In today's rapidly developing telecommunications technologies, mobile communication services are widely penetrating into all segments of society. The 5th generation network (5G) will cover a broader range of usage scenarios - enhanced mobile broadband (eMBB), ultra-reliable low-latency communications (URLLC), and massive machine-type communications (mMTC). This paper considers a model of joint service of URLLC and eMBB trafic within a single base station. The first type of trafic is supposed to have priority due to its high latency requirements, while the second type of trafic has a high speed. If there are no available resources to serve URLLC trafic, the allocated resources for eMBB trafic will decrease until they are completely withdrawn. We model the above mentioned system as a queuing system with the bit rate degradation and service interruption.
The life of society has been rapidly developing due to mobile communications, that has become an integral part of the everyday life of each person in society. Due to the rapid increase in the number of users and devices connected to the network, there is the growth of the load on communication networks. At the same time, network delays should be reduced and occur as rarely as possible.
The paper addresses the usage scenarios in 5G networks and joint service of eMBB and URLLC trafic. We apply methods of queuing theory and mathematical teletrafic theory. The research tasks are the following: 1. to analyze works related to joint transmission of eMBB and URLLC trafic; 2. to build a mathematical model with adaptive change in the speed of eMBB trafic; 3. to derive the probabilistic-temporal characteristics of the model.
Workshop on information technology and scientific computing in the framework of the XI International Conference CEUR
The paper is organized as follows. In Section 2, the characteristics of the coexistence of serving narrow-band URLLC and broadband eMBB trafic are explained. Section 3 presents the system model and probabilistic-temporal attributes for it. Furthermore, Section 4 shows us the numerical analysis and the discussion. As a result, the conclusions are in Section 5.
According to ITU-R recommendation [
It should be noticed that additional use cases are expected to emerge, which are may currently not foresee. Nevertheless, 5G network will encompass many diferent features. Figure 1 illustrates some examples of envisioned usage scenarios.
Since both eMBB and URLLC are essential components of communication trafic in 5G networks, various studies have looked at the coexistence of these services. In terms of network bandwidth, eMBB generates a massive amount of data trafic. Unlike eMBB, URLLC produces fewer data because of its stringent latency and reliability requirements. Consequently, the coexistence of these two services is associated with achieving the suficient eMBB throughput while meeting the URLLC requirements. Due to the time sensitivity of critical applications such as UAV automation, autonomous vehicle control, and critical medical equipment management, URLLC takes precedence over eMBB for scheduling. Typically, eMBB scheduling involves increasing the network capacity to improve the spectral eficiency, while the packet delivery reliability is ensured through re-transmissions. However, eMBB scheduling approaches may not guarantee the reliability and latency thresholds required for URLLC and thus cannot be applied in URLLC scheduling. In contrast, URLLC involves the transmission of short packets with specified latency and reliability margins.
URLLC VR/AR
Self driving car
Industry automation 5G
Cloud computing
Voice
Smart city mMTC
Smart house
Medicine
3D video UHD screens inGiagasebcyotensd eMBB
By the way, 3GPP has proposed a short and long transmission within the time interval
(TTI) frame allocation for these coexistence scenarios. In this frame allocation, eMBB trafic is
scheduled for a long TTI, and URLLC is automatically scheduled for a short TTI over existing
eMBB trafic by adopting a puncture or overlay scheme [
Exceptionally, the authors of [
Regarding that paper, we consider that an incoming eMBB session arrives and takes the whole slot while URLLC trafic takes a mini-slot. Indeed, one PRB equals one slot, and the duration of it is 5 ms. We will describe the model with an adaptive change of eMBB trafic rate in more detail in the next section.
There are several approaches to URLLC and eMBB coexistence. So [
In [
Let us consider a model where resources are allocated within one base station to service URLLC and eMBB trafic. eMBB trafic requests are distributed by the scheduler to the resource units of each slot. When a URLLC request is received, it can be assigned to any free resource unit of any subsequent mini-slot due to the delay requirements. Furthermore, the structure of the frame is presented in Figure 2.
Sessions of eMBB trafic occupy one resource block or 1 resource units (RU), while URLLC sessions occupy one resource unit. There are resource blocks in the system, i.e. the maximum number of eMBB sessions. Then the maximum number of URLLC sessions in the system is = 1 ⋅ .
Sessions of both trafic arrive according to the Poisson process, and the arrival rates are equal to and respectively. An eMBB session occupies the maximum speed. However, if all the resources of the system are busy when a URLLC session arrives, the rate of a eMBB session can be reduced since the second trafic is in priority. At the same time, the maximum speed will ifrst decrease, and then the lower ones in descending order. If several applications are served at the same speed, then the choice is made randomly.
Let us denote the total number of URLLC sessions active at a certain moment of time. To designate the number of eMBB sessions, we introduce the vector ⃗= ( 1, .., ), where the number of speeds at which eMBB sessions can be serviced, equal to the number of resource units 1. Also, we will use a service rate vector ⃗= ( 1, .., ) such that 1 > 2 > ... > . Then the state of the system will take the form ( 1, ..., , ) = (,⃗) = ⃗ . Since the requests are served according to an exponential distribution, the eMBB service rate will be , and URLLC – . All the main parameters of the system used in this work are presented in Table 1.
