Selective Periodic Compoet Carrier Assigmet Techique i LTE ad LTE-A Systems Husu S. Narma ad Mohammed Atiquzzama School of Computer Sciece, Uiversity of Oklahoma, Norma, OK 73019 Email: {husu, atiq}@ou.edu Abstract Iteret usage over mobile devices is o the rise. The badwidth demad for mobile Iteret access is also icreasig with the umber of mobile users. To aswer users demad, carrier aggregatio is proposed i LTE-A system. I carrier aggregatio, the best available oe or more compoet carriers of each bad are assiged to each user to provide efficiet services. Several works have bee reported i the literature o madatory ad periodic compoet carrier assigmet methods. Although the previous works, especially periodic compoet carrier assigmet methods, have sigificatly improved performace of LTE ad LTE-A systems, may limitatios still exist. Oe limitatio of previous works is that data trasfer is iterrupted durig periodic compoet carrier assigmet operatio which ca decrease performace of the system. Therefore, i this paper, selective periodic compoet carrier assigmet techique which allows cotiues data trasfer durig periodic carrier assigmet operatios is proposed. Results show that the proposed techique icreases throughput rate up to 25% ad decreases average delay time up to 35%. Idex Terms LTE, LTE-A, 4G, compoet carrier assigmet, resources allocatio, aalysis. I. INTRODUCTION I LTE-A multi-bad architecture, each bad has several Compoet Carriers (CCs). User Equipmet (UE) ca simultaeously coect oe or multiple carriers from differet bads. Base statios (EodeB) arrage the umber of simultaeous coectios of from each bad. Iefficiet Compoet Carrier Assigmet (CCA) sigificatly decreases system performace [1], [2]. There are several proposed compoet carrier assigmet methods ad their aalysis [2] [9] for LTE ad LTE-A systems. I [3] [6], full or partial feedback for Chael Quality Idicator (CQI) is used i order to fid the best available carriers for each user. I [2], distributio of carriers to users are balaced. I [7], [8], service-based carrier assigmet methods are proposed by givig priority for some services while assigig carriers to users. I [9], a uplik carrier assigmet method has bee proposed by cosiderig a ratio fuctio, traffic type ad CQI to icrease throughput while sedig data from users to a enodeb (enb). While uplik carrier assigmet methods try to optimize badwidth ad power usage, dowlik carrier assigmet methods try to optimize oly badwidth usage. I additio to the above carrier assigmet methods, there exist traditioal assigmet methods, Least Load (LL) (LL ca be called as Roud Robi (RR)) ad Radom (R) methods [10]. LL method assigs users to least loaded carriers thus, LL method well balaces users loads across carriers i short ad log terms, ad R method radomly selects carriers for users hece, R method oly well balaces users loads across carriers i log term. However, both of them igore Quality of Service (QoS) requiremets of each UE ad CQI of chaels. All of works [2] [10] ca be grouped uder Madatory Compoet Carrier Assigmet (mcca) methods which oly update carriers based o madatory chages (iclude path lost, CQI chages, etc). I [11], a Periodic Compoet Carrier Assigmet (pcca) method is proposed ad carriers are periodically assiged to each UE i specified time iterval. Mi-delay, higher CQI, etc. algorithms ca be used i periodic carrier assigmet methods i order to optimize delay or throughput of systems. For example, periodic carrier assigmet method i [11] is a kid of mi-delay-based method, which tries to miimize delay which is experieced by users. I periodic carrier assigmet, CCs of all users are updated periodically i additio to madatory CCs assigmet. As preseted i [11], periodic carrier assigmet sigificatly improves the performace of LTE ad LTE-A systems. However, oe kow limitatio of such system is iterruptio of data trasfer durig periodic carrier assigmet process. This is due to reassigmet all carriers of users at the same time i periodic carrier assigmet [11]. Such techique ca be called as Joit Periodic Compoet Carrier Assigmet Techique (j-pcca). The performace of periodic carrier assigmet ca be improved more