Topology Control for C-RAN Archtecture Based on Complex Network Zhanun Lu, Yung He, Yunpeng L, Zhaoy L, Ka Dng Chongqng key laboratory of moble communcatons technology Chongqng unversty of post and telecommuncaton Chongqng, Chna Abstract Amng at the adaptve topology control problem n wreless access network, ths paper proposes that the Remote Rado Unt (), the User Equpment (UE), and the channel between and UE are mapped as class node A, class node B, and sde respectvely by usng the centralzed processng characterstcs of Centralzed Random Access Network (C-RAN), and then bulds the complex network model. Furthermore, ths paper establshes the optmzaton model wth the mnmum total number of opened as the obectve functon, whