A Mechanism for Uplink Packet Scheduler in LTE Network in the Context of Machine-to-Machine Communication Adyson M. Maia∗ , Dario Vieira† , Miguel F. de Castro∗ , Yacine Ghamri-Doudane‡ ∗ GREat Lab - Federal University of Ceará (UFC), Fortaleza, Brazil School of Information and Digital Technologies (EFREI), Villejuif, France ‡ University of La Rochelle, La Rochelle, France Email: adysonmaia@great.ufc.br, dario.vieira@efrei.fr, miguel@great.ufc.br, yacine.ghamri@univ-lr.fr † Engineering Abstract—This paper proposes a mechanism for uplink packet scheduler in LTE network in the context of Machine-to-Machine (M2M) communication. The proposed approach uses the current and past information of the system to satisfy the Quality of Service (QoS) requirements, to ensure fairness in resource allocation and to control the congestion caused by M2M devices. We carried out some network simulations by using a NS-3 simulator so as to show the effectiveness of the proposed approach. The results indicate that our solution can reduce the impact of M2M communication on Human-to-Human (H2H) communication and avoid the problem of starvation, when compared to related approaches. Keywords—LTE; Internet of Things; IoT; Machine-to-Machine; M2M; uplink packet scheduler I. I NTRODUCTION In the Internet of Things (IoT), a great variety of smart objects (home appliances, automobiles, cell phones, etc.) will be connected to the Internet. In this new scenario, where these smart objects will be connected and interacting with each other, Machine-to-Machine (M2M) communication will play an important role in the deployment of IoT [1]. M2M communication refers to the transmission and automatic collection of data from a remote source without, or with only limited, human intervention through a public network infrastructure (Wi-Fi, WiMax, UMTS, LTE, etc.) [2]. There is a large market potential for M2M communication, and the application fields of this communication are very wide [2]. Furthermore, the quantity of objects that will be connected in the future IoT is estimated at billions or even trillions [1]. Due to this vast number of smart objects, a large volume of traffic generated by the M2M communication is expected. Accordingly, this will exceed the volume generated by human-oriented communications (VoIP, media streaming, web browsing, etc.), also called Human-to-Human (H2H) communication, which are responsible for most of the traffic in the Internet at present. It is expected also that cellular network infrastructures will play a key role in the deployment of IoT, as well as for M2M communication, highlighting the significance of the LTE (Long Term Evolution) network for this deployment [3]. However, improvements in the LTE are necessary because of the intrinsic characteristics of M2M communication (discussed in more detail in Section II), which are different from H2H communication, for which the network was originally designed. Among these improvements, the uplink channel packet scheduler requires special attention. Existing solutions in the literature [4] that address scheduling for the current communication (H2H) in LTE are not suitable for M2M communication, because of the assumption of the low volume of services and the Quality of Service requirements (QoS) of H2H communication [5]. Subsequently, this is not the case for the majority of M2M applications, which have a huge amount of services and therefore different QoS requirements [5]. Furthermore, these solutions do not cope with the congestion caused by the deployment of a huge number of smart objects in the same area and, therefore, using the same shared resources. The current solutions that handle M2M communication [5]–[10], which are discussed in Section III, have some shortcomings. One of them is that some of these solutions do not ensure the fair allocation of resources and may lead to the problem of starvation. Another weakness is related to either the lack of control over the impact caused by the introduction into the network of M2M communication on the performance of the H2H communication, or this kind of impact control being poorly carried out. Within this context, in this paper, we discuss the packet scheduler in the LTE radio network, taking into account both M2M and H2H communication, and we propose a new mechanism for scheduling packets to uplink channel, details of which are presented in Section IV. The proposed scheduler uses the