Analysis of Scheduling Algorithms base on Fairness Index, Packet Loss Rate & Delay in 4G Wireless Network

Transcription

1 Analysis of Scheduling Algorith base on Fairness Index, Packet Loss Rate & in 4G Wireless Network Zalmai Zormatai 1, Sunera Kargathara 2 1 M.E student, Marwadi Education Foundation, Rajkot city, Gujarat, India 2 Assistant Professor, Marwadi Education Foundation, Rajkot city, Gujarat, India 1 zalmai.zormatai13456@marwadieducation.edu.in 2 sunera.kargathara@marwadieducation.edu.in Abstract Packet loss rate and end to end delay are parameters due to which end users can t achieve better data services especially in wireless mobile communication system. Packet loss rate occurs due to losses, interference in wireless media and end to end delay is related to the user distance from NodeB/eNodB which also causes loss of data. In this paper, we prefer to analyze fairness index, packet loss rate and end to end delay using different scheduling algorith like Round Robin, Proportional Fair, Best CQI and MLWDF algorith in 4G wireless network and try to find out better algorithm which can able to schedule users with little packet loss rate and little delay. Keywords: Wireless Mobile Communication System, Packet Scheduling, Scheduling Algorith, Fairness Index, Packet Loss Rate, 1. Introduction Wireless mobile communication system are utilizing frequently by users due to its higher mobility. But, the QoS of it is causing due to wireless media. Because it can able to provide better mobility but can t able to provide properly better services especially data services according to user requirements. When utilizing wireless mobile communication system, users suffer with high packet loss rate. Packet loss rate is directly related to losses, interference which is observed much in wireless media. High packet loss rate and high end to end delay can degrade especially QoS of data services. There has been performed much research to improve QoS of wireless mobile communication system and in result we are looking 4th generation of wireless mobile communication system and still researches are continued related to 5th generation of wireless mobile communication system. These all researches are based on, to improve QoS in wireless mobile communication system. 1G and 2G can t able to provide data services. 2.5G was the first step towards data services which was not also able to provide better data services according to user requirements. 3G and 3.5G was better than previous generations which was introduced by 3GPP (3rd Generation Partnership Project) and can able to provide data services up to 2Mbps. But due to increased amount of user data, 3G and 3.5G could not afford properly. So, according to user requirements 3GPP introduced 4th generation of wireless communication system which works base on LTE (Long Term Evolution) and LTE-A (Long Term Evolution - Advanced) standards. 4G wireless mobile communication 142

2 system can able to provide 75Mbps downlink data rate. It is used OFDMA access technology for downlink and SC-FDMA access technology for uplink. It is totally IP based network, its performance is much better than 3G. t supports o service ideo streamin service service v and web browsin. t provides hi h mobility than previous enerations. ts spectral efficiency is much better than previous enerations [1]. 4G technology perfor scheduling mechanism for resource allocation base on OFDMA. Scheduling mechanism is performed by scheduling algorith. Different types of scheduling algorith are available which have specific characteristic. Some of the algorith can able to schedule users base on low packet loss rate and little end to end delay. 2. Packet Scheduling in LTE Packet scheduling is a type of mechanism which is responsible to allocate resource to a user in a given time interval. It distributes resource in efficient way which results better throu hput low pac et loss rate and fairness [2]. The task of scheduling is performed by packet scheduler, which is installed in NodeB/eNodeB. Packet scheduler receives information about user channel quality condition which is known as CQI (Channel Quality Information) in every TTI. CQI is selected base on SNR. [3] Below figure represent packet scheduling in LTE. i.1. ac et schedulin in [10] When scheduler receives CQI report of each user then AMC (Adaptive Modulation and Coding) selects modulation and coding scheme for transmission of that users. Packet scheduler uses PDCCH (Physical Downlink Control Channel) to inform users about assigned resource block and selected modulation and coding scheme. As scheduler uses scheduling algorith for scheduling purpose. So, some of the scheduling algorith are defined as below. 3. Packet Scheduling Algorith Packet scheduling algorith are used for scheduling. Each packet scheduling algorithm has its own characteristic. Here we considered Round Robin, Proportional Fair, Best CQI and MLWDF algorith as below: 3.1. Round Robin (RR) Algorithm 143

