Design of Clutches and Brakes in Design of Machine Elements.pptx
Traffic models and estimation
1. “Traffic Models and Estimation
For 5G networks”
Mina Youssre Yonan
Eng.minayoussre@gmail.com
2. Traffic Models
• Stochastic geometry model [1],[2]
• Temporal-spatial traffic Model [3],[4],[5]
• Traffic-aware Model [6]
ELSAWY, Hesham, et al. Modeling and analysis of cellular networks using
stochastic geometry: A tutorial. IEEE Communications Surveys & Tutorials, 2017.
[1]
[2]
Feng, M., Mao, S. and Jiang, T., Base Station ON-OFF Switching in 5G Wireless
Networks: Approaches and Challenges. IEEE Wireless Communications ,2017.
LEE, Dongheon, et al. Spatial modeling of the traffic density in cellular
networks. IEEE Wireless Communications, 2014, 21.1: 80-88.
[3]
[4]
[5]
ZHAO, Zhifeng, et al. Temporal-spatial distribution nature of traffic and base
stations in cellular networks. IET Communications, 2017
WANG, Shuo, et al. An approach for spatial-temporal traffic modeling in mobile
cellular networks. In: Teletraffic Congress (ITC 27), 2015 27th International. IEEE,
2015.
SAXENA, Navrati; ROY, Abhishek; KIM, HanSeok. Traffic-aware cloud RAN: a key
for green 5G networks. IEEE Journal on Selected Areas in Communications, 2016.
[6]
3. Stochastic geometry Model
The stochastic geometry framework was widely used to
analyze the theoretical system performance the key factors
include :
• User/Traffic arrival rate
• user distribution over space
• file/task size
• service rate
4. • Possion Point Process
- The most widely used model for spatial locations of node
- No dependence between
node locations.
- Random number of nodes
Stochastic geometry Model
5. • Possion Point Process
A stationary Poisson point process 𝛷 of density 𝜆 is characterized by :
1- The number of points in a bounded set 𝐴 ⊂ R2 has a Poisson
distribution with mean 𝜆 𝐴
P( 𝛷 𝐴 = 𝑛 ) =
𝜆 𝐴 𝑛
𝑛!
𝑒−𝜆 𝐴
2- The number of points in disjoint sets are
independent, i.e., For 𝐴 ⊂ R2
, 𝐵 ⊂ R2
and A∩B = ∅ .
𝚽 𝐀 totally independent of 𝚽 𝐁
Stochastic geometry Model
6. • Properties of PPP
1- The density of the PPP (as defiend in previous slide) is 𝜆.
Observe that the above expression does not depend on the set A.
Stochastic geometry Model
7. • Traffic Model Example In [7]
[7] E. Oh, K. Son, and B. Krishnamachari, “Dynamic Base Station Switching-On/Off Strategies for Green Cellular Networks,” IEEE Trans. Wireless
Commun., vol. 12, no. 5, pp. 2126–2136, May 2013.
1- assume the traffic arrival rate of UE located x at time t is modeled
as an independent Poisson distribution with mean arrival rate λ(x, t).
2- Its average requested file size is assumed to be an exponentially
distributed random variable with mean 1/μ(x, t).
3- The traffic load of UE is then defined as γ(x, t) = λ(x, t)/μ(x, t)
[inbps].
4 - the service rate of UE at location x from BS b at time t is calculated
as:
𝑠 𝑏 (𝑥, 𝑡) = 𝐵𝑊 ・ log2 (1 + 𝑆𝐼𝑁𝑅 𝑏(𝑥, 𝑡))
Stochastic geometry Model
8. Stochastic geometry Model
5- The system load of BS b at time t is defined as the fraction of
resource to serve the total traffic load in its coverage
where 𝐴 𝑏 represents BS b’s coverage. The system load denotes the
fraction of time required to serve the total traffic load in his coverage.
9. Stochastic geometry Model
• Problem Formulation
BS switching algorithm that minimizes the total energy expenditure in
cellular networks during T .
- 𝐸 𝐵𝑆 is the BS power consumption
- 𝑎 𝑏 𝑡 ∈ 0,1 is the activity indicator of BS b at time [0, T[
• Optimization Problem
energy saving problem considering the BS switching can be
11. Stochastic geometry Model
• A Notion of Network-impact
1- let examine the possibility whether a particular BS can be turned off
or not. We define the set of neighboring BSs of BS b by 𝑁𝑏, and further
denote by 𝑛 ∈ 𝑁𝑏 the neighboring BS providing the best signal strength
(except BS b) to the UE at the location 𝑥 ∈ 𝐴 𝑏 as follows:
𝑛 = iϵNb
arg max
g(i,x) ・ Pb , for 𝑥 ∈ 𝐴 𝑏.
2- Note that the BS n can be interpreted as the BS to which the traffic
loads will be transferred turning off BS b. The BS b will be able to switch
off only if all its neighboring BSs satisfy the following feasibility constraint
13. Temporal-spatial traffic Model
• PPP model can reflect the randomness of (users or base
stations) distribution but, it cannot describe the
convergence of user behavior and traffic [5].