current traffic information of each device as well as allocation logs to classify the traffic and also to prioritize the allocation of radio resources. We argue that the proposed approach can (i) control the impact of M2M communication on H2H, (ii) ensure fairness and thus avoid the problem of starvation and (iii) meet the QoS requirements of applications. We present in Section V the experiments through simulations and we also discuss the results gained. Finally, in Section VI, we discuss the conclusions and future works. II. OVERVIEW This section briefly describes some technologies on M2M communication and LTE, in order to provide to required background of this article’s proposal. A. M2M Communication M2M and H2H applications have some features in common such as: secure connection, mobility and package switching [6]. However, since M2M applications perform their functions without (or with only limited) human intervention, several other characteristics differentiate them from H2H applications. Amongst these characteristics can be mentioned (i) the infrequent transmission at regular intervals of a small amount of data, (ii) delay tolerance and (iii) the main traffic is in uplink (devices to the server) [11]. The majority of M2M applications can be classified into two categories [12]. In the category of event-based applications, the devices send messages to the server when an event occurs in the monitored environment. Furthermore, applications in this category require reliable real-time transmission. Most M2M applications are included in the second category, called time-based. In this case, the devices send their collected data to the server at regular intervals of time (which can vary from a few milliseconds to several minutes or hours). As the devices in M2M communication generally send more data than they receive, we focus in this paper on the uplink packet scheduler. In the next subsection, we present a brief description of the packet scheduler in the LTE. B. Packet Scheduler for LTE Network The scheduling in the two channels (downlink and uplink) is performed separately (one scheduler for the downlink and another one for uplink) in the Medium Access Control (MAC) layer of the base station. The scheduling of both channels is performed each TTI (Transmission Time Interval) of 1 ms, and it is decided which Resource Blocks (RBs), the smallest unit that the scheduler can allocate to a device, are allocated to which device. This decision aims to meet the needs of device resources, as well as to maximize global system performance. An important characteristic that differentiates the downlink scheduler from the uplink scheduler is that the latter has the constraint that all the RBs allocated to a single device should be continuous with respect to frequency. III. R ELATED W ORKS In this section, we discuss some of the solutions that take into account M2M communication in the network. These solutions are compared according to their goals and the parameters used to accomplish them. In [6], the authors propose a scheduler group-based system in which devices are grouped into classes, dynamically created, according to their QoS requirements. The QoS requirements of the devices are placed into a class made up of two parameters: (i) the arrival rate of packets and (ii) the maximum tolerable jitter. However, in scenarios where there are many devices, the scheduler proposed in [6] is not fair to devices that belong to classes with lower priorities. In addition, the solution does not exploit information regarding the quality of the channel between the device and the base station, which influences the throughput. Two solutions are presented in [5] aiming to guarantee QoS requirements. For that, the authors apply two parameters as follows: (i) the channel quality between the device and the base station, and (ii) the maximum tolerable delay as QoS metric. The two solutions differ between them by the weight of the two parameters used in order to prioritize the QoS metric (either the channel quality or maximum tolerable delay). However, the two solutions are not fair to the devices that have the worst conditions in the relevant parameter used as metric. In [10], the authors propose a scheduler that prioritizes the devices that have a critical deadline. For this, the solution prioritizes the allocation of resources for devices that have more data to transmit and the waiting time for resources is