3 DOI: /ijss/ t is a basic al orithm desi ned especially for fairness purpose. t doesn t consider other parameters like delay, packet loss rate and throughput. So, it is better for fairness than other parameters [6] Proportional Fair (PF) Algorithm It is a scheduling algorithm which considers better channel quality condition, fairness and throu hput of users durin pac et schedulin. t doesn t consider delay and pac et loss rate. t can able to provide better user throu hput and fairness [7] Best CQI Algorithm It is one of the scheduling algorith which considers channel quality, i.e.: if the quality of channel is best It is a scheduling algorithm which schedules end users base on best channel uality condition. f channel uality of users will best then users will be scheduled otherwise users will not be scheduled [6] Modified Largest Weighted First (MLWDF) Algorithm 4. Simulation Work This section explains our simulation work. We used MATLAB as a simulation tools. Our simulation work is based on analysis of fairness index, packet loss rate and end to end delay using Round Robin, Proportional Fair, Best CQI and MLWDF scheduling algorith. After analysis we will conclude that which scheduling algorithm is better to provide low packet loss rate and little end to end delay. Simulation parameter is given as below: Table1. Simulation parameter Parameter Value Scheduling TTI 1 Number of Users Carrier Frequency 2GHz` System Bandwidth 20MHz Number of Resource Block 100 Modulation QPSK, 16QAM, 64QAM Path loss Model Okumura Hata Model Area Urban Area Traffic Model VoIP VoIP standard G.711 Scheduling Algorithm Round Robin, Proportional Fair, Best CQI, MLWDF Threshold 0.06sec User distance from NodeB/eNodeB Random As we discussed above that scheduler is responsible to assign resource to each user base on their channel quality condition. First we try to obtain information about channel quality condition of each user using Okumura Hata Model in Urban area. Channel quality condition of each user can be found base on user distance from NodeB/eNodeB, Path loss and SNR. After it, resource block will assign to each user base on their channel quality condition. Number of user in this simulation is 50, 75, 100 and 125 and their distance from NodeB/eNodeB is considered randomly. 144

4 After finding channel quality condition of each user we analyze fairness index, packet loss rate and delay of users using different scheduling algorith as below: 4.1 Analysis of Fairness Index Before starting analysis of Fairness Index, First we should introduce Fairness Index. Fairness Index is one of the parameter which shows the capability of scheduling algorithm. i.e.: it shows that how many users can be scheduled by each scheduling algorithm. First we analyze fairness index of Round Robin algorithm using 50, 75, users to know that how many user can be scheduled by Round Robin algorithm. After analysis it results that if scheduler will use Round Robin scheduling algorithm then all users will be scheduled without channel quality consideration. Round Robin algorithm is better if we want to assign resource to each user. If the channel quality condition of any user will better then throughput will better. But, if the channel quality condition of any user will not better then throughput will not better and user will achieve high packet loss rate. If we want to assign resource to each user then we should use this algorithm. But, if we want throughput, low packet loss rate and low delay then this algorithm is not suited and we should find better algorith instead of it. After Round Robin algorithm, we try to find fairness index of Proportional Fair algorithm for same amount of users. Proportional Fair algorithm schedules end users base on better channel quality condition. If channel quality condition will better it will assign resource else it will not assign resource to the users. When we apply this algorithm, we obtain fairness Index less than fairness index of Round Robin al orithm. Because this al orithm doesn t has a capability to assi n resource to all users. Some of the users remain from resource allocation due to weak channel quality condition. Just those users obtain resources which have better channel quality condition. This algorithm provides better result if we use it to achieve better throughput and low packet loss rate. After Proportional Fair algorithm, we try to find fairness index of Best CQI scheduling algorithm for same amount of users. We obtain lowest fairness index by using this algorithm and in result least of the users schedule by this algorithm. Best CQI algorithm schedules end users base on best channel quality condition. If the channel quality condition will best it will assign resource else it will not assign resource to end users. This algorithm is better if we require high throughput and low packet loss rate. After Best CQI algorithm, we apply MLWDF algorithm and find out fairness index of it for same amount of users. By using this algorithm we result better fairness index relative to Best CQI algorithm. But, achieve lower fairness index relative to Round Robin and Proportional Fair. MLWDF algorithm schedules users base on least or threshold delay. If the end to end delay of user is equivalent to threshold delay then it will schedule users else it will suspend and assign transmit power base on users channel quality condition. After analysis of fairness index, the result of each algorithm is represented in below plot graph, which also shows some variations with increasing the number of users. 145

