|Ali Ghubaish (A paper written under the guidance of Prof. Raj Jain)||Download|
Machine-to-Machine (M2M) communication is the first step technology for machines communicating with each other without the human factor. Third-Generation Partner- ship Project (3GPP) Long-Term Evolution (LTE) and Long-Term Evolution-Advanced (LTE-A) cellular network technologies provide the resources that M2M communication required. However, LTE/LTE-A is required to satisfy the requirements of M2M communication, such as power management and supporting massive number of devices while maintaining the quality of service (QoS) requirements for these devices. In this paper, we are going to present four scheduling techniques that satisfy these requirements: power efficient, QoS, multi-hop, scalability.
LTE/LTE-A Scheduling Techniques, M2M Communication, D2D Communication, Power Efficient Scheduling, QoS Scheduling, LTE/LTE-A Relaying, Multi-hop Scheduling, Scalable Network Scheduling.
Machine-to-Machine (M2M) communication is the technology that will dominate the world communications technologies because it will connect everything around us to the Internet, where the whole world is connected and can be seen by everyone. Today, one percent of the things around us is connected to the Internet, which is called Internet of Things (IoT), that equals to 10 Billion devices, including things and computer users [Bradley13]. However, this number according to the statistics will increase to 50 Billion in 2020, which will require a network technology that can handle that amount of connected devices to the Internet while maintain the M2M communication requirements, such as limited batteries. A general example of IoT network architecture is presented in Fig. 1. The Figure shows how M2M communications occur through cellular network and the ability of Device-to-Device (D2D) communications.
To achieve that, M2M communication requires an infrastructure that could handle the massive number of devices which can be possible through 3GPP LTE/LTE-A cellular network technology. LTE/LTE-A provides a set of features that make it the perfect candidate for M2M communication, such as IP native connectivity, large capacity, flexible allocation of radio resources, and scalability [Mehaseb15].
|Figure 1. A General Example of IoT Network Architecture.|
In this paper, we explain scheduling techniques that satisfied M2M requirements over LTE/LTE-A. In Section 2, a background for M2M communication and LTE/LTE-A scheduling techniques classifications are presented. In Section 3, we presented a scheduling techniques for power efficient of M2M communications. In Section 4, throughput and delay scheduling techniques are discussed to satisfy the QoS requirements of M2M communications. In Section 5, relaying and D2D communications scheduling techniques are presented for M2M multi-hop requirements. In Section 6, we explained a scheduling technique to satisfy the massive number of devices in M2M communications. Finally, a summary of this paper is presented.
In this section, we explain some of the characteristics for M2M communications and the scheduling techniques that are used via LTE/LTE-A.
M2M communications have recently become an interesting topic because of the ability of this type of communication to communicate without human interaction. However, this type of communication has a set of characteristics need to be achieved in order to be able to work efficiently [Ghavimi14]. These characteristics can be addressed in seven points:
LTE/LTE-A scheduling techniques classifications are based on M2M communication requirements to achieve optimal integration. Based on these characteristics, we can conclude that M2M communications require four main requirements [Mehaseb15]. First, power efficient scheduling uses Single-Carrier Frequency Division Multiple Access (SCFDMA) for uplink and Orthogonal Frequency Division Multiple Access (OFDMA) for downlink. Second, QoS based scheduling is required to handle different systems QoS requirements such as latency, jitter, error rate, and Guaranteed Bit Rate (GBR). Third, multi-hop based scheduling uses multiple hops to send data over short distances instead of sending it over long distance using single hop, which helps saving power. Finally, scalable network based scheduling uses LTE-A IP native connectivity feature to support a massive number of M2M communications.
In this section, we explain two different categories that explain different methods of using energy efficiently through LTE/LTE-A communications [Mehaseb15]. First, allocating fewer Physical Resource Blocks (PRBs) per device. Second, transmitting data in a low data rate.
