller BaseP rice
will limit the potential population of the sellers. We
Fig. 3: Market simulation results in total.
simulated the market for a duration of 10 minutes, which
involves auctioning 4,800,000 consecutive frames, each
of 125 µs duration. The simulation results are illustrated
in Fig. 2 and have been separated into three sub-figures
based on the buyers’, sellers’ and InP’s utility. In all sub-
figures the X-axis denotes the difference between the
supply and demand (Supply Demand). Each point
in the X axis is a bin that contains the result of 100
auctions with a (Supply Demand) difference falling
within the range of x±when = 4.
To analyze the results statistically we used box plots
for our numerical results. In a box plot the quartiles
of the data are shown graphically. The red line in the
middle of the box is the median, and the lower and
higher horizontal edges of the box denote the first and
third quartile of the data, respectively. The dashed lines
are the whiskers and the red pluses are the outliers.
Fig. 2 shows the box plots of the per FU utility of the
buyers and the sellers. As expected in Fig. 2.a we see that
significantly lower supply than demand leads to very low
utility for buyers because the bid values are becoming
very close to each other. So, the winner will pay a price
which is very close to its valuation. The utility increases
as the supply increases. While more supply than demand
leads to higher average utility, it also means that all the
bidders will pay the reserved price. This leads to a wide
range of positive utilities for buyers; some will have high
utility and others will have very low. This effect is visible
on the right end of Fig. 2.a. In Fig. 2.b we see a similar
effect on the left side where demand is much higher than
supply. In Fig. 2.c we present the utility of InP which is
non-zero only when the demand is greater than supply.
For InP Supply Demand 0means that the buyers
pay the reserved price which is equal to the procurement
price from sellers. Thus, there is no revenue in these
scenarios. On the other hand, InP utility is positive when
the demand is higher than supply.
Finally, Fig. 3 represents the outcome of the auction
throughout the whole duration of the simulation. Fig. 3
suggests that the system satisfies both individual ratio-
nality and weak budget balance, since the sellers’, buy-
ers’, and InP’s utility is never negative. Therefore, none
of the VNOs nor the InP will ever regret participating
in the auction.
In this paper, we proposed an auction model to address
the Inter-operator dynamic capacity sharing in virtual-
ized multi-tenant PONs. The proposed auction mecha-
nism is incentive compatible and satisfies all the desired
economic properties. The auction was designed in a way
to minimize any communication delay that could affect
the network scheduling operations, which typically occur
every 125 µs. For this purpose, VNOs send all the
required information for conducting the auction at once,
along with the bandwidth map. The market simulation
results confirm our claim that our model is individual
rational and weak budget balanced. These properties
provide incentives for the seller VNOs to participate in
the auction by selling their excess capacity and for the
buyers to acquire it while implementing their desired
bandwidth allocation mechanism. The InP also benefits
from the auction in two ways: first by receiving revenue
for facilitating the auction; second, by operating the
network resources more efficiently, which can lead to the
possibility to accommodate a higher number of VNOs
in the same PON. A natural progression of this work
is to analyze the effects of adjusting the BaseP rice
according to the Supply Demand on the utilities of
the sellers, buyers, and the InP.
This publication has emanated from research con-
ducted with the financial support of Science Founda-
tion Ireland (SFI) and is co-funded under the Euro-
pean Regional Development Fund under Grants Number
14/IA/2527 (OSHARE) and 13/RC/2077 (CONNECT).
[1] Broadband Forum SD-315, “Stage 1 analysis of
fixed access network sharing,” Tech. Rep., June
2015. [Online]. Available: http://www.broadband-
[2] J. van de Belt, H. ahmadi, and L. Doyle, “Defining
and surveying wireless link and network virtual-
ization,” IEEE Communications Surveys Tutorials,
vol. PP, no. 99, pp. 1–1, 2017.
[3] M. Ruffini, “Multidimensional convergence in fu-
ture 5g networks,Journal of Lightwave Technol-
ogy, vol. 35, no. 3, pp. 535–549, Feb 2017.
[4] T. Orphanoudakis, E. Kosmatos, J. Angelopoulos,
and A. Stavdas, “Exploiting pons for mobile back-
haul,” IEEE Communications Magazine, vol. 51,
no. 2, pp. S27–S34, February 2013.
[5] I. Macaluso, B. Cornaglia, and M. Ruffini, “An-
tenna, spectrum and capacity trade-off for cloud-
ran massive distributed mimo over next generation
pons,” in 2017 Optical Fiber Communications Con-
ference and Exhibition (OFC), March 2017, pp. 1–
[6] T. Tashiro, S. Kuwano, J. Terada, T. Kawamura,
N. Tanaka, S. Shigematsu, and N. Yoshimoto, “A
novel dba scheme for tdm-pon based mobile fron-
thaul,” in OFC 2014, March 2014, pp. 1–3.