5 Fairness Index DOI: /ijss/ Figure.2. Analysis of Fairness Index This plot graph provides clear information about fairness index of Round Robin, Proportional Fair, Best CQI and MLWDF algorith and shows that fairness index of Round Robin algorithm is equal to 1 and highest than other algorith. But in other end plot shows that the fairness index of Best CQI is least than other scheduling algorith and equal to 0.2. When we look the middle portion of plot, it shows that the two remaining algorith have medium fairness index with little variation. The difference is that the fairness index of Proportional fair is higher than MLWDF algorithm, equal to 0.6 and increase with increment the number of users. But, fairness index of MLWDF algorithm is lower than Proportional Fair, equal to 0.5 and decrease with increment the number of users. We can conclude fairness index of each scheduling algorithm in tabular form which is as below: Table 2 Analysis of Fairness Index Algorith Number of Users Round Robin % Proportional Fair % Best CQI % MLWDF % The result of fairness index in this table explains specifically that how many users have scheduled by each scheduling algorith. So, from this table we result that Round Robin scheduling algorithm schedule all users. Second, when we look to Proportional Fair algorithm in this table, it is shown that 61% of users are scheduled by Proportional Fair algorithm. Third, when we look to Best CQI algorithm in this table, it is shown that 20% of users are scheduled by Best CQI algorithm. Fourth, when we look to MLWDF algorithm in a table, it is shown that 48% of users are scheduled by MLWDF algorithm. 4.2 Analysis of Packet Loss Rate 146

6 First we should introduce packet loss rate. Packet Loss Rate is the ratio of lost data and total transmitted data. It is denoted by PLR and its unit is percentage (%). If the value of packet loss rate is 1, its meant 100% of data has lost. Packet Loss Rate (PLR) should less than 1 otherwise it shows worst QoS. We are finding packet loss rate in LTE/LTE-A networks. Equation of packet loss rate is represented as below: PLR = 1-(1-BER) ^dtransmitted [4] (1) In equation (1) PLR is packet loss rate, BER is Bit Error Rate, dtransmitted is transmitted data. As we see that PLR rely on BER. So, BER is calculated base on complementary error function (erfc) and SNR or CQI. The equation use for BER is as: BER = (1/2)*erfc ( R [4] (2) In equation (2) erfc is complementary error function, SNR is Signal to Noise Ration. SNR can be replaced to CQI, and in this case BER can be calculated as: BER = (1/2)*erfc ( [4] (3) Now we can find PLR of each algorithm for different number of users. First we find PLR of users (50, 75, 100 and 125) using Round Robin algorithm. As we know that Round Robin al orithm doesn t consider channel quality condition. So this algorithm calculates PLR of each user and provides average PLR of end users base on the number of users. If the numbers of users are 50, it will calculate PLR of 50 users and if the numbers of users are 75, it will calculate PLR of 75 users and so on. So we can say that average PLR of this algorithm is high. Second we consider Proportional Fair algorithm and calculate PLR of end users base on this algorithm. This algorithm schedules end users base on better channel quality condition. Channel quality condition is inversely proportional to packet loss rate. If channel quality condition is better, PLR will be low else PLR will be high. In general PLR of each user using Proportional Fair algorithm is lower. Third we consider Best CQI algorithm and try to calculate packet loss rate of each user using this algorithm. It also provides least packet loss rate because it schedules users base on best channel condition. Fourth, we consider MLWDF algorithm, and calculate packet loss rate of each user. This algorithm schedules end users base on threshold delay. End to end delay is related to distance from NodeB/eNodeB. If distance of user from NodeB/eNodeB is less then delay will be less, and SNR will be high. By using this algorithm we noted that PLR of users is less relative to Round Robin and Proportional Fair algorithm and high relative to Best CQI algorithm. The result of user PLR using mentioned scheduling algorithm is given in below plot graph. 147