15. Temporal-spatial traffic Model
• Spatial Traffic Model
In [3] , The target area
of the measurement is
an area of 160 × 180
km which includes all
types of areas (urban,
rural, and etc.). There
are about 5763 cell in
the target area. Cell
types include macro-
cells as well as small
cells, such as micro- and
pico-cells.
18. Temporal-spatial traffic Model
• the real traffic data from a mobile operator in one big city of China
is analyzed.
• The data contains information of 185 base stations from one base
station controller (BSC) during three weeks
• 6km * 2.5km area and the unit of traffic volume is Byte.
• The research area is located in dense urban area and contains
three typical regions: park, university campus and central business
district.
20. Spatial-temporal Traffic Modeling in Mobile Cellular Networks
• Spatial modeling of traffic
(Left) Spatial distribution of traffic in park region at spare time.
(Right) Spatial distribution of traffic in park region at busy time.
22. Spatial-temporal Traffic Modeling in Mobile Cellular Networks
• Spatial-temporal traffic modeling approach
- Step 1: Calculate the average traffic volume of all base
stations in the region using the sinusoid superposition model
- Step 2: According to the empirical value of 𝜎, compute the parameter
of lognormal distribution at each time with the following expression:
23. Spatial-temporal Traffic Modeling in Mobile Cellular Networks
Step 3: Generate the traffic value of every base station
in the region at each time using lognormal distribution with
parameters 𝜇 𝑡 and 𝜎 expressed as follows:
Where 𝑉𝑖(𝑡) is the traffic volume of the 𝑖 𝑡ℎ
base station at time t.
24. Traffic Learning
• Traffic Learning by using game with information-theoretic approach[2]
Optimizing the traffic
entropy
learn the
traffic patterns over
time
Efficient forecast of
the expected near-
future traffic
trigger smart
power saving
operations.
Saxena, Navrati, Abhishek Roy, and HanSeok Kim. "Traffic-aware cloud RAN: a key for green 5G networks." IEEE Journal on Selected Areas in
Communications 34.4 (2016)
25. Network Model
• Individual BBUs in C-RAN are often termed as Virtual Base Station
(VBS) and the entire pool is termed as Virtual Base Station Cluster
(VBSC).
• switches help in traffic distribution and load balancing among BBU-
RRH pairs
26. Information-theoretic approach
• The traffic estimation paradigm needs to consider the
joint traffic patterns across multiple VBSs in 5G C-RAN.
• At a specific regular time interval 𝑡sample, every VBS
samples and quantizes the traffic into discrete traffic
profiles of identical range of 100 kbps.
• Represent the quantized traffic by sequence of k symbols
𝜏1, 𝜏1, … , 𝜏 𝑘 where every symbol 𝜏 𝑘 𝜖 ℵ actually represents
a quantized traffic profile.
27. Information-theoretic approach
• Every individual VBS independently fails to consider the
correlation between the traffic patterns across all RRHs.
In fact, independent entropy optimization of each
individual VBS increases the overall joint uncertainty
(entropy)
𝐻 𝜏1, 𝜏1, … , 𝜏 𝑘 ≤
𝑖=1
𝑘
𝐻 𝜏i
28. Game theory (intro.)
• Stochastic games are widely used to model multiagent systems,
where individual agents pursue their own goals.
• An n-player stochastic game, G, is a tuple < 𝑆, 𝐴𝑖, 𝑟𝑖, 𝑝 >, ∀𝑖
∈ [1, … , 𝑛] where,
- S is state space
- 𝐴𝑖 is the action space
- 𝑟𝑖 Payoff Function
- 𝑝 Transition Probability
• If 𝜋𝑖 denotes the strategy of a player 𝑖, then for a given initial state
𝑠 = 𝑠0 and 𝛽𝑡 ∈ [0; 1], at any time t, the objective of any player 𝑖 is
to maximize the sum of expected rewards.
29. Game theory (intro.)
• If 𝜋𝑖 denotes the strategy of a player 𝑖
• for a given initial state 𝑠 = 𝑠0 and 𝛽𝑡 ∈ [0; 1],at any time t
• objective of any player 𝑖 is to maximize the sum of expected rewards
• A Nash equilibrium is a joint strategy, where each player is the
best response to the others. Thus, it is a tuple of n strategies
(𝜋1
∗
, 𝜋2
∗
, … , 𝜋 𝑛
∗ ) ∀𝑖 ∈ 1, … , 𝑛
31. Traffic-aware with game theory
• Nash Traffic Entropy Learning (NTEL)
• At each time t, every RRH observes the current state and makes
its traffic estimation (action).
• After that, it observes its own reward, actions taken by other RRH
and their rewards, and the new state.
• It then estimates it’s own Nash Equilibrium strategies
(estimations) at that stage and updates its own utility using the
iterative equation .
• The parameters 𝛼 𝑡 and 𝜓 are in the range [0,1]
34. Traffic-aware RRH Switch On/Off
• objective is to selectively switch off the relevant RRHs at low traffic
load. Our proposed NTEL-based proactive traffic-estimation assists
in designing an efficient strategy for RRH switch on/off.