because joit techique causes frequet packet trasfer iterruptios ad that results delay ad packet drops durig periodic carrier assigmet operatios. Therefore, the aim of this work is to improve the performace of periodic compoet carrier assigmet methods by elimiatig frequet packet trasfer iterruptios. The objective of this paper is to cosider packet drops ad delay which are experieced by users durig periodic carrier assigmet process ad propose selective periodic carrier assigmet techique (s-pcca) to icrease the performace of periodic carrier assigmet methods for LTE ad LTE-A systems. The key cotributios of this work are as follows: (i) proposig selective periodic carrier assigmet techique, (ii) the system models for joit ad selective techiques are explaied by usig queuig system, (iii) comparig joit ad selective techiques by applyig both techiques to two well-kow carrier assigmet methods, modified LL ad R (modified based o CQI ad periodic assigmet) with extesive simulatio. R ad LL methods are selected because
CCs Assiger Packets Scheduler of simplicity ad commo usage of them i the literature. Results show that the proposed techique icreases throughput rate up to 25% ad decreases average delay time up to 35% comparig to joit techique i the system. Our proposed techique ad related aalysis will help service providers build efficiet periodic compoet carrier assigmet methods i order to icrease throughput ad decrease average delay time. The rest of the paper is orgaized as follows: I Sectio II, the system model of carrier assigmet procedure for joit ad selective techiques are explaied ad followed by queuig aalysis of both techiques i Sectio III. Simulatio eviromets with parameters are described i Sectio IV. I Sectio V, simulatio results are preseted ad aalyzed. Fially, Sectio VI has the cocludig remarks. II. SYSTEM MODEL WITH JOINT AND SELECTIVE TECHNIQUES Fig. 1 demostrates a simple model of carrier assigmet methods ad packet schedulers. There are umber of users ad each user ca oly coect up to m umber of CCs. Today, LTE-A system ca oly support up to five simultaeous CCs coectio for each user providig IMT-A level service [12]. Oe to two of CCs are primary compoet carriers for uplik ad dowlik, ad ca oly be updated durig hadover [12], ad the rest of carriers is Secodary Compoet Carriers (SCCs) which are updated for each user based o CQI of chaels. However, as stated i [11], periodic carrier assigmet is a ew method try to reassig all CCs periodically i additio to madatory carrier assigmets. Therefore, both PCC ad SCC are updated durig the periodic carrier assigmet operatios for all users [11]. After carrier assigmet process fiishes, UE 1 UE 2 UE CC 1 CC 2 CC 3 CC m Fig. 1. Geeral System Model with users ad m available CCs. Packed Scheduler trasfers packets over selected carriers i time ad frequecy domais. Curretly Proportioal Fairess ad max-mi are commo packet schedulers i LTE systems [2], [11]. A. Joit Periodic Compoet Carrier Assigmet (j-pcca) Madatory carrier assigmet methods allocate users to carriers based o mobility of users (icludig path ad coectio lost, low CQI, etc.). Therefore, whe UE i moves from oe positio to aother positio, uplik ad dowlik carriers are updated to maitai coectio. O the other had, periodic carrier assigmet methods allocate users to carriers based o time ad periodically updates carriers i specified time itervals [11] regardless of users mobility. Durig joit periodic carrier assigmet process, all carriers are simultaeously updated for all users; packet trasfer of users is thus iterrupted. After joit periodic carrier assigmet process is completed, packet trasfer is recommeced. B. Proposed Selective Periodic Compoet Carrier Assigmet (s-pcca) As explaied i Sectio II-A, disadvatage of joit techique is simultaeous reassigmet of all carriers to users resultig i iterruptio of packet trasfer. I order to provide better service, we have proposed a ovel method, selective periodic carrier assigmet, to solve the disadvatage of joit techique. I selective techique, oly selected carriers of users are periodically updated. However, it is possible to update all carriers durig selective