7 1.47% 1.48% 1.45% 1.36% Average Packet Loss Rate % % % % 7.34% 7.93% 7.49% 6.85% 44% % % 44% Algorith DOI: /ijss/ Figure.3. Analysis of Packet Loss Rate If we look Round Robin algorithm, it shows that the average PLR of users is greater than using other scheduling algorith. Now we look Proportional Fair algorithm. It shows user PLR is less than Round Robin, but greater than MLWDF and Best CQI algorithm. Best CQI algorithm shows least user PLR and MLWDF algorithm shows lower PLR than Round Robin and Proportional Fair algorithm. We can represent this analysis in tabular form, if we want to know specific value of users PLR using mentioned scheduling algorithm. Table 3 Analysis of Packet Loss Rate Number of Users Round Robin Proportional Fair Best CQI MLWDF 148

8 This table provides information that if there will be a 50 users and scheduler will use Round Robin algorithm 44% packet loss will be observed by each users. If the numbers of user increase to 75, 100 and 125 packet loss rate will be 44.32%, 43.80% and 44%. So, it is concluded that by using Round Robin algorithm users will observe high packet loss rate. When analyze packet loss rate using Proportional Fair algorithm. It results less packet loss rate than Round Robin algorithm. If the numbers of users will 50, Packet loss rate will be 7.3%. If we increase the number of user to 75, 100 and 125, Packet loss rate will be 7.93%, 7.49% and 6.85%. When analyze packet loss rate using Best CQI algorithm, it results least packet loss rate. If the numbers of users will 50, packet loss rate will be %, and when number of users are increasing to 75, 100 and 125, packet loss rate will be %, % and %. When analyze packet loss rate using MLWDF algorithm. it results greater packet loss rate than Best CQI algorithm, but less packet loss rate than Round Robin and Proportional Fair algorithm. If the numbers of users will 50, packet loss rate will be 1.47%, if the number of users will increase to 75, 100 and 125, the average packet loss rate will be 1.48%, 1.45% and 1.36%. 4.3 Analysis of First we introduce delay. is the time taken by a packet to travel across a network from the source to the destination. End to end delay includes Transmission delay, Propagation delay, Processin delay and ueuin delay. o we should now all of these delays. nd to end delay can be calculated as below [5]. Total = D Tran + D Prop + D Proc + D Que [5] (4) In equation (4) D Tran is Transmission delay, D Prop is Propagation delay, D Proc is Processing delay and D Que is Queuing delay. Transmission : Transmission is also called encoding delay. Transmission delay is the ratio of data sent (bits sent) and speed of link. It is calculated as: D Tran = Data sent (bits) /Link speed [9] (5) Link speed of each media is different e.g.: link speed of fast Ethernet channel is 100Mbps. So in our case we consider link of LTE and the standard downlink speed of LTE is 75 Mbps. Propagation : Propagation is the ratio of length of link and speed of light. So, it is calculated as: D Prop en th of in peed of li ht [9] (6) In this type of delay we consider length of link as distance of user from NodeB/eNodeB, and the speed of light is 3*10^8m/s. Processing Processing is a type delay which is a time taken to process a data in a system for transmission, It is generally negligible to calculate because it is much less time. So, in general it is counted as 1 microsecond. Queuing 149