closest to exceed the maximum allowed delay. Nonetheless, this solution also has no mechanism to ensure fairness in the allocation. In a real scenario, both H2H communication and M2M communication will be present on the network. In this context, the shortage of radio resources, and consequently, the network congestion is a problem to take into account. The main cause of this problem is that although each M2M device transmits small amounts of data resulting in less need for radio resources, the large amount of devices requesting resources leads to a shortage of resources. Accordingly, the H2H communication will be seriously affected if the scheduling does not take into account this problem. In this context, none of the solutions aforementioned deal with the impact of M2M communication on the H2H communication. The solution presented in [7] controls this impact by dividing the scheduler into two parts. In the first part, the scheduling allocates resources for H2H devices. In the second part, the scheduling is carried out for the others devices by using the remaining resources after the first part. However, it may occur that after the first part of the scheduling there are no more available resources for M2M devices in scenarios where there is heavy H2H traffic. As a result, this may lead to the problem of starvation for M2M devices. In [8], the authors propose some improvements on the work introduced in [6]. In this proposed approach [8], the impact of M2M communication on H2H communication is controlled by reserving a constant amount of resources for M2M communication. Moreover, to classify traffic in predetermined QoS classes, the Poisson process, applied to model the burst transmission of event-based applications as proposed in [13], is used instead of transmission at regular intervals in the time-based applications. However, in [8], there is neither guarantee of fairness nor use of the channel quality information. Moreover, the constant separation of resources for both M2M and H2H communications cause a poor utilization of these resources. In [9], the proposed scheduler also reserves a constant percentage of resources for M2M communication. But in this case, this percentage refers to an upper limit of resources that can be allocated for M2M communication. Moreover, this solution prioritizes the devices with less tolerance to delay, which results in no guarantee of fair allocation to devices more delay tolerable. Based on the shortcomings of the solutions discussed above, we propose a new mechanism for the packets scheduler that (i) controls the impact of M2M communication on H2H, (ii) ensures fairness in allocating resources and tries to (iii) maximize the satisfaction of QoS requirements. TABLE I. Index 10 11 12 13 14 15 16 17 IV. Q O S CLASSES FOR M2M APPLICATIONS Maximum Delay 50 ms 50 ms 100 ms 150 ms 300 ms 400 ms 500 ms 1s Packet Loss 10−6 10−2 10−2 10−2 10−2 10−2 10−2 10−2 Applications Event Based Time Based Time Based Time Based Time Based Time Based Time Based Time Based P ROPOSED M ECHANISM FOR S CHEDULING In this section, we describe in detail the mechanism proposed in this paper. In Subsection IV-A, we present the classification of the device according to its traffic flows. In Subsection IV-B, we describe the mechanism to control the impact on the performance of H2H communication with the introduction of M2M communication. Finally in Subsection IV-C, we present the scheduler algorithm. B. Calculation of Demand for Resources One goal of our work is to control the impact of M2M communication on H2H. For this purpose, the proposed mechanism will prioritize the H2H communication in the scheduling. However, the amount of resources allocated to the H2H devices so that they do not suffer large reductions in their performances is not an easy task. The main reason for this is that the amount of data (in bits) to be transmitted is not directly related to the amount of RBs needed for this transmission. Indeed, the throughput is influenced, beyond the amount of RBs, by the quality of the transmission channel between the device and each RB allocated. To calculate the current demand for resources by H2H devices, we use the ratio of the average amount of resources allocated and the average size of data in the transmission buffer as shown below: avg d H (u, t) = BS(u, t) × RB (u, t − 1) RB avg BS (u, t − 1) (1) A. QoS Classes for M2M The LTE standard supports end-to-end