Ultra Dense Network: Techno- Economic Views. By Mostafa Darabi 5G Forum, ITRC July 2017

Transcription

1

2 Ultra Dense Network: Techno- Economic Views By Mostafa Darabi 5G Forum, ITRC July 2017

3 Outline Introduction 5G requirements Techno-economic view What makes the indoor environment so very different? Beyond ultra-dense barrier Architectural considerations Managing hotspots in ultra dense network Future trends 3

4 Introduction Current design paradigm for mobile broadband systems is not a scalable and economically feasible way to solve the expected future capacity crunch. Advanced signal processing and antenna techniques and more millimeter wave spectrum, indeed provide more capacity, but are not the answer to the three to four orders of magnitude more capacity at today s cost. Solving the engineering problem of providing high data rates alone is not sufficient. Instead, we need to solve the techno-economic problem to find both business models and scalable technical solutions that provide extreme area capacity for a given cost and energy consumption. 4

5 Introduction Achieving very high capacities is feasible in indoor environments. However, to become economically viable, approaches with radically different fundamental cost factors compared to those used in today s cellular systems are needed. To reach very high capacity we must venture beyond the ultra-dense barrier, that is, networks where the number of access points in an area is larger than the active number of mobile terminals. In such networks area capacities of more than 1 Gb sm 2 are perfectly feasible. The problem set encountered in such UDN is very different from conventional cellular systems and their solution requires conceptually new tools. 5

6 Requirements for 5G 5G cellular networks will need to accommodate 1000x more data traffic and 50x more devices. The capacity can be increased by orders of magnitude via the following approaches: Increasing the bandwidth by using millimeter wave frequency Increasing the spectral efficiency with the help of Massive MIMO and Full-duplex communication. Using dynamic spectrum management techniques such as CRN, Unlicensed bandwidth Densifing the wireless networks, i.e. to increase the number of base stations per area unit. Densifing is not only an engineering problem we are dealing with a techno-economic problem. 6

7 Requirements for 5G The requirements for 5G boil down to two main areas: Containing the data tsunami, that is, more capacity and higher data rates to quench the thirst of more users for more data. Catering to efficient machine-type communication (MTC), systems that allow both billions on things to communicate as well as provide high-reliability, low-delay wireless communication in real-time control-loops. Here, we will mainly address the first challenge. 7

8 Techno-economic view In densification, the problem is to provide the high access data rate under given resources constraints. Some of the sub-problems involved are technical (e.g. available bandwidth, energy consumption, noise), but most of them are economic (e.g. the number and cost of base stations, wired infrastructure). Traditional cellular concept does not scale well. This worked well in the era of mobile telephony, as more capacity meant more paying customers. Now the same users expect much higher data rates without paying more. It is clear that devising a new network architecture is necessary to break this vicious cost circle. 8

9 Techno-economic view The exponentially growing traffic is not uniformly distributed. Most traffic is found in dense population centers and mainly indoors [1]. Reaching users inside the buildings by penetrating walls and windows with high power transmissions is neither very reliable nor very efficient. In order to increase the base station density further we need to deploy them indoors. Inside the walls, however, a radically different technoeconomic landscape opens up in which very different rules apply. The traditional operators are unable to use their usual toolbox for deployment, nor do they have the business models for indoor operation. [1] This case is described in the amazingly fast scenario of the 5GPPP METIS project, or in the 9 pervasive video scenario of the NGNM, or the gigabit in a second scenario by the ITU-R.

10 What makes the indoor environment so very different? In the indoor environment, high propagation losses created for the signals from access points (APs) to user terminals. From an interference perspective, the wall attenuation becomes a blessing, as very little interference escapes from a building, the same spectrum can be reused in adjacent buildings. Equipment cost for indoor operation is significantly lower, and the cost of maintenance is only a fraction of the cost of outdoor operation. Indoor access point hardware cost is likely to be in the 100 s of USD, rather than in the 1000 s of USD expected for outdoor deployment. In high-density plug-and-play deployments, the failure of a single access point may not even be noticed by the users, as it results only in a moderate loss in performance. The access point may be replaced at some convenient time. 10

11 What makes the indoor environment so very different? Outdoor system costs are dominated by towers, antennas, energy, backhaul connections, etc. What remains in indoor AP is the cost for the wired infrastructure, the backhaul for the access points. As the number of APs becomes very large, the backhauling cost becomes the dominant cost factor and few users will share it. This means that, despite the low cost of equipment, the cost per user still is significant in indoor systems. The usual competitive public operation business model used in mobile cellular collapses. The only economically sensible thing is that the facility/home owner deploys their own network, which then could be shared by the public operators to serve their indoor customers. 11

12 What makes the indoor environment so very different? Traditional mobile operators build their business on having exclusive access to nationwide blocks of spectrum. The competition for this spectrum will be fierce and large amounts of money are paid at spectrum auctions for cellular spectrum. The result of spectrum scarcity has been high equipment cost and energy hungry solutions. In indoor environment, the signals hardly leave or enter the building and the spectrum can be reused next door. There should be no competition for the spectrum as the facility owner is the only one that can make use of it. mm-wave band have several advantages 12

13 What makes the indoor environment so very different? A reasonable licensing regime would be to give the facility owner access to all spectrum, on the condition that public operators are given the opportunity to use the indoor network. Moving from the per-minute charge paradigm to a flat rate permonth charging model is radically changing the business of the operators. In the old paradigm off-loading a user to some other network meant losing call-minutes and revenue. In the flat-rate paradigm, keeping the customer s business while letting someone else provide the network access makes off-loading an interesting business proposition. The mobile operator s business is transitioning from solely providing infrastructure for communication to managing the connectivity of their customers. 13