9 DOI: /ijss/ Queuing is a type of delay which shows that the time is taken for a data to wait in queue for transmission. So, it is the product of transmission delay and queue length. Queuing is calculated as: D Queu = D Trans x L Queu [9] (7) Average Length of Queue is considered less than ½. So, we consider it as 1/3. L Queue. [9] (8) Now we start analysis of delay for mentioned scheduling algorithm base on end to end delay equation. To analyze delay we should know user distance from NodeB/eNodeB and link speed of network, which is most important for calculating delay. The unit of delay we consider here is millisecond () and second (sec). First, we analyze delay of each user from 50 to 125 using Round Robin algorithm, this algorithm provides us higher delay compare to threshold delay. In this paper threshold delay is considered 0.06 seconds which is equal to 66. The higher delay in Round Robin algorithm is due to its characteristics. This algorithm schedules all users and it doesn t consider threshold delay. o there are many users who have delay higher than threshold delay and in result delay of users using this algorithm is higher. Second, we analyze delay of users from 50 to 125 using Proportional Fair algorithm, this algorithm shows end to end delay of each user near to threshold delay. So, it is better algorithm to be used for delay purpose. Third, we analyze delay of users from 50 to 125 users using Best CQI algorithm, it shows that end to end delay is less for users using Best CQI algorithm. Fourth, we analyze end to end delay of each user from 50 to 125 using MLWDF algorithm, it provides end to end delay equal to threshold delay. This algorithm has one characteristic that if the delay of user greater than threshold value, it tries to assign transmission power base on user CQI. Result of user delay using mentioned scheduling algorith are represented in given plot graph: Figure.4. Analysis of It shows that while using Round Robin algorithm for delay purpose, user will observe larger delay and the delay will be higher than threshold delay. 150

10 If we look to Proportional Fair algorithm, It shows that user delay is near to Threshold delay and little variation also observes in user delay by increasing number of users. When we look to MLWDF algorithm, it shows that user delay is near to threshold delay and variation of user delay is little compare to Proportional Fair and Round Robin algorithm. In the last if we look to Best CQI algorithm, it shows least delay compare to all other delay, and no any variation is occurred by increasing the number of users. But if we want to know specific value of user delay then we can represent it in tabular form as below: Table 4 Analysis of Algorith Round Robin Proportional Fair Best CQI MLWDF User Total Average Total Average Total Average Total Average Number of Users By this table we know that how much delay is observed by each user using mentioned schedfuling algorith. In this table we look two types of delay which is total delay and average delay. Total delay infor us about the waiting time of each user for resource allocation, and average delay is a time which is taken to transmit and receive data of each users. In this table, it shows that the average delay of Best CQI is not varied by increasing the number of users and delay is fixed, and in other side average delay of MLWDF, Proportional Fair, and Round Robin algorith are varied by increasing the number of users. If we check the total delay of users using Round Robin algorithm, it shows that if the number of user will 50, each user will be waiting for 5sec, if the number of user will 75 each user will be waiting for 7.7 sec. If the number of user will 100, each user will be waiting for 10 sec and if the number of user will 125 each user will be for waiting 12.8 sec for resource allocation. And the average delay of users using Round Robin algorithm is 103, it means that the end to end delay of each user using Round Robin algorithm will be 0.1 sec. 151