periodic carrier assigmet process accordig to selectio algorithm. Selective techique is takig ito accout time ad CQI durig periodic carrier assigmet process i additio to Compoet Carrier Assigmet Method. For example, LL method with selective techique is processed as follows for each periodic time: The threshold of CQI is predetermied for selectio Algorithm (5 is selected as predetermied threshold for s- pcca). Here, the threshold ca be dyamically arraged by usig user profile iformatio for each user as is doe i our past work [13]. Partially or fully CQI feedback is obtaied to measure the quality of CCs for each user. Note that, although CQI is low, the chael ca trasfer oly a limited umber of packets. The CCs, which have lower CQI tha the predetermied threshold, are reseted ad selected to be updated for each user. selective techique first selects ew CCs, which have the least umber of active users ad have higher CQI tha the threshold. It is very importat to ote that the umber of ew CCs may ot be equal to previous umber of CCs for each user. To make equal, more CCs, which have the least umber of active users, ca be assiged. For example; assume that UE i receives data by usig C 1, C 2, ad C 3 compoet carriers ad the CQI of C 1 ad C 2 are lower tha the threshold. Therefore, selective techique chooses C 1 ad C 2 to update. However, selective techique oly fids CQI of C 4 is higher tha the threshold from all available CCs for UE i. Therefore, LL method with selective techique assigs C 4 ad the CC, which has the least umber of active users ad the highest CQI (accordig to UE i ), to UE i. To icrease the efficiecy ad QoS, packet trasferrig priority is give to the CC, which is the closest to the enb. Similarly, R method with selective techique is processed as above except that radomly assigig CCs from the CCs which have higher CQI tha the threshold. III. ANALYSIS I this sectio, aalytic expressios will be derived for performace metrics of joit ad selective techiques durig 2
periodic carrier assigmet by usig queuig system. A. Notatios The otatios used for the aalysis i the rest of the paper are listed i Table I. TABLE I NOTATIONS i P t1, 2,..., u j P t1, 2,..., mu Q i Queue of UE i N Size of each queue p k Probability of k umber of packets i the system µ j Service rate of CC j λ i Packet arrival rate of UE i δ Avg. delay durig periodic carrier assg. Avg. queue legth durig periodic carrier assg. D Drop probability durig periodic carrier assg. B. Queuig Models of j-pcca ad s-pcca for Dowlik Fig. 2 illustrates dowlik process for users i LTE-A. The queuig model scheduler is Joit Queue Scheduler. We have used Joit Queue Scheduler because the performace of Joit Queue Scheduler is better tha Disjoit Queue Scheduler for LTE-A [14]. While Joit Queue Scheduler allows all users to have disjoit buffers as showed i Fig. 2, Disjoit Queue Scheduler allows all CCs to have disjoit buffers. Q 1 λ1 μ 1 λ2 Q 2 μ 2 UE 1 UE 2 Q λ μ m UE Fig. 2. Dowlik Joit Queue Model with users ad m available CCs. Dowlik packet arrival rate for UE i is λ i, each CC represeted by a server ad service rates of CCs are µ j where j P t1, 2,..., mu ad each buffer, Q j, ca hold at most N packets. Packet schedulers equeue a arrived packet which is requested by a user to oe of assiged CCs. Durig joit periodic carrier assigmet operatio, packet trasfer of UE i is termiated all the time. However, packet trasfer of UE i is termiated if all carriers eed to be updated or PCC eeds to be updated durig selective periodic carrier assigmet operatio (If PCC is updated the all carriers eed to be updated). Therefore, there are two cases i the system for joit ad selective techiques: Case 1: PCC is required to be updated, therefore SCCs are required to be updated (This case also icludes whe all carriers are required to be updated). Case 2: SCCs are required to be updated but PCC is ot required to be updated. The performace metrics of joit ad selective techiques are same for Case 1. Hece, oly Case 2 is explaied to distiguish differeces betwee joit ad selective techiques. Durig periodic assigmet operatio (Case 2) i joit techique for UE i, the packet trasfer operatio