Breaking the Economic Barrier of Caching in Cellular
Networks: Incentives and Contracts
Kenza Hamidouche, Walid Saad, and Mérouane Debbah
Large Networks and Systems Group (LANEAS), Centrale-Supélec, Gif-sur-Yvette, France
[email protected]
Tyrrhenian International Workshop on Digital Communications
September 12, 2016
LANEAS
Large Networks and Systems Group
Kenza Hamidouche (CentraleSupélec)
Breaking the Economic Barrier of Caching in SCNs
12/09/2016
Introduction
Outline
1
Introduction
2
System Model
3
Proposed Incentive Mechanism for Caching
4
Simulation Results
5
Conclusions and Perspectives
2/27
Introduction
Outline
1
Introduction
2
System Model
3
Proposed Incentive Mechanism for Caching
4
Simulation Results
5
Conclusions and Perspectives
3/27
Introduction
Motivation
Exponential growth of mobile video traffic [1].
Proliferation of social networks.
Increasing number of mobile devices.
Limited capacity of the deployed wireless networks [2]
Dense deployment of small base stations (SBSs).
Capacity-limited heterogeneous backhaul links in small cell networks
(SCNs).
Distributed caching
[1] Cisco visual networking index: Global mobile data traffic forecast update, 2014-2019. White Paper, [Online] http://goo.gl/uQ0DJQ, 2014.
[2] Jakob Hoydis, Mari Kobayashi, and Mérouane Debbah, ”Green small-cell networks,” IEEE Vehicular Technology Magazine, 6(1):37-43, 2011.
4/27
Introduction
Motivation
Exponential growth of mobile video traffic [1].
Proliferation of social networks.
Increasing number of mobile devices.
Limited capacity of the deployed wireless networks [2]
Dense deployment of small base stations (SBSs).
Capacity-limited heterogeneous backhaul links in small cell networks
(SCNs).
Distributed caching
[1] Cisco visual networking index: Global mobile data traffic forecast update, 2014-2019. White Paper, [Online] http://goo.gl/uQ0DJQ, 2014.
[2] Jakob Hoydis, Mari Kobayashi, and Mérouane Debbah, ”Green small-cell networks,” IEEE Vehicular Technology Magazine, 6(1):37-43, 2011.
4/27
Introduction
Distributed Caching
Idea: Duplicate data and store it in the small cells [3].
Caching involves, users, mobile network operatos (MNOs), and content
providers (CPs).
How can MNOs incite CPs to cache their content at the SBSs?
Radio links
The core network
Small base station
Backhaul links
[3] N. Golrezaei, A. G. Dimakis, and A. F. Molisch, “Wireless device-to-device communications with distributed caching,” in Proc. of IEEE
International Symposium on Information Theory, Cambridge, MA, USA, Jul. 2012, pp. 2781–2685.
5/27
Introduction
Literature Overview
Distributed caching:
Minimize the expected user’s delay [4].
Consider user’s mobility [5].
Consider the geographical position of SBSs [6].
Create MIMO cooperation opportunities between SBSs [7].
Business of distributed caching:
Stackelberg game to maximize the revenue of the MNO and users’ QoS [8].
Stackelberg game to jointly maximize the profit of MNOs and CPs [9].
Minimize the SBSs’ cost by incitivizing users to share their cached content
[10].
Presence of asymmetric information
[4] N. Golrezaei, A. M. Molisch, A. G. Dimakis, and G. Caire, “Femto-caching and device-to-device collaboration: A new architecture for wireless
video distribution,” IEEE Communication Magazine, vol. 51, no. 4, pp. 142-149, 2013.
[5] K. Poularakis and L. Tassiulas, “Exploring user mobility for wireless content delivery,” IEEE International Symposium on Information Theory
Proceedings, 2013, pp. 1017-1021.
[6] B. Blaszczyszyn and A. Giovanidis, “Optimal geographic caching in cellular networks,” ICC, 2015.
[7] A. Liu and V. Lau, “Exploiting base station caching in MIMO cellular networks: Opportunistic cooperation for video streaming,” IEEE
Transactions on signal processing, vol. 63, no. 1, pp. 57-69, Jan. 2015.