14 What makes the indoor environment so very different? 14

15 Beyond ultra-dense barrier If we aim at serving a fixed user population with a large traffic demand and keep increasing the density of access points, at some point there will be more access points than terminals. We call this point the ultra-dense barrier, or more precisely, where the access point density exceeds the user density. As we push far beyond this point, the character of the system changes radically, and the behavior becomes similar to a distributed system of antennas. 15

16 Beyond ultra-dense barrier With received power inversely proportional to the distance to the power α, the SINR becomes proportional to ( D d) α The expected distance d and D are proportional to 1 λ AP and 1 λ U respectively. Using stochastic point processes [2] derive the area capacity ( b sm 2 ) as [2] J. Park, S.-L. Kim, and J. Zander, Asymptotic Behavior of Ultra-Dense Cellular Networks and Its Economic Impact, Proc. IEEE Global Commun. Conf. (GLOBECOM) 2014, Austin, 16 TX, USA, Dec

17 Beyond ultra-dense barrier The system energy consumption per user can be written as [3] P = c 1 λ AP α + C 2 λ 2 AP λ U The first term corresponds to the transmit power that approaches zero with increasing λ AP. The second term corresponds to the idle power consumption of the APs that are currently not used. [3] S. Tombaz, A. Vstberg, and J. Zander, Energy- and Cost-Efficient Ultra-High-Capacity Wireless 17 Access, (Invited Article), IEEE Wireless Commun., Oct

18 There is no (theoretical) limit to the area capacity as the number of users (λ U ) increases, as long as we keep operating beyond the ultra-dense barrier, that is, the ratio λ AP is kept large. λ U Beyond ultra-dense barrier A concern is that the rate only increases with the logarithm of the access point density and beamforming gain c. 18

19 Architectural considerations Area capacity is an increasing function of the number of access points. Thus, to achieve a certain capacity, a certain number of APs per area unit is needed. The cost to provide this capacity consists of the cost of the APs and the cost of the backhaul network. There exist two distinct groups of systems: Systems with more or less distributed control functions, e.g. WiFi Centralized solutions (also know as cloud RAN ):, e.g., Coordinated Multipoint, 19

20 Architectural considerations The WiFi-type systems have a very low fixed cost as they use existing backhaul and do not need centralized control. Systems with advanced coordinated interference exhibit a high fixed cost for the fronthaul/backhaul, but can achieve high capacities. The exact design principle will depend on the effective amount of available spectrum for our indoor system. 20

21 Architectural considerations If large amounts of spectrum are available, the required area spectrum efficiency will be low (region A ) and low complexity, WiFi-type systems will dominate the scene. If there is a shortage of spectrum, we are likely to operate in regions B or C, where centralized system designs with more advanced interference mitigation methods are needed. 21

22 Architectural considerations In which region would we be if today s off-the-shelf technology would be used? Below table shows the estimated amount of spectrum needed to achieve 1 Gb/sm2 in some simple scenarios. 22

23 Managing hotspot in ultra dense network Ultra-dense heterogeneous networks with non-uniform traffic hotspots are important scenarios for future wireless networks, which, however, impose challenges on efficient framework and algorithm design due to the heavy congestion incurred by massive hotspot users. QoE-aware optimization problem that enfolds connectivity and delay 23

24 Managing hotspot in ultra There are three types of data transmissions, differentiated by the number of involved relay Aps: Direct transmission One-relay transmission Two-relay transmission dense network QoE of hotspot users is defined as the combination of the connectivity and average delay. 24

25 QoE-Aware Optimization To simplify the substraction QoE expression, weights (credits) for users are introduced and QoE is transformed to a multiplication expression. To avoid the computationally complicated heuristic searching, the problem is solved by an iterative matching process. 25

26 QoE-Aware Optimization BS Allocation 1) Each user selects the BS that gives the maximum weighted utility. 2) If the number of users associated with a BS is smaller than its capacity, the allocation of this BS to these users is completed and the users are marked as settled users. Otherwise, these users will remain unsettled users. 3) BSs with surplus capacity will be allocated to the unsettled users. Similarly, each unsettled user selects the BS that gives the maximum weighted utility from those BSs. 4) Repeat step 3 until either all BSs reach their capacity or there are no more unsettled users. 26

27 QoE-Aware Optimization AP Allocation The direct link between users and associated BS will be tested, and direct transmission will be allocated if the error probability is below ε. By this criterion, the settled users are divided into direct-transmission users and non-direct-transmission users. The objective of the relay AP allocation is: 27

28 QoE-Aware Optimization 28

29 QoE-Aware Optimization capacity-aware approach, the maximum-harmonic-mean (MHM) method allocates the relays and BSs to maximize the harmonic mean SNRs on the allocated links. 29

30 QoE-Aware Optimization 30

31 Future trends UDNs represent a feasible way to reach the 1000x capacity target in indoor environments. From a strict engineering perspective, there are no limitations on the capacity that can be achieved, and area capacities in excess of 1 Gb/sm2 are thus clearly within reach. Instead, the question is if this can be done in a scalable and affordable way. Future trends: Secondary Spectrum Availability UDN Cloud RAN UDN Energy Management UDN Complexity and Backhaul Cost 31

32 Thanks for your attention Any questions? 32