is as follows: (i) Packet trasfer is iterrupted for the user. (ii) All servers of the user are updated. (iii) Packet trasfer is recommeced for the user over ew carriers. O the other had, durig periodic carrier assigmet operatios (Case 2) i selective techique, the process is as follows: (i) For all users, some carriers (CCs) are selected to be updated accordig to the selective algorithm (here, it is based o chael quality idicator) (ii) Packet trasfer is oly iterrupted o carriers which are eeded to be updated for each user. (iii) New carriers are assiged to users. (iv) Packet trasfer is commeced o ew carriers for the users. C. Assumptios To make the model aalytically tractable, it is assumed that there is oly oe UE i the system as demostrated i Fig. 3, all servers are capable of servig all type of packets, the queuig system is uder heavy traffic flows, packet arrivals follow Poisso distributio, ad service times for packets are expoetially distributed. Type of queue disciplie used i λ i Q i Fig. 3. Dowlik Joit Queue Model with oe user ad m available CCs. the aalysis is FIFO. Badwidth ad CQI of carriers ca be differet, so ca service rates of all servers. D. Performace Metrics I this sectio, we will approximately derive drop probability, average queue legth ad average delay durig periodic carrier assigmet process for Case 2 i joit ad selective techiques. I joit periodic carrier assigmets, all carriers are updated for UE i. Therefore, the service rate is zero ad the system is ot i steady state. Hece, we oly metio the possibilities o performace of joit techique. O the other had, we approximately derive performace metrics of selective techique. 1) State Probability: The service rate of the system is state depedet. Whe oe packet is i the system, the service rate is µ t1 µ 1 ad whe two packets are i the system, the service rate is µ t2 µ 1 µ 2. Service rate of the system icreases util all carriers are utilized (c carriers for UE i ). The the total server rate of the system is fixed at µ tc. It is importat to ote that there is at least oe carrier which serves icomig traffic i selective techique which meas 1 c m. By usig above approach, the state trasactio diagram for selective techiques ca be obtaied as i Fig. 4. By usig state trasactio diagram, state probabilities ca be obtaied as follow: p k $ '& '% p 0 λ k i μ 1 μ 2 μ m k± k c µ p c tc ρk 0 UE i c k c N (1) 3
λ i λ i λ i λ i λ i p 0 p 1 p 2 p c p c+1 p c+n μ t1 μ t2 μ tc μ tc μ tc Fig. 4. State trasactio diagram for the model. where ρ λ i {µ tc. Usig c N p k 1, we get p 1 0 $ '& '% 1 1 c c k± λ k i k0 µ c tc k1 kc 1 λ k i k± N µ c tc k1 c N ρ k ρ 1 ρ 1 where ρ λ i {µ tc. 2) Drop Probability: Drop probability of the model is the fial state probability which is p c N. Therefore, drop rate for selective techique ca be obtaied as follow: D p 0 µ c tcρ c N (2) (3) O the other had, drop probability for joit techique ca ot be obtaied because of usteady state. It may be obtaied by usig lim ρñ8 D or lim µñ0 D. Therefore, D 1 3) Average Queue Legth ad Average Delay: Average queue legth ad average delay ca be formulated by usig state probability. Average queue legth, for M{M{1{N queue is as follows: Ņ k1 kp k (4) However, M{M i {c{n the system has c servers ad from the above state probabilities (Eq. (1)), will be; c N kc 1 which gives the followig expressios for : $ '& '% p 0 µ c tc ρ c 1 1 pn 1qρ µ p c tc 0 NpN 1q 2 pk cqp k (5) N Nρ N 1 p1 ρq 2 ρ 1 ρ 1 Usig Little s law ad Eqs. (3) ad (6), average delay ca be obtaied as follows: δ (7) 1 D O the other had, for joit techique, average queue legth () will be N. Therefore, average delay (δ) will be δ 8. However, because periodic carrier assigmet time duratio is limited to a umber (assume τ), the δ τ (6) IV. SIMULATION OF THE SYSTEM I Sectio III, we approximately derive aalytic performace metrics for selective techique ad possible performace values for joit techique. We show that selective has improved performace of the system durig periodic carrier assigmet operatios. However, overall system performace metrics ca be differet because service rate of carriers for each user are time ad positio