[7] A. Elrasad, N. Afraz, and M. Ruffini, “Virtual
dynamic bandwidth allocation enabling true pon
multi-tenancy,” in 2017 Optical Fiber Communi-
cations Conference and Exhibition (OFC), March
2017, pp. 1–3.
[8] A. Banerjee, G. Kramer, and B. Mukherjee, “Fair
sharing using dual service-level agreements to
achieve open access in a passive optical network,”
IEEE Journal on Selected Areas in Communica-
tions, vol. 24, no. 8, pp. 32–44, Aug 2006.
[9] W. Vickrey, “Counterspeculation, auctions, and
competitive sealed tenders,Journal of Finance,
vol. 16, no. 1, pp. 8–37, 1961.
[10] L. M. Ausubel and P. Milgrom, “The lovely but
lonely vickrey auction,” in Combinatorial Auctions,
chapter 1. MIT Press, 2006.
[11] H. Ahmadi, I. Macaluso, I. Gomez, L. DaSilva,
and L. Doyle, “Virtualization of spatial streams for
enhanced spectrum sharing,” in 2016 IEEE Global
Communications Conference (GLOBECOM), Dec
2016, pp. 1–6.
[12] Q. Wu, M. Zhou, Q. Zhu, and Y. Xia, “Vcg auction-
based dynamic pricing for multigranularity service
composition,” IEEE Transactions on Automation
Science and Engineering, vol. PP, no. 99, pp. 1–
10, 2017.
[13] I. Koutsopoulos and G. Iosifidis, “Auction mecha-
nisms for network resource allocation,” in 8th Inter-
national Symposium on Modeling and Optimization
in Mobile, Ad Hoc, and Wireless Networks, May
2010, pp. 554–563.
[14] ITU-T, “G.9807.1 : 10-gigabit-capable symmetric
passive optical network (xgs-pon),” International
Telecommunication Union, Geneva, Recommenda-
tion 9807.1.
[15] P. Sarigiannidis, D. Pliatsios, T. Zygiridis, and
N. Kantartzis, “Dama: A data mining forecasting
dba scheme for xg-pons,” in 2016 5th International
Conference on Modern Circuits and Systems Tech-
nologies (MOCAST), May 2016, pp. 1–4.
... Thus, we put our best effort to improve the non-asymptotic efficiency. The preliminary results of this research was presented in OFC 2018 [24] and PIMRC 2017 [25] Conferences where we employed a VickreyClarkeGroves auction (VCG) in the buyers side and uniform price procurement on the sellers side. ...
Full-text available
In this paper we address the sharing incentive issue in multi-tenant passive optical networks (PONs). We propose an economic-robust and efficient sharing platform for new emerging multi-tenant PON networks. This platform is capable of accommodating a diverse range of service providers and enhancing the network utilization. We propose a sharing platform that provides sharing incentives for the incumbent network operators through monetization of inter-operator network sharing. Meanwhile, the platform allows the incumbent operators to operate a virtual instance of the bandwidth scheduling algorithm which enables them to meet their quality of service and latency requirements. Therefore, the proposed sharing platform grants a high degree of control to the operators co-operating the same network while, thanks to the higher resource efficiency, reduces the initial investment. We first model the multi-tenant PON as a market and define the roles of the virtual network operators (VNOs) and the infrastructure provider (InP) along with their utility functions. We propose a double auction mechanism to facilitate the trading of excess resources. The proposed double auction satisfies the crucial economic properties of a market while it achieves more efficient resource allocation among the market players. We have theoretically proven the economic robustness of the mechanism including incentive compatibility, individual rationality and weak budget balance. Through extensive market simulations, we confirmed that the proposed mechanism achieves superior allocative efficiency compared to a reference baseline mechanism.
Full-text available
Virtualization is a topic of great interest in the area of mobile and wireless communication systems. However the term virtualization is used in an inexact manner which makes it difficult to compare and contrast work that has been carried out to date. The purpose of this paper is twofold. In the first place, the paper develops a formal theory for defining virtualization. In the second instance, this theory is used as a way of surveying a body of work in the field of wireless link virtualization, a subspace of wireless network virtualization. The formal theory provides a means for distinguishing work that should be classed as resource allocation as distinct from virtualization. It also facilitates a further classification of the representation level at which the virtualization occurs, which makes comparison of work more meaningful. The paper provides a comprehensive survey and highlights gaps in the research that make for fruitful future work.
Conference Paper
Full-text available
We propose a virtual-DBA architecture enabling true PON multi-tenancy, giving Virtual Network Operators full control over capacity assignment algorithms. We achieve virtualization enabling efficient capacity sharing without increasing scheduling delay compared to traditional (non-virtualized) PONs.