The scheme of the described model is presented in Figure 3. The system state space is: = ⎧
If the system is in a state (,⃗) , then various events are possible to occur with diferent intensities. All possible situations are presented in Table 2. For the convenience of describing transitions, we introduce the unit vector ⃗ = (0, ..., 0, 1, 0, ..., 0), where 1 is at the ℎ place.
First, consider the possible situations occurring with the intensity , i.e. when a new request eMBB arrives: • If the system has free resources, namely a free resource block, the request will be accepted for service. In Table 2, this transition from the central state is presented as number 1. • If there are no free resources in the system, then the session will be blocked.
In the case when an eMBB session service ends, i.e. an event with intensity which is presented as under number 2 from Table 2 occurs, the request leaves the system, and the resources are released.
We turn to events that occur with intensity , i.e. upon an receipt of the URLCC application: • If the system has at least one free resource unit, the session is accepted for service. In • If there are no free resources, but at least one eMBB session served not at the minimum speed, then the service speed of the eMBB session decreases, and the URLLC session is accepted for service. In Table 2 it is displayed as number 4. • If there are no free resources, but at least one eMBB session served at the minimum speed, then an eMBB session service is interrupted, and the URLLC session is accepted for service. In Table 2 this is confirmed as number 5. • If there are no free resources and no eMBB sessions in the system, the session will be blocked.
as number 6.
When a URLLC session leaves the system, events occur with an intensity : • If there are no eMBB sessions in the system or they are served at the maximum speed, then the URLLC session leaves the system and frees up resources. In Table 2 it is presented • If the system has at least one eMBB session that is not served at full speed, then after the termination of the URLLC session service, the speed of the active eMBB session is restored. In Table 2 it is shown as number 7.
Now let us talk about performance indicators of priority service of URLLC trafic. Having formed an infinitesimal generator matrix and solving the resulting system of equilibrium equations, we can find the stationary probability distribution (,⃗), (,⃗) ∈ following probabilistic-temporal characteristics of the model: . Based on it, we obtain the 1. Average number of eMBB sessions and URLLC sessions : = ∑ ⋅ ( 1, … , , ), = 1, … ; ∈⃗ = ∑ ⋅ ( 1, … , , ); ∈⃗ 2. Blocking probability of eMBB sessions and URLLC sessions : = ∑ ( 1, … , , ), ℬ2 = { ⃗∈ ∶ +
∑ + 1 > } ; ∈⃗ℬ 2 =1 = (0, … 0, ).
In this section, we present the results of a numerical analysis, namely the average number of eMBB and URLLC sessions and the blocking probability of eMBB/URLLC sessions. We will use two scenarios, which are presented in Table 3. The idea is that firstly is gradually increased for various constant values of , and vice versa. So we have that = 5 which is the total number of the resource blocks. We have = 5 , which is the number of speeds, so we will have 1 = 5, 2 = 4, 3 = 3, 4 = 2, 5 = 1. (a) Scenario 1
The following is the blocking probability of eMBB sessions. Obviously, the higher the intensity of incoming eMBB sessions, the greater the blocking probability of them. Figure 6.a clearly shows that with a very high rate of arrival of URLLC sessions, the rate of arrival of eMBB practically does not afect the blocking probability of eMBB sessions, which is very large and tends to be 1, since all resources are occupied by URLLC trafic. At the same time, when the (a) Scenario 1 intensity of URLLC sessions is low, the blocking probability of eMBB will be small since the system will cope with the load from the eMBB stream. The graph in Figure 6.b confirms the assumptions made. It can also be concluded that eMBB blocking is more influenced by the intensity of incoming URLLC sessions since the schedule behaves almost the same for the selected constant intensities of incoming eMBB trafic.
The blocking probability of URLLC sessions is afected by the rate of only this type of trafic since the system prioritizes it.
This paper is focused on a mathematical model for eMBB and URLLC coexistence in the form of a queuing system with priority service for URLLC trafic – reducing and interrupting the transmission rate of eMBB. In particular, we formulated indicators of priority access eficiency such as the probability of reducing the transmission rate, the probability of service interruption, the average transmission rate of eMBB trafic. Moreover, the numerical results show that URLLC has a significant impact on eMBB. In the future, we will consider the model that considers the spatial location of devices generating URLLC and eMBB trafic.
This paper has been supported by the RUDN University Strategic Academic Leadership Program (recipients Alexander Chursin, Petr Kharin, and Anna Kushchazli). The work was supported by the RFBR, project 20-37-70079 (recipients Irina Kochetkova and Petr Kharin). of joint urllc and embb transmission in 5g networks, in: Internet of things, smart spaces, and next generation networks and systems, Springer, 2019, pp. 635–648. [12] E. Makeeva, N. Polyakov, P. Kharin, I. Gudkova, Veroyatnostnaya model’ dlya analiza harakteristik sovmestnoj peredachi trafika urllc i embb v besprovodnyh setyah [probability model for perfomance analysis of joint urllc and embb transmission in 5g networks] (2020) 33–42. [13] P. Kharin, E. Makeeva, I. Kochetkova, D. Efrosinin, S. Shorgin, Sistema massovogo obsluzhivaniya s orbitami dlya analiza sovmestnogo obsluzhivaniya trafika s malymi zaderzhkami urllc i shirokopolosnogo dostupa embb v besprovodnyh setyah pyatogo pokoleniya [retrial queuing model for analyzing joint urllc and embb transmission in 5g networks] 14 (2020) 17–24.