depedet. Therefore, we have implemeted simulatio to observe the overall system performaces of joit ad selective techiques ad show results i Sectio V. A. Assumptios for enbs It is assumed that there is oly oe enb which has three bads to provide service to users. The parameters of enb is give i Table II. TABLE II enb PARAMETERS Num. of enb 1 Used Bads 800MHz, 1.8GHz, 2.6GHz Num. of CCs i Each Bad 4 Total Num. of CCs 12 Queue Legth of Each Queue 50 packets Badwidth of CCs 10MHz Modulatios QPSK, 16QAM, ad 64QAM CQI 3, 5, 7, ad 11 Trasmissio Time Iterval 10ms Time for CCA 20ms CQI Threshold 5 B. Assumptios for There are two types of equipmet, LTE ad LTE-A types i the system. Half umber of equipmet is LTE type ad ca oly use oe carrier ad the other half umber of equipmet are LTE-A type ad ca use multiple carriers (up to five). I simulatio, four CCs ca be simultaeously used by LTE- A type equipmet because maximum five CCs ca be used by LTE-A type equipmet, ad oe of them must be used for upload primary compoet carriers (see Sectio II). Users are iitially o-uiformly distributed i area which meas that mostly users are located close to enb. 50% of users ca move aroud of the enb i specified time iterval. Each user ca oly dowload oe type of traffic. Packet arrivals follow Poisso distributio with 250 (packets per secod) arrival rate for each user. Ad total arrival rates of traffic are elarged whe the umber of users is icreased. Selected Trasmissio Time Iterval (TTI) for a packet is 10ms. C. Carrier Assigmet Process ad Packet Schedulig I simulatio, modified LL ad R methods are used for periodic carrier assigmet by selectig carriers which have higher CQI tha the threshold i additio to policies of LL ad R methods. Therefore, durig the simulatio, LL method assigs carriers, which have the least umber of users ad have higher CQI tha the threshold. If CQI of all of available carriers for a UE is ot higher tha the threshold, the least 4
Utilizatio 0.90 0.85 0.80 0.75 LL S 0.70 R S 0.65 LL J R J 0.60 Fig. 5. Utilizatio of bads for joit ad periodic techiques i R ad LL methods. Delay (ms) 0.30 0.25 0.20 0.15 0.10 0.05 LL S R S LL J R J 0.00 Fig. 6. Average delay of a packet for joit ad periodic techiques i R ad LL methods. loaded carriers are assiged to the UE. Similar ways are followed for R method i the simulatio except that R method radomly assigs carriers to users. We have used a mi-delay packet schedulig method i order to compare joit ad selective techiques. Packet arrival traffics are kept same for all test cases. Because of ad enb positios, CQI Idex for all carriers ca be oe of four optios which are give i Table. II. Each packet is trasferred by usig oe of assiged carriers. To icrease the efficiecy ad QoS, packet trasferrig priority is give to the CC, which is the closest to the enb ad miimizes packet delay if multiple carriers are available. If there is o available assiged carriers to serve arrived packets, packets are equeued to correspodig user queues. If there is ot ay empty spaces i queues, arrived packets are dropped. D. Observatio Methodology The results i Sectio V are average of 200 realizatios for differet size of users. The impact of light ad heavy users loads o joit ad selective techiques is ivestigated by usig modified Radom (R) ad Least Load (LL) methods. R ad LL methods are selected for test cases because of commo usage i the literature ad simplicity. Radom periodic carrier assigmet with joit techique (R J ), Radom periodic carrier assigmet with selective techique (R S ), Least Load periodic carrier assigmet with joit techique (LL J ), ad Least Load periodic carrier assigmet with selective techique (LL S ) have bee compared (Superscripts J ad S i the figures represet joit ad selective techiques, respectively). V. RESULTS I this sectio, overall system performace ad experieced performace by each device type are preseted for joit ad periodic techiques. A. Overall Performace of the System I this subsectio, the overall system performace for both LL ad R methods with joit ad selective techiques is preseted by usig bad utilizatio, delay ad throughput rate. 1) Utilizatio: Fig. 5 shows the utilizatio of bads