[8] J. Li, H. Chen, Y. Chen, Z. Lin, B. Vucetic, and L. Hanzo, ”Pricing and resource allocation via game theory for a small-cell video caching system,”
IEEE Jour. on Sel. Areas in Comm., 2016.
[9] J. Li, H. Chen, Y. Chen, Z. Lin, B. Vucetic, and L. Hanzo, ”Pricing and resource allocation via game theory for a small-cell video caching system,”
IEEE Jour. on Sel. Areas in Comm., 2016.
[10] Z. Chen, Y. Liu, B. Zhou, and M. Tao, ”Caching incentive design in wireless D2D networks: A stackelberg game approach,” IEEE Int. Conf. on
6/27
Comm., 2016
System Model
Outline
1
Introduction
2
System Model
3
Proposed Incentive Mechanism for Caching
4
Simulation Results
5
Conclusions and Perspectives
7/27
System Model
System Model
Content Providers
Macro base station
Small base station
Backhaul links
A set M of M macro base stations (MBSs) deployed by an MNO.
A set S of S small base stations (SBSs).
A set U of U user equipments (UEs).
A set C of C content providers (CPs) with different traffic loads and content
popularity.
8/27
System Model
System Model
Content Providers
Macro base station
Small base station
Backhaul links
The CPs’ types are sorted in an ascending order and classified into K types θ1 , ..., θK
with K ≤ C :
θ1 < ... < θk < ... < θK , k ∈ {1, ..., K }.
(1)
Specify a suitable performance-reward bundle contract
(π, ρ) = (price, storage allocation).
9/27
System Model
System Model
Content Providers
Macro base station
Small base station
Backhaul links
The data rate from SBS i to user j is:
ï
αij (ρ(θ), θ) = Et wij log 1 +
ò
pij |hij |2 .
σ 2 + I (ρ, θ)
(2)
10/27
System Model
System Model
Content Providers
Macro base station
Small base station
Backhaul links
The achievable rate by an SBS i from MBS m:
ï
ò
pmi |hmi |2 0
αmi
= Et wmi log 1 + 2
,
σ + I0
P
0
2
(3)
where I = l∈M\m pli |hli | is the interference experienced by SBS i from all the other
transmitting MBSs.
11/27
System Model
System Model
Content Providers
Macro base station
Small base station
Backhaul links
The achievable rate of user i from SBS j in a cache-enabled network:
0
rij (θ) = (1 − βif (ρ(θ), θ)) min {αij (ρ(θ), θ), αmi
} + βif (ρ(θ), θ)αij (ρ(θ), θ),
(4)
where β ∈ {0, 1}S×Fk is the outcome of the MNO’s storage allocation ρ.
12/27
System Model
System Model
Content Providers
Macro base station
Small base station
Backhaul links
The total rate of the users of CP k can be given by:
XX
rk (ρ(θ), θk ) =
rij (θ),
(5)
i∈S j∈Uki
where Uki ⊆ Uk is the set of users that request at least one file from CP k by using SBS
i, and Uk is the set of users requesting files of CP k.
13/27
Proposed Incentive Mechanism for Caching
Outline
1
Introduction
2
System Model
3
Proposed Incentive Mechanism for Caching
4
Simulation Results
5
Conclusions and Perspectives
14/27
Proposed Incentive Mechanism for Caching
Utilities of the MNO and the CPs
The utility function of a CP k of type θk :
uk (θ) = rk (ρk (θ), θk ) − πk (θ).
(6)
A proper utility function for the MNO is given by:
vk (θ) = πk (θ) − ck (θ, θk ),
(7)
where πk is the price that the operator charges CPs of type k.
15/27
Proposed Incentive Mechanism for Caching
Utilities of the MNO and the CPs
The total utility of the operator can be given by:
v=
X
vk (θ).
(8)
k∈C
Maximizes the global benefit of the CPs:
max
(πk ,ρk )
subject to
X
uk (ρk (θ), θk )
k∈C
(9)
v ≥ 0.
16/27
Proposed Incentive Mechanism for Caching
Feasibility of a Contract
Definition
Individual Rationality (IR): The contract that a CP selects should guarantee that the
utility of the CP is nonnegative for any θ−k declared by the other CPs,
rk (ρk (θk , θ−k ), θk ) − πk ≥ 0, ∀k ∈ {1, ..., K }.