quality of service for the main types of H2H applications. Traffic flows can be classified into one of nine classes of QoS defined according to their requirements as specified in [14]. The QoS requirements are defined by the following parameters: (i) the maximum tolerable delay and (ii) the rate of packet loss. However, these nine classes do not encompass the wide range of maximum delay (a few milliseconds to several minutes) of time-based applications. Furthermore, the event-based category requires high priority, low tolerance for delay and high reliability (low rate of packet loss). Thus, the only possible relationship for this category is a class that is used exclusively for signaling messages. To address this issue, we propose two approaches to extend the QoS classes to support M2M communication. Both proposals have the advantage of identifying the devices (M2M or H2H) only by the QoS class to which their traffic is related. The proposed approaches are defined as follows: 1) 2) In the first approach, two classes are added to the LTE standard. One class is used for event-based applications and the other ones is used for time-based applications. Nevertheless, the maximum allowable delay parameter is not used in the latter class. Instead, each device of time-based applications sends to the network (base station) its own value for this parameter. In the second approach, 1 + N classes are added to the standard. As in the previous approach, one class is used for event-based applications, but now N classes for time-based applications are used. These N classes differ in the value of the maximum tolerable delay. Therefore, a device that transmits its data in the time interval equal to x chooses the class with the largest maximum allowable delay less or equal to x. Table I illustrates an example of the proposed approaches. The first approach uses only the first two classes (Index 10 and 11) and the second one all other classes illustrated in the table. d H (u, t) is the current demand for RBs from dewhere RB vice u at TTI t, BS is the current size of the data in the buffer, RB avg is the average number of RBs allocated to the device and BS avg is the average size of data in the transmission buffer. Moreover, each H2H device has a min minimum threshold of RBH (greater than zero) required resources. The functions RB avg , BS avg are calculated using an Exponential Moving Average (EMA). With the demand for resources calculated, we can reserve an amount of RBs for H2H communication equal to the sum of the demand of each H2H device. However, the H2H devices may require all available resources causing starvation for M2M devices. To avoid this, a constant minimum percentage of resources (0 ≤ ρ ≤ 1) is ensured for M2M. Nevertheless, this minimum percentage of resources is reserved only if there are sufficient devices request them. Furthermore, as M2M devices transmit small amounts of data (less than 1000 bits) [13], then a M2M device will receive a fixed and small number of min max resources RBM . Therefore, at most UM (t) M2M devices will receive resources at TTI t. The Equations (2a) and (2b) formalize this discussion, where RB is the set of available resources, UM is the set of M2M devices requesting resources and RBM (t) is the amount of resources reserved for M2M communication. Consequently, the amount of resources reserved for H2H is equal to the remaining resources as shown in Equation (2c). max UM (t)  = max %!  $ d H (t) |RB| − RB ρ|RB| , (2a) min min RBM RBM min max RBM (t) = RBM × min (|UM |, UM (t)) RBH (t) = |RB| − RBM (t) (2b) (2c) C. Scheduler Algorithm Once the demands for resources are calculated, the RBs are divided into two consecutive groups regarding the frequency, one for each type of communication (H2H and M2M) and with size equal to its demand. Thereby the scheduling for H2H and M2M can be handled separately. Therefore the mechanism proposed in this paper focuses only on scheduling for M2M devices, benefiting from the use of any existing uplink scheduler in the literature for H2H communication. Based on the model specified in [15], we divide the proposed algorithm into two phases as detailed below. 