Full-text available
Future 5G services are characterised by unprecedented need for high rate, ubiquitous availability, ultra-low latency and high reliability. The fragmented network view that is widespread in current networks will not stand the challenge posed by next generations of users. A new vision is required, and this paper provides an insight on how network convergence and application-centric approaches will play a leading role towards enabling the 5G vision. The paper, after expressing the view on the need for an end-to-end approach to network design, brings the reader into a journey on the expected 5G network requirements and outlines some of the work currently carried out by main standardisation bodies. It then proposes the use of the concept of network convergence for providing the overall architectural framework to bring together all the different technologies within a unifying and coherent network ecosystem. The novel interpretation of multi-dimensional convergence we introduce leads us to the exploration of aspects of node consolidation and converged network architectures, delving into details of optical-wireless integration and future convergence of optical data centre and access-metro networks. We then discuss how ownership models enabling network sharing will be instrumental in realising the 5G vision. The paper concludes with final remarks on the role SDN will play in 5G and on the need for new business models that reflect the application-centric view of the network. Finally, we provide some insight on growing research areas in 5G networking.
Full-text available
The growing popularity of mobile data services necessitates a rapid rise in network capacity not only on the air interface to the end user, but also in the backhaul network. The latter is quite important for the mobile operator business model, affecting capital investment, operational expenses, service deployment, and customer experience. Fiber infrastructure is inevitably the only long-term solution, and the deployment of passive optical networks presents an opportunity for a cost-effective, scalable, and future-proof solution. In this article we investigate the use of PONs for mobile backhaul and propose a resource allocation framework building on the efficiency of PONs to share resources, dynamically allocate bandwidth in real time, and enhance efficiency by improved statistical multiplexing. The main objective of this work is to exploit existing standardized technologies, and provide design and deployment guidelines regarding PON MAC operation, enabling a gradual and future-safe infrastructure upgrade of mobile backhaul systems.
When a single service on its own cannot fulfill a sophisticated application, a composition of services is required. Existing methods mostly use a fixed-price scheme for service pricing and determine service allocation for composition based on a first-price auction. However, in a dynamic service market, it is difficult for service providers to determine a fixed price that is profitable while attractive to customers. Meanwhile, this mechanism cannot ensure that the providers who require the least cost to provide services would win the auction, because the pricing strategy of service providers is unpredictable. To address such issues, in this paper, we propose Vickrey-Clarke-Groves auction-based dynamic pricing for a generalized service composition. We consider fine-grained services as candidates for composition as well as coarse-grained ones. In our approach, service providers bid for services of different granularities in the composite service and based on received bids, a user decides a composition that minimizes the social cost while meeting quality constraints. Experimental results at last verify the feasibility and effectiveness of the proposed approach.
Conference Paper
In this work we propose a virtualized network architecture for an infrastructure provider that shares the physical resources of a Massive MIMO cell among several virtual network operators (VNOs) using spatial multiplexing. In this architecture the infrastructure provider allocates spatial streams to the VNOs, which enables each VNO to select its own scheduling policy and user priority to differentiate its service from the other VNOs. To assign the spatial streams to the VNOs that value them the most, we propose an auction-based spatial stream allocation approach. We show that the proposed auction-based approach performs very close to the optimal (fixed) approach in the case of homogeneous static VNOs demand. In case of heterogeneous demands, the auction mechanism is able to dynamically allocate the resources according to the needs of different VNOs.
Conference Paper
Passive Optical Networks (PONs) provide a cost effective, efficient, and flexible solution for addressing the bottleneck in last mile in modern access networks. End users are interconnected with the Central Office (CO) by means of optical fibers, where pure optical paths are created allowing upstream and downstream communication directions. The latest new generation PON (NG-PON) standard, known as 10-gigabit-capable passive optical network (XG-PON), enables a very promising architecture that offers 10 Gbps nominal data delivery ratio in the downstream direction. In the upstream direction, a channel supporting 2.5 Gbps rate is shared among optical network units (ONUs) that constitute the connecting interface between the optical path and the final users. The optical line terminal (OLT) is located within the CO and constitutes the main decision-making tank of the PON. OLT applies a dynamic bandwidth allocation (DBA) scheme for coordinating the transmission opportunities, especially in the upstream direction. According to the standard, a differential fibre distance of 40 km, between ONUs and OLT, is allowed. This outspread deployment implies high propagation delays which should be taken into account of designing the bandwidth allocation. This work is focused on proposing a cognitive DBA scheme which is capable of forecasting the additional bandwidth, which arrives in ONUs, during the transmission coordination between OLT and ONUs. The k-nearest neighbors algorithm is applied for forecasting the additional bandwidth requests of each ONU. In addition, the adopted algorithm is enhanced with an adaptive reward/penalty learning method which efficiently selects the most appropriate k value based on the traffic dynamics. Simulation results indicate notable improvements when the proposed scheme is applied in terms of jitter and latency.
Conference Paper
We propose a mobile-DBA with low-latency for a TDM-PON based mobile fronthaul. It utilizes mobile-scheduling information and reduces the latency to about 1/20 of conventional one. Measured latencies (< 50 μs) are enough for LTE.