for joit ad selective techiques for LL ad R methods. Utilizatio slowly icreases whe the umber of users is higher tha 25 for all cases. However, utilizatio is a suddely icrease for all cases whe the umber of user is chagig from 10 to 25. Throughput Rate 1.0 0.9 0.8 0.7 0.6 0.5 LL S R S LL J R J 0.4 Fig. 7. Throughput rate for joit ad periodic techiques i R ad LL methods. That is because of icreases i packet arrivals ad the bads are ot busy whe the umber of users is 10. Though a large umber of users (100 users), utilizatio of all cases does ot reach peak rate (=1) because all users are i the coverage area of Bad-a but ot i coverage areas of Bad-b ad Bad-c. Thus, the utilizatio of bads does ot reach peak rate although utilizatio of Bad-a is high. Moreover, there is o sigificat differece betwee bad utilizatio of joit ad selective techiques for R method except whe the umber of users is 25 or 50. While the umber of users is 25 or 50, selective techique has higher bad utilizatio tha joit techique for R method. O the other had, while the umber of users are lower tha 50, selective techiques has higher bad utilizatio tha joit techique for LL method ad it is vice verse whe the umber of users is higher 50. 2) Delay: Fig. 6 demostrates average delay per packet for joit ad selective techiques. The average delay is icreasig from 0.03 up to 0.27 ms. While the umber of users is growig, delay is regularly gettig higher for all cases. This is due to high umber of packet arrivals. For both LL ad R methods, selective techique has lower delay tha joit techique. It is worth to metio that while the umber of users is icreased, the average delay gab betwee selective ad joit is geerally growig for R method ad decreasig for LL method. 3) Throughput: Fig. 7 shows throughput rate for joit ad selective techiques. The throughput rate is decreasig from 1 to 0.45 because icreases i the umber of users leads to gradually reduce throughput rate for all cases. For both LL ad R methods, selective techique has higher throughput rate tha joit techique. It is importat to metio that the throughput gab betwee selective ad joit techiques is almost same for R method except whe the umber of user is 50. While the umber of the users is 50, selective techique has much higher throughput rate tha joit techique for R method. However, the throughput gab betwee selective ad joit techiques is decreasig for LL method while the umber of users is greater tha 50. B. Experieced Performace by Device Types We also ivestigate the experieced performace by device types (LTE ad LTE-A device types). Subscripts L ad F represet LTE ad LTE-A type equipmet, respectively i Figs. 8 5
Delay (ms) 0.35 0.30 0.25 0.20 0.15 0.10 0.05 0.00 Fig. 8. Average delay of LTE ad LTE-A type pakets for joit ad periodic techiques i R ad LL methods. ad 9. For example, RL S represets Radom periodic carrier assigmet with selective techique for LTE type equipmet ad RF S represets Radom periodic carrier assigmet with selective techique for LTE-A type equipmet. 1) Delay: Fig. 8 shows average delay per packet which is experieced by LTE ad LTE-A type equipmet for joit ad selective techiques. Average delay of LTE type equipmet is higher tha average delay of LTE-A type equipmet for both R ad LL methods because there is oly oe assiged CC to serve for LTE type equipmet ad multiple CCs for LTE- A type equipmet. Moreover, selective techique remarkably decreases average delay of LTE type equipmet ad slightly improves average delay time of LTE-A type equipmet comparig to joit techique for both R ad LL methods because of packet trasfer iterruptio i joit techique. 