1) First Phase: This phase has the role of choosing which devices will receive resources. The devices that received fewer resources over time and that are closest to overcoming the maximum tolerable delay have higher priority to be selected for the next phase. Thus, the goal of the first phase is both to ensure fairness in the allocation of resources and to satisfy the QoS requirements. M2M devices are prioritized according to the function TD MM shown in Equation (3), where T avg (u, t) is a moving average of the throughput that indicates whether the device u has received a fair amount of resources until the TTI t. ∆D(u) is a function that measures whether this device is close to not satisfying its QoS requirements. Lastly, the constant ω (0 ≤ ω ≤ 1) indicates the weight to ensure fairness (first component of the Equation (3)) and to satisfy the QoS requirements (second component of the Equation (3)) in prioritizing the devices. TD MM (u, t) T avg (u, t − 1) = (1 − ω) 1 − maxn∈UM T avg (n, t − 1)   ∆D(u) +ω 1− maxn∈UM ∆D(n)   (3) ∆D(u) is calculated in two ways. If the last request for resources was either granted or no request was made, then ∆D(u) is equal to the maximum tolerable delay. Otherwise, ∆D(u) is either equal to the difference between the maximum delay and the expected time since the last request not granted or zero if this difference is negative. min Given this, at most RBM /RBM M2M devices that have TD higher values by MM are selected for the next phase. The devices not selected will need to wait x TTIs to make a new request. The value of x is chosen, for each device, randomly in the interval [0, σ∆D], where σ is a constant (0 ≤ σ ≤ 1). Therefore the next requests are spread over time in order to reduce possible network congestion generated by the large amount of devices requesting resources. 2) Second Phase: This phase creates groups with the same amount of continuous RBs; one for each device selected by the previous phase. After the creation of the groups, it remains only to decide which group of resources will be allocated to which device. For this, we use the following greedy approach: 1) 2) 3) 4) For each resources group g and device u, calculate the expected throughput if the device use this group. Let S be a set with these values of throughput. If S 6= ∅, then remove the highest value from this set. Otherwise, the algorithm terminates. Allocate resources group g for the device u that is associated with the removed value only if the device has not yet received resources. Repeat step 2. V. E XPERIMENTS A. Simulation Environment In order to evaluate our proposed mechanism, we evaluated it against two other solutions in the literature. The first solution is the scheduler Proportional Fair (PF), which is one of the most widely used and researched solutions in the literature for fair resource scheduling of H2H communication [16]. The PF was also used for the scheduling of H2H devices in the proposed mechanism. The second approach used in our simulation is the second algorithm proposed in [5]. In order to carry out the experiments, we had to define the values for two input parameters of the second algorithm proposed in [5]. The first input parameter was the RBs available for M2M communication, the value of which we have set as equal to the same parameter as that of our proposed scheduler. The second input parameter that we have set as a value is related to the amount of resources required by each M2M device. In this case, we have assumed that the available resources are divided equally between the M2M devices. The simulated scenario was of a near future urban environment, perhaps in the IoT, where there is a high traffic H2H and M2M. The H2H applications that we used were VoIP, Video and FTP. The modelings of their traffic flows were based on the works of [17] (VoIP and FTP) and [18] (Video). These three traffic flows have been chosen to cover the various QoS classes in the LTE standard. Moreover, we assumed both that each device generates only one type of traffic and that each H2H traffic is generated for an equal number of devices. The burst transmissions of event-based applications were modeled by a Poisson process with rate γ = 0.02 packets/TTI [8]. The transmission interval of each time-based device was uniformly distributed between 50 and 550 ms. The two categories of M2M applications have a packet size of 125 bytes [13]. Aiming to simulate an environment with high H2H traffic, the experiment was composed of 30 H2H devices, 10 for each type of traffic. The devices (H2H and M2M) were uniformly distributed around a single base station. We have defined the number of devices per category of M2M applications such that