This category uses a scheduler that uses dynamic scheduling technique that offers resources based on the channel state, which means better usage of the resources that are provided by the network [Ghandour11]. This technique consists of two stages:
A. Time Domain Packet Scheduling (TDPS): In this stage, a set of users will be selected to be served based on priority order, according to GBR value of each user by using a metric system. This metric, which can be expressed in Equation (1), is updated in every new Transmission Time Interval (TTI). Where the rate achieved by user i at the k subframe is Ri, k. The priority of scheduling user i in subframe j is expressed in Mi,j. the GBR value of user i is GBRi, as can be seen in Table 1.
|Mi, j (t) =||∑ j-1 k=1 Ri, k||, where j ∈ [1, m]||(1)|
B. Frequency Domain Packet Scheduling (FDPS): In this stage, each user gets assigned satisfied number of PRBs, which means that the best or second best number of PRBs each user requires, according to their Channel Quality Indicator (CQI) value to save power energy, which can be expressed in Equation (2). Where the total saved power when reducing the number of PRBs from N is ΔPi,N,M. N represents the highest CQI for user i to M, which represents the satisfied CQI for user i.
ΔPi,N,M = 10log10(N/M) (2)
Table 1. CQI Based Resource Type Classifications (Based upon [Jang13])
|QCI||Bearer Type||Priority||Packet Delay Budget||Example|
|3||3||50 ms||Online Gaming|
|4||5||300 ms||Video Streaming|
|5||Non-GBR||1||100 ms||IMS Signaling|
|6||6||300 ms||TCP, FTP|
This category reduces the power consumption and adopts the transmission rate based on the channel state, the queued packets in the buffer, and the average delay [Li09, Salodkar10]. Based on the channel state, the consumption power can be reduced if the average delay is slightly increased. The scheduler in this category selects Un packets to fill the queue, which is expressed in Equation (3), for transmission at the beginning of every new TTI and uses Pin power, input power, for transmission. Where the number of generated packets in the buffer is xn. The number of packets to be scheduled is determined by k. P*(D ave i ) is the minimum average power that is required to achieve average delays per packet no greater than D ave i , which can be expressed in Equation (4). The scheduler works if each user generates only one type of traffic with average arrival rate λi.
Un = min(xn, log⌊kxn⌋) (3)
In this section, we explain two different categories that explain different methods of using QoS requirements through LTE/LTE-A communications [Mehaseb15]. First, throughput-based scheduling. Second, delay-based scheduling.
This category uses a scheduler that uses priority order of devices is based on the QoS class identifier (QCI), which means that the priority of every device is based on its priority value in Table 1 [Jang13]. This technique consists of three stages:
A. TDPS: A In this stage, the scheduler differentiates between devices is according QCI in Table 1, which can be expressed in Equations (5) and (6) [Mehaseb15]. The devices with GBR resources type are allocated PRBs first then devices with non-GBR resources type.
B. FDPS: In this stage, the scheduler admits new connection is based on the number of RPBs that is available, which should satisfy the channel quality of that device.
C. Adaptive Contiguous PRBs: In this stage, the scheduler assigns a group of contiguous RPBs to each device according to the CQI of each device.
Where NGBR (i,t) and NNon-GBR (i,t) are the number of PRBs for user i at time t for GBR and Non-GBR. GBRi is the GBR of user i. Qqi (i) is the number of the queued packets of user i in queue qi. NRE is the number of resource elements of PRBs. u is type of the queue corresponding to QCI ranks. nqi (i,t) is the number of the queued packets of user i in queue qi at time t. L is the length of packet data. d is the packet delay budget. MClevel (i,t) is the number of data bits that can be transmitted by a resource element of user i at time t [Mehaseb15].
This category uses a scheduler that provides Human-to-Human (H2H) and M2M traffics with the required QoS requirements for both traffics [Zhenqi13]. The scheduler divides these two traffics into two queues:
H2H Queue: This queue provides services for H2H and Time-sensitive M2M communications, which means that the users in this queue have higher priority than the another queue. The scheduler uses a computational formula that guaranties no delay, as expressed in Equation (7). Where ric (t) represents the instantaneous channel rate of user i at time t. Mi (t) is the satisfaction, which is a new concept aspect for this formula over frequency-domain promotional fair algorithm. Di (t) represents the waiting delay of user’s queue. THi is the service delay threshold of user i. Ri (t) represents the average transmission rate that user i can reach within the time window before time t. Ti is the target rate when user i satisfies the QoS of service.
|λic (t) = (||ric (t)||) . exp (||Di (t)||) , where Mi (t) = (||Ri (t)||) (7)|
M2M Queue: This queue provides services for none Time-sensitive M2M communications. The system’s timeline is divided into cycles of period T. The length of T is based on the packet delay in Table 1. The scheduler uses round-robin algorithm to find M2M device’s PRBs if they exceeded the delay threshold.