2) Throughput Rate: Fig. 9 demostrates throughput rate which is experieced by LTE ad LTE-A types equipmet for joit ad selective techiques. Throughput rate of LTE type equipmet is lower tha throughput rate of LTE-A type equipmet for both R ad LL methods because of oe CC assigmet for LTE type devices. Furthermore, selective techique sigificatly icrease throughput rate of LTE type equipmet ad slightly improves throughput rate of LTE-A type equipmet for both R ad LL methods. LL S L R S L LL J L R J L LL S F R S F LL J F C. Summary of Results Based o the results, we make the followig observatios: (i) joit techique leads that LTE type equipmet traffic suffers higher delay tha LTE-A type equipmet traffic, (ii) selective techique sigificatly ehaces the performace for LTE ad LTE-A type equipmet, ad (iii) selective techique remarkably decreases overall (up to 35%) average delay ad improve (up to 25%) throughput rate comparig to joit techique i modified R ad LL methods. VI. CONCLUSION I this paper, selective periodic compoet carrier assigmet techique is proposed by cosiderig behavior of system durig the compoet carrier assigmet operatios. The performaces of curret joit ad proposed selective compoet carrier assigmet techiques are compared by usig aalytic aalysis based o queuig algorithm ad a extesive simulatio. Results show that the proposed techique efficietly uses system resources ad improves throughput rate up to 25% ad average delay time up to 35% i LTE ad R J F Throughput Rate 1.0 0.9 0.8 0.7 0.6 0.5 0.4 0.3 Fig. 9. Throughput rate of LTE ad LTE-A type equipmet for joit ad periodic techiques i R ad LL methods. LTE-A systems. Our proposed techique ad related aalysis will help service providers build efficiet periodic compoet carrier assigmet methods i order to improve performaces metrics such as throughput rate ad delay. REFERENCES [1] I. F. Akyildiz, D. M. Gutierrez-Estevez, ad E. C. Reyes, The evolutio to 4G cellular systems: LTE-Advaced, Physical Commuicatio, vol. 3, pp. 217 244, Mar 2010. [2] Y. Wag, K. Pederse, T. Sorese, ad P. Mogese, Carrier load balacig ad packet schedulig for multi-carrier systems, IEEE Trasactios o Wireless Commuicatios, vol. 9, o. 5, pp. 1780 1789, May 2010. [3] L. xiag Li, Y. a Liu, F. Liu, G. Xie, K. mig Liu, ad X. yag Ge, Resource schedulig i dowlik LTE-Advaced system with carrier aggregatio, The Joural of Chia Uiversities of Posts ad Telecommuicatios, vol. 19, o. 1, pp. 44 49, Feb 2012. [4] N. Kolehmaie, J. Puttoe, P. Kela, T. Ristaiemi, T. Hettoe, ad M. Moisio, Chael quality idicatio reportig schemes for UTRAN log term evolutio dowlik, i IEEE Vehicular Techology Coferece, Sigapore, May 11-14, 2008, pp. 2522 2526. [5] S.-B. Lee, S. Choudhury, A. Khoshevis, S. Xu, ad S. Lu, Dowlik MIMO with frequecy-domai packet schedulig for 3GPP LTE, i INFOCOM, Rio de Jaeiro, Apr 19-25, 2009, pp. 1269 1277. [6] S. Dothi ad N. Mehta, Performace aalysis of subbad-level chael quality idicator feedback scheme of LTE, i Natioal Coferece o Commuicatios, Cheai, Ja 29-31, 2010. [7] F. Liu, W. Xiag, Y. Zhag, K. Zheg, ad H. Zhao, A ovel QoEbased carrier schedulig scheme i LTE-Advaced etworks with multiservice, i Vehicular Techology Coferece, Quebec City, Caada, Sep 3-6, 2012. [8] W. Fu, Q. Kog, W. Tia, C. Wag, ad L. Ma, A QoS-aware schedulig algorithm based o service type for LTE dowlik, i Iteratioal Coferece o Computer Sciece ad Electroics Egieerig, Hagzhou, Chia, Mar 22-23, 2013, pp. 2468 2474. [9] R. Sivaraj, A. Pade, K. Zeg, K. Govida, ad P. Mohapatra, Edgeprioritized chael- ad traffic-aware uplik carrier aggregatio i LTE- Advaced systems, i Iteratioal Symposium o a World of Wireless, Mobile ad Multimedia Networks, Sa Fracisco, CA, Jue 25-28, 2012. [10] T. Dea ad P. Flemig, Trukig efficiecy i multi-carrier CDMA systems, i 56th Vehicular Techology Coferece, Vacouver, Caada, Sep 24-28, 2002, pp. 156 160. [11] X. Cheg, G. Gupta, ad P. Mohapatra, Joit carrier aggregatio ad packet schedulig i LTE-Advaced etworks, i Commuicatios Society Coferece o Sesor, Mesh ad Ad Hoc Commuicatios ad Networks, New Orleas, LA, Jue 24-27, 2013, pp. 469 477. [12] J. Wastrom. (2013, Jue) LTE-Advaced. Accessed: Mar. 18, 2015. [Olie]. Available: http://www.3gpp.org/techologies/keywordsacroyms/97-lte-advaced [13] H. Narma ad M. Atiquzzama, Carrier compoets assigmet method for LTE ad LTE-A systems based o user profile ad applicatio, i IEEE GLOBECOM Workshop o Broadbad Wireless Access, Austi, TX, Dec. 12, 2014. [14] L. Che, W. Che, X. Zhag, ad D. Yag, Aalysis ad simulatio for spectrum aggregatio i lte-advaced system, i 70th Vehicular Techology Coferece, Achorage, AK, Sep 20-23, 2009. LL S L R S L LL J L R J L LL S F R S F LL J F R J F 6