the time-based category has a larger number of the devices, since most M2M applications belong to this category. Moreover, the bandwidth was set to 5 MHz, 25 RBs available per TTI (1 ms) in order to simulate an environment with limited resources even using a small number of devices. We used Table I to classify the M2M traffic according to the two approaches discussed in Section IV-A. The minimum min min amount of resources required per device (RBH and RBM ) was chosen to be able to transmit all buffered data in a single TTI if the device has good quality in the channel and the amount of data is small. The waiting time after the request is denied was established as a maximum of 10% (σ = 0.1) of the value of ∆D. We chose the value of the constant ω of Equation (3) through previous simulations. In these simulations, we varied the value of ω between 0 and 1, we used 3 H2H devices, 250 M2M devices and the second approach of Subsection IV-A. Furthermore, we configured the value of ρ equal to 0.48 so min that there were at least 4 (25ρ/RBM ) M2M devices getting resources per TTI. The decision of the value of ρ is discussed TABLE II. Simulation time Runs Distribution of devices Number of base stations Bandwidth Number of M2M devices Number of H2H devices H2H Traffic Video VoIP FTP M2M Traffic Event Based TIme Based PARAMETERS OF THE SIMULATIONS General Parameters 3 s (3000 TTIs) 30 Fixed and uniform within 1400 1 5 MHz (25 RBs) 0, 50, 100, 150, 200, 250; 1/3 Event Based, 2/3 Time Based 10 VoIP, 10 Video, 10 FTP Packet Size Inter-arrival Time 1200 bytes 40 bytes 256 bytes 75 ms 20 ms 16.625 ms 125 bytes Poisson Process, γ = 50 ms 125 bytes Uniformly distributed [50, 550] ms Parameters of the Proposed Mechanism min min RBH , RBM 3 2M ρ 0, 0.24, 0.48, 0.72, 0.96 ω 0, 0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9, 0.95, 1 σ 0.1 Constant of EMA 99 B. Results As discussed in Section IV-A, two approaches are proposed in this paper to classify M2M traffic according to its QoS requirements. In the remainder of this paper, the first approach is referred to as “P roposalV 1” and the second one as “P roposalV 2”. The second scheduler proposed in [5] is designated as “Lioumpas2”. Finally, the “Proportional Fair” scheduler is identified as P F . 1) Impact of M2M communication on H2H: Figure 2(a) presents the mean values of the throughput for H2H devices in relation to variation of the number of M2M devices. Note that since the P roposalV 1, V 2 and Lioumpas2 schedulers use the same subroutine to separate resources between H2H and M2M communications, then only the P roposalV 2 and P F schedulers are shown in the Figures 2(a), 2(b) and 2(c). Furthermore, to analyze how the parameter ρ influences the performance of H2H communication, we varied its value according to Table II. By analyzing Figure 2(a), we observed that the P F presented a drastic drop in throughput as the number of M2M devices increased. P roposalV 2 also presents decreases in its rates in almost all cases and this decrease is greater with the increase of value of ρ. But even with ρ = 0.96 the drop in throughput is less steep than the P F . Fig. 1. Level of non-satisfaction of QoS requirements (ρ = 0.48, 30 H2H devices, 250 M2M devices) - M2M Traffic. in the next subsection. Figure 1 presents the results of these simulations regarding the level of non-satisfaction of QoS requirements for M2M traffic flows. Therefore, as we can see in Figure 1, ω = 0.95 had the best results. Moreover, we can observe in this figure that occurs improvements in QoS satisfaction when the second component of Equation (3) has greatest weight. This is explained by the fact that eventbased devices, which are less delay tolerant, begin to have higher priority in scheduling. However, when the fairness is disregarded (ω = 1), the satisfaction of QoS is also compromised. In order to evaluate the performance of our proposed solution, we used three metrics: (i) the throughput, (ii) the percentage of packets that do not satisfy the constraint of the maximum tolerable delay and (iii) the Jain’s fairness index [19]. We have defined the simulation time (3 s) to contain a reasonable amount of schedules (3000). To obtain reliable results as discussed in [19], we carried out 30 simulations, generating random and independent samples. In addition, the confidence interval of 95% was used to analyze the results. Finally, Table II summarizes the main parameters used