In this section, we explain two different categories that explain different methods of using multi-hop schedulers through LTE/LTE-A communications [Mehaseb15]. First, LTE Relaying. Second, D2D communications.
In this category, the transmitter uses three different relaying schemes: Layer 1 (Amplified and Forward (AF)), Layer 2 (Decode and Forward (DCF)), and Layer 3 (Demodulate and Forward (DF)). Table 2 summarizes the advantages and disadvantages of each scheme [Lo09, Yang09]. The transmitter uses three types of links for each scheme: the link that connects the user equipment (UE) and backhaul link(eNB) is the direct link; the link that connects the UE and Relay Node(RN) is the access link; and the link that connects the RN and eNB is the relay link.
Table 2. Comparison of Different Relaying Schemes (Based upon [Mehaseb15])
|Layer 1 (AF)|
|Layer 2 (DCF)|
|Layer 3 (DF)|
LTE-A release 12 supports D2D communications. D2D communications support direct connection between two nodes without the need to eNB. Two types of Spectrum sharing: in-band spectrum and out-of-band spectrum [Wang13]:
5.2.1 In-band spectrum: In this spectrum, the nodes use the same band as cellular networks for D2D communications, where cellular network users and D2D communication are sharing the same resources. This spectrum has two categories:
A. Overlay: In this category, D2D can communicate with each other by using cellular network resources with orthogonal resource mode to keep the D2D nodes from interfere with other cellular users. Only one D2D link can use the same resources even if it is in inactive mode. However, a new study shows that multiple D2D links can use the same cellular network resource only if they do not interfere with other cellular users, which does not apply in the underlay category [Phunchongharn13].
To make D2D communication possible, it requires a three step procedure: peer discovery, link quality, and feedback control. In [Yang13], a study conducted about the first procedure. A distributed peer discovery protocol based on an adaptive resource allocation algorithm is used to allow peer discovery for D2D communications through random access channel. To achieve D2D communication, each UE needs to follow two steps. First, PRB allocation by the eNB. In this step, every UE needs to get a permission from eNB to connect directly to another UE through allocated PRBs, which are a number of blocks in a time slot that each UE can transmit/advertise for himself and scan for other UEs on different time slots, from the eNB. Second, beacon transmission by UEs. In this step, every UE can communicate with another UE based on two models services: client-server service where there are UE sever, who is the only can transmit a beacon, and UE client, who only can scan for UE servers to get their services. peer-to-peer service where all the UEs can transmit beacons and at the same time can scan for other UEs for information, but they can see each other if they are using different PRBs in the same time slot.
In [Bae14], a study conducted about the second procedure of making D2D communications. Mode switching is introduced to make a better connecting between the UEs. D2D communication uses direct connection if one of the two connected devices has better power signal over the eNB through a cellular network link, which can be accomplished by using cell reference signal. In every duration of Time-to-Trigger (TTT), mode switching from cellular network link to D2d link occurs when the signal-to-noise-plus- interference ratio of D2D link becomes better than cellular network link and visa versa. TTT duration need to be set carefully because if it is long, it would lead to a delay in the mode switching criteria, which results of losing the benefit of the mode switching criteria.
B. Underlay: In this category, D2D can communicate with each other by using cellular network resources with non-orthogonal resource mode that may cause interference if other cellular users use the same resources.
In [Liu14], In [Liu14], D2D communication required to be based on load balancing algorithm, which focuses on macro cells, pico cells, and femto cells in LTE-A networks, to reduce the congestion on some of tire-cells and distribute it among nearby uncongested cells. The algorithm follows four steps to accomplish load balancing among all cells. In addition, this algorithm will not move to the next step unless it fails to serve the new requesting UE in the current step. First step, the congested macro eNB tries to forward the new requesting UE to a nearby uncongested macro, pico, or femto cell to allocate required PRBs for that UE through an idle UE as a relay between the uncongested cell and the requesting UE. Second step, the congested macro eNB tries to forward a currently connected UE to an uncongested cell through an idle UE as a relay between the uncongested cell and the forwarded UE, so the congested macro eNB can allocate required PRBs for the new requesting UE. Third step, the congested macro eNB tries to forward the new requesting UE to a congested cell, which can forward a currently connected UE to an uncongested cell, so the second congested cell can allocate required PRBs for the new requesting UE. Forth step, the congested macro eNB tries to forward a currently connected UE to a congested cell, which can forward a currently connected UE to an uncongested cell, so the first congested cell can allocate required PRBs for the new requesting UE.