11 DOI: /ijss/ If we check the total delay of users using Proportional Fair algorithm, it shows that if the number of user will 50, each user will be waiting for 1.8 sec. If the number of user will 75, each user will be waiting for 2.8 sec. If the number of user will 100, each user will be waiting for 3.8 sec and if the number of user will 125 each user will be waiting for 4.8 sec for resource allocation. And the average delay of users using Proportional Fair algorithm is 61, it means that the end to end delay of each user using Proportional Fair algorithm will be sec. If we check the total delay of users using Best CQI algorithm, it shows that if the number of user will 50, each user will be waiting for 0.33 sec. If the number of user will 75, each user will be waiting for 0.5 sec. If the number of user will be 100, each user will be waiting for 0.66 sec and if the number of user will be 125 each user will be waiting for 0.83 sec for resource allocation. And the average delay of users using Best CQI algorithm is 33, it means that the end to end delay of each user using Best CQI algorithm will be sec. If we check the total delay of users using MLWDF algorithm, it shows that if the number of user will 50, each user will be waiting for 1.3 sec. If the number of user will 75, each user will be waiting for 1.9 sec. If the number of user will 100, each user will be waiting for 2.6 sec and if the number of user will 125 each user will be waiting for 3.2 sec for resource allocation. And the average user delay of users using MLWDF algorithm is 53, it means that the end to end delay of each user using MLWDF algorithm will be sec. 5. Conclusion and Recommendations We performed simulation to find scheduling algorithm which will provide low packet loss and low delay. We analyzed packet loss rate and delay of Round Robin, Proportional Fair, Best CQI and MLWDF algorithm. All of these algorith provided different packet loss rate and delay. Round Robin algorithm provided highest packet loss rate and delay, Best CQI algorithm provided least packet loss rate and delay. Proportional Fair and MLWDF algorith provided lower packet loss rate and delay. So, it is concluded that Proportional Fair algorithm is a better algorithm than other algorith which has capability to provide better fairness index, low packet loss rate and low delay in 4G wireless networks. References [1] F. Capozzi, G. Piro, Student Member, IEEE, L.A.Grieco, Member, IEEE, G. Boggia, Senior Member, and. amarda. Downlin ac et chedulin in ellular etwor s: Key Desi n ssues and a Survey, IEEE Communication Surveys & Tutorials, VOL. 15, NO. 2, Second Quarter [2] Ammar Hafeez Mohammad Masoom Zafar ir Ahmad Mudasir A urvey on Analysis and Design of Scheduling Algorithm for LTE-Advanced Hetero eneous etwor s Journal of etwor in Technology, [3] mam Osman erformance valuation of Downlin chedulin Al orith in on erm valuation ( etwor repository.sustech.edu, [4] harles Jumaa Katila hiara Buratti Melchiorre Danilo Abri nani and Roberto erdone ei hbors- Aware roportional air schedulin for future wireless networ s with mixed MA protocols EURASIP Journal on Wireless Communications and Networking, Springer, [5] He ma haari Kais Mnif otfi Kamoun Multimedia uality transmission evaluation over wireless networ s: a survey prin er [6] MC Thayammal, MM Linda, «A Comprehensive Study on Efficient Resource Allocation by QoS in Wireless etwor s A Journal on ommunication

12 [7] Ms. allavi Umesh ai Ms. hahad hir eri Dr.G.R.Udupi rof. lasin rancis Dias Desi n and Development of Medium Access ontrol cheduler in e odeb nternational Journal of Computational Engineering Research, [8] U.. wawelu1.. Ani2 and M. A. Ahane u3 omparative Analysis of the performance of Resource Allocation Al orith in on erm volution etwor s i erian Journal of echnolo y (NIJOTECH), [9] Network delays and losses, University of Minnesota Duluth.edu. [10] Farhana Afroz1, Roshanak Heidery2, Maruf Shehab3, Kumbesan Sandrasegaran and Sharmin Sultana hompa omparative Analysis of Downlin pac et schedulin al orith in 3G networ s International Journal of Wireless & Mobile Networks (IJWMN) Vol. 7, No. 5, October