in the simulations. Other parameters not shown in the table have default values of NS-3 [20] simulator used in the experiments. Figure 2(b) presents the mean values of the percentage of packets for H2H traffic flows that exceed the maximum tolerable delay. Just as in the graph of throughput (Figure 2(a)), we observed that P roposalV 2 performed better than P F in satisfying QoS requirements (low percentage of packets that exceeded the maximum delay) of H2H applications. In addition, we observed that with the increase of value of ρ the level of non-satisfaction of QoS requirements also tends to increase. Analyzing Figures 2(a) and 2(b), we can conclude that low values of ρ result in low impacts on performance of H2H communication. However, as can be seen in Figure 2(c), these low values have negative effects on satisfaction of QoS requirements of M2M traffic flows. Hence, we choose ρ = 0.48, for the remaining simulations of this paper, because it have a good trade-off between the metrics discussed in this paragraph. 2) Satisfaction Level of QoS requirements: Figure 3(a) presents the mean percentage of packets of event-based applications that exceeded the maximum tolerable delay. Similarly, Figure 3(b) shows the results obtained for the time-based applications. In Figure 3(a), we can see that the P roposalV 1 and P roposalV 2 present an increase in its values. This increase is associated with the guarantee that the time-based devices received resources. Also for this reason, the P roposalV 1 and P roposalV 2 present better results, in Figure 3(b), than Lioumpas2 since this later gives low priority to time-based devices and does not ensure fair allocation of resources. However, Lioumpas2 presented better results for event-based applications when the number of devices was high (greater than 180). 3) Fairness in the allocation of resources: Figures 4(a) and 4(b) show the mean of the fairness index values for event- (a) Throughput - H2H Traffic (b) Level of non-satisfaction of QoS requirements - H2H Traffic (c) Level of non-satisfaction of QoS requirements - M2M Traffic Fig. 2. Impact of M2M communication on H2H communication. (a) Event-based applications Fig. 3. (b) Time-based applications Level of non-satisfaction of QoS requirements for M2M communication (ρ = 0.48, ω = 0.95). and time-based applications respectively. In order to better understand the results shown in these figures, the fairness index can be interpreted as follows: the closer the value of the index to 1, the fairer the result. Consequently, values closer to zero indicate that the result is more unfair, hence a greater the probability of starvation occurs. The Lioumpas2 scheduler presented reductions in fairness index for both categories of M2M applications. This reduction is more severe for time-base applications, which have lower priorities in this scheduler. P roposalV 1 and P roposalV 2 presented a slight increase in the index of fairness for time- based applications. This result is related to the increase in traffic that has the highest priority in these schedulers as defined in Section V-A. For this same reason, the P F scheduler has shown an increase in their values, but in this case, for the time-based applications. VI. C ONCLUSION AND F UTURE W ORKS In this paper, we present a mechanism for the packet scheduler in the uplink of LTE network to treat M2M communication using historical information about resource allocations, channel quality and QoS requirements of devices to (i) control (a) Event-based applications Fig. 4. (b) Time-based applications Index of fairness for M2M communication (ρ = 0.48, ω = 0.95). the impact of M2M communication on H2H, (ii) avoid the problem of starvation with the fair allocation of resources and (iii) satisfy the QoS requirements. From the analysis of the results obtained by the simulations, we observe that our approach to separating the resources between H2H and M2M communications, so that the amount of resources allocated to the M2M devices is controlled, is a viable alternative to controlling the impact of M2M communication on H2H communication. Furthermore, the proposed scheduler avoids the problem of starvation by ensuring fair allocation of resources. However, in an environment with a high volume of H2H traffic and in order to meet the objectives (i) and (ii) aforementioned, the satisfaction of QoS requirements of event-based applications is compromised. Moreover, the