5.2.2 Out-of-Band spectrum: In this spectrum, the nodes use unlicensed bands for D2D communications such as 2.4 and 5.8 GHz ISM band. The nodes use Time Division Duplexing (TDD) to guarantee that each device doesn’t communicate with two devices at the same time.
In this section, we explain an enhanced Proportional Fair (PF) scheduling through LTE/LTE-A communications. The scheduler uses user grouping to satisfy the scalability objective [Songsong09]. In addition, the scheduler allocates PRBs to increase the ratio of the maximum data bit rate to the average received data bit rate. The enhanced PF scheduling is used to achieve the fairness among the users in the network by using a metric system as expressed in Equation (8) [Mehaseb15]. Where β is the weighting factor. ri (t) is the data bit rate over all PRBs allocated to user i to time t, which can be calculated by using Shannon’s theorem. Ri (t) is the long-term average data rate of user i up to time t.
Mi = βri (t)/Ri (t) (8)
In this paper, we provide a survey about scheduling techniques for M2M communication through LTE/LTE-A cellular network technology. These techniques are based on M2M communications requirements, such as power management, QoS, multi-hop, and scalable network. The power efficient is the first scheduling technique that focuses on how to allocate less PRBs for each device and using low data rate transmission to conserve energy. Next, QoS scheduling technique uses the throughput and the delay that the network link reliability is based on to satisfy M2M connection requirements. Multi-hop scheduling is the third among the requirements that investigate the LTE/LTE-A relaying nodes to support longer reach for machines. Also, this scheduling looks further in D2D communications, where each device can directly communicate with each other without the need for eNB. Finally, scalable network based scheduling that concerns about how to provide services for a massive number of devices through LTE/LTE-A network.
[Bradley13] Bradley, J.; “The Internet of Everything: Creating Better Experiences in Unimaginable Ways,” Nov 21, 2013. Available: blogs.cisco.com/ioe/the-internet-of-everything-creating-better-experiences-in-unimaginable-ways/#more-131793
[Mehaseb15] Mehaseb, M.; Gadallah, Y.; Elhamy, A.; El-Hennawy, H.; “Classification of LTE Uplink Scheduling Techniques: An M2M Perspective,” IEEE Communications Surveys and Tutorials, Vol. PP, No. 4, 2015, pp. 1. Available: http://ieeexplore.ieee.org/document/7339417/
[Ghavimi14] Ghavimi, F.; Hsiao-Hwa Chen; “M2M Communications in 3GPP LTE/LTE-A Networks: Architectures, Service Requirements, Challenges, and Applications,” IEEE Communications Surveys and Tutorials, Vol. 17, No. 2, 2014, PP. 525-549. Available: http://ieeexplore.ieee.org/document/6916986/
[Ghandour11] Ghandour, F.; Frikha, M.; Tabbane, S.; “A Fair and Power Saving Uplink Scheduling Scheme for 3GPP LTE Systems,” International Conference on the Network of the Future (NOF), 2011, PP. 6-9. Available: http://ieeexplore.ieee.org/document/6126686/
[Li09] Li, Z.; Yin, C.; Yue, G.; “Delay-Bounded Power- Efficient Packet Scheduling for Uplink Systems of LTE,” 5th International Conference on Wireless Communications, Networking and Mobile Computing (WiCom), 2009, PP. 1-4. Available: http://ieeexplore.ieee.org/document/5303491/
[Salodkar10] Salodkar, N.; Karandikar, A.; Borkar, V. S.; “A Stable Online Algorithm for Energy-Efficient Multiuser Scheduling,” IEEE Transactions on Mobile Computing, Vol. 9, No. 10, 2010, PP.1391-1406. Available: http://ieeexplore.ieee.org/document/5482581/