(10)
Definition
Incentive Compatibility (IC): A contract satisfies incentive compatibility constraint if each
CP of type θk prefers to reveal its real type θk rather than another type θ̂k , i.e.,
rk (ρk (θk , θ−k ), θk ) − πk ≥ rk (ρk (θ̂k , θ−k ), θk ) − πk .
(11)
17/27
Proposed Incentive Mechanism for Caching
Incentive Mechanism Analysis
Optimization problem
max
(πk ,ρk )
subject to
X
uk (ρk (θ), θk )
k∈C
rk (ρk (θk , θ−k ), θk ) − πk ≥ 0, ∀k ∈ {1, ..., K },
(12)
rk (ρk (θk , θ−k ), θk ) − πk ≥ rk (ρk (θ̂k , θ−k ), θk ) − πk ,
v ≥ 0.
18/27
Proposed Incentive Mechanism for Caching
Solution of the incentive optimization problem
Theorem
The unique efficient solution of the optimization problem (12) can be given by:
X
∗
ρk ∈ arg max
ρk
πk (θ̂) =
î
max
ρi
|
X
ri (ρi (θ̂), θ̂i ) − ci (θ̂) , ∀k,
i
ó îX
ri (ρi (θ̂−k ), θ̂i ) − ci (θ̂−k ) −
(13)
ó
ri (ρ∗i (θ̂), θ̂i ) − ci (θ̂) ,
(14)
i6=k
i6=k
{z
(a)
}
|
{z
(b)
}
where (a) represents the maximized social welfare when CP k is not considered while in (b), CP
k is considered. Moreover, θ̂ represents the revealed type by the CPs while θ is the real type of
the CPs.
19/27
Proposed Incentive Mechanism for Caching
Proof of the Theorem
Lemma
The proposed mechanism is incentive compatible.
Lemma
Truth telling is a dominant strategy under (14).
Lemma
The proposed mechanism (14) is individually rational.
Lemma
The proposed mechanism (14) is weakly budget-balanced.
20/27
Simulation Results
Outline
1
Introduction
2
System Model
3
Proposed Incentive Mechanism for Caching
4
Simulation Results
5
Conclusions and Perspectives
21/27
Simulation Results
Simulation Parameters
We consider 5 CPs with 5 different traffic loads from 1 to 5.
For each CP, a set of 100 files with a popularity that follows a Zipf
distribution.
The MNO deploys one MBS and 10 SBSs.
The storage capacity of each SBS is 1Gbits.
22/27
Simulation Results
Amount of content served from the backhaul (Mbits)
Served content via backhaul with respect to CPs’ type
700
Type-5 CP
Type-4 CP
Type-3 CP
Type-2 CP
Type-1 CP
600
500
400
300
200
100
0
1
1.5
2
2.5
3
3.5
4
4.5
5
Types of the CPs (from 1 to 5)
Figure: The amount of served content via backhaul with respect to CPs types.
23/27
Simulation Results
CPs utility vs CPs type
250
Proposed mechanism
Equal allocation of the storage space
Utility of the CPs
200
150
100
50
0
1
1.5
2
2.5
3
3.5
4
4.5
5
Types of the CPs (from 1 to 5)
Figure: Utility of CPs as a function of the CPs’ type and storage allocation policy.
24/27
Conclusions and Perspectives
Outline
1
Introduction
2
System Model
3
Proposed Incentive Mechanism for Caching
4
Simulation Results
5
Conclusions and Perspectives
25/27
Conclusions and Perspectives
Conclusions and Perspectives
Proposed a novel cache incentive framework.
Used contract theory to account for the presence of asymmetric information.
Proposed a model that accounts for the interdependence between contracts.
Showed via simulations the effectiveness of the proposed framework.
Perspectives
Extend the model by explicitly defining the caching policy.
Extend the framework for multiple competing MNOs.
26/27
Thanks
K. Hamidouche, W. Saad, M. Debbah, ”Breaking the Economic Barrier of Caching in Cellular Networks: Incentives and
Contracts”, IEEE Global Communications Conference, Washington, DC, USA, 2016.
Thank you for your attention.
Questions?
http://www.laneas.com/kenza-hamidouche
[email protected]
27/27