results demonstrated that the approach of extending the QoS classes of the LTE standard to add more n classes for M2M applications in the time-based category has similar results to the approach of adding a new control message which the device uses to send its own values of QoS requirements. Therefore, the first approach avoids the generation of more traffic in a congested environment. [6] [7] [8] [9] [10] [11] [12] [13] Interesting future works that we can mention are: (i) the use of the characteristic of group base of M2M devices in scheduling and (ii) the adaptation of the constants of the mechanism proposed in this paper according to the state of the system. [14] [15] R EFERENCES [1] G. Wu, S. Talwar, K. Johnsson, N. Himayat, and K. Johnson, “M2m: From mobile to embedded internet,” Communications Magazine, IEEE, vol. 49, no. 4, pp. 36–43, 2011. [2] D. Boswarthick, O. Elloumi, and O. Hersent, M2m communications: a systems approach. Wiley. com, 2012. [3] D. Miorandi, S. Sicari, F. D. Pellegrini, and I. Chlamtac, “Internet of things: Vision, applications and research challenges,” Ad Hoc Networks, vol. 10, no. 7, pp. 1497 – 1516, 2012. [Online]. Available: http://www.sciencedirect.com/science/article/pii/S1570870512000674 [4] R. Kwan and C. Leung, “A survey of scheduling and interference mitigation in lte,” JECE, vol. 2010, pp. 1:1–1:10, Jan. 2010. [Online]. Available: http://dx.doi.org/10.1155/2010/273486 [5] A. Lioumpas and A. Alexiou, “Uplink scheduling for machine-tomachine communications in lte-based cellular systems,” in GLOBECOM Workshops (GC Wkshps), 2011 IEEE, 2011, pp. 353–357. View publication stats [16] [17] [18] [19] [20] S.-Y. Lien, K.-C. Chen, and Y. Lin, “Toward ubiquitous massive accesses in 3gpp machine-to-machine communications,” Communications Magazine, IEEE, vol. 49, no. 4, pp. 66–74, 2011. S. Zhenqi, Y. Haifeng, C. Xuefen, and L. Hongxia, “Research on uplink scheduling algorithm of massive m2m and h2h services in lte,” in Information and Communications Technologies (IETICT 2013), IET International Conference on, 2013, pp. 365–369. A. G. Gotsis, A. S. Lioumpas, and A. Alexiou, “Analytical modelling and performance evaluation of realistic time-controlled m2m scheduling over lte cellular networks,” Transactions on Emerging Telecommunications Technologies, vol. 24, no. 4, pp. 378–388, 2013. [Online]. Available: http://dx.doi.org/10.1002/ett.2629 I. Abdalla and S. Venkatesan, “A qoe preserving m2m-aware hybrid scheduler for lte uplink,” in Mobile and Wireless Networking (MoWNeT), 2013 International Conference on Selected Topics in, Aug 2013, pp. 127–132. N. Afrin, J. Brown, and J. Khan, “Performance analysis of an enhanced delay sensitive lte uplink scheduler for m2m traffic,” in Telecommunication Networks and Applications Conference (ATNAC), 2013 Australasian, Nov 2013, pp. 154–159. 3GPP, “Service requirements for machine-type communications,” 3GPP, Tech. Rep. 22.368, March 2012. A. Amokrane, A. Ksentini, and Y. Hadjadj-Aoul, “Congestion control in the context of machine type communications in long term evolution networks,” Master’s thesis, ENS Cachan Bretagne, 2011. 3GPP, “Analysis on traffic model and characteristics for mtc and text proposal,” 3GPP, Technical Report R1-120056, February 2012. [Online]. Available: http://www.3gpp.org ——, “Policy and charging control architecture,” 3GPP, Technical Specification 23.203, Septembre 2013. A. Pokhariyal, K. Pedersen, G. Monghal, I. Z. Kovacs, C. Rosa, T. Kolding, and P. Mogensen, “Harq aware frequency domain packet scheduler with different degrees of fairness for the utran long term evolution,” in Vehicular Technology Conference, 2007. VTC2007-Spring. IEEE 65th, 2007, pp. 2761–2765. S.-B. Lee, I. Pefkianakis, A. Meyerson, S. Xu, and S. Lu, “Proportional fair frequency-domain packet scheduling for 3gpp lte uplink,” in INFOCOM 2009, IEEE, 2009, pp. 2611–2615. M. Salah, N. Ali, A.-E. Taha, and H. Hassanein, “Evaluating uplink schedulers in lte in mixed traffic environments,” in Communications (ICC), 2011 IEEE International Conference on, 2011, pp. 1–5. T. Potsch, S. Marwat, Y. Zaki, and C. Gorg, “Influence of future m2m communication on the lte system,” in Wireless and Mobile Networking Conference (WMNC), 2013 6th Joint IFIP, 2013, pp. 1–4. R. Jain, The art of computer systems performance analysis. John Wiley & Sons Chichester, 1991, vol. 182. NS-3, “The network simulator ns-3,” http://www.nsnam.org, 2013.