[Jang13] Jang, H.; Lee, Y.; “QoS-constrained Resource Allocation Scheduling for LTE network,” International Symposium on Wireless and Pervasive Computing (ISWPC), 2013, PP. 1-6. Available: http://ieeexplore.ieee.org/document/6707440/
[Zhenqi13] Zhenqi, S.; Haifeng, Y.; Xuefen, C.; Hongxia, L.; “Research on Uplink Scheduling Algorithm of Massive M2H and H2H Services in LTE,” IET International Conference on Information and Communications Technologies (IETICT), 2013, PP. 365-369. Available: http://ieeexplore.ieee.org/document/6617513/
[Lo09] Lo, A.; Niemegeers, I.; “Multi-hop Relay Architectures for 3GPP LTE-Advanced,” IEEE 9th Malaysia International Conference on Communications (MICC), 2009, PP.123-127. Available: http://ieeexplore.ieee.org/document/5431478/
[Yang09] Yang, Y.; Hu, H.; Xu, J.; Mao, G.; “Relay Technologies for WiMAX and LTE-Advanced Mobile Systems,” IEEE Communications Magazine, Vol. 47, No. 10, 2009, PP. 100-105. Available: http://ieeexplore.ieee.org/document/5273815/
[Wang13] Wang, P.; Wei, W.; Zhuoming, L.; “System Performance of LTE-Advanced Network with D2D Multi-hop Communication,” 2013 3rd International Conference on Consumer Electronics, Communications and Networks (CECNet), 2013, PP. 645-648. Available: http://ieeexplore.ieee.org/document/6703413/
[Phunchongharn13] Phunchongharn, P.; Hossain, E.; Kim, D. I.; “Resource Allocation for Device-to-Device Communications Underlaying LTE-Advanced Networks,” IEEE Wireless Communications, Vol. 20, No. 4, 2013, PP. 91-100. Available: http://ieeexplore.ieee.org/document/6590055/
[Yang13] Zhu-Jun Yang; Jie-Cheng Huang; Chun-Ting Chou; Hung-Yun Hsieh; Chin-Wei Hsu; Ping-Cheng Yeh; Hsu, C.-C.A.; “Peer discovery for device-to-device (D2D) communication in LTE-A networks,” Globecom Workshops (GC Wkshps), 2013 IEEE, 2013, PP. 665-670. Available: http://ieeexplore.ieee.org/document/6825064/
[Bae14] Hyung-Deug Bae, Jae-Wook Shin, and Pyung-Joong Song, “Mode Switching for Device-to-Device Communication in LTE-A Network,” MoMM '14 Proceedings of the 12th International Conference on Advances in Mobile Computing and Multimedia, 2014, PP. 7-10. Available: http://dl.acm.org/citation.cfm?id=2684108&CFID=599883770&CFTOKEN=23757807
[Liu14] Liu, J.; Kawamoto, Y.; Nishiyama, H.; Kato, N.; Kadowaki, N.; “Device-to-device communications achieve efficient load balancing in LTE-advanced networks,” IEEE Wireless Communications, Vol. 21, No. 2, 2014, PP. 57-65. Available: http://ieeexplore.ieee.org/document/6812292/
[Songsong09] Songsong, S.; Chunyan, F.; Caili, G.; “A Resource Scheduling Algorithm Based on User Grouping for LTE-Advanced System with Carrier Aggregation,” International Symposium on Computer Network and Multimedia Technology (CNMT 2009), 2009, PP. 1-4. Available: http://ieeexplore.ieee.org/document/5374801/
|3GPP||Third Generation Partnership Project|
|AF||Amplified and Forward|
|CQI||Channel Quality Indicator|
|DCF||Decode and Forward|
|DF||Demodulate and Forward|
|eNB||evolved Node B|
|FDPS||Frequency Domain Packet Scheduling|
|GBR||Guaranteed Bit Rate|
|IoT||Internet of Things|
|OFDMA||Orthogonal Frequency Division Multiple Access|
|PRBs||Physical Resource Blocks|
|QoS||Quality of Service|
|QCI||QoS class identifier|
|SCFDMA||Single-Carrier Frequency Division Multiple Access|
|TDPS||Time Domain Packet Scheduling|
|TDD||Time Division Duplexing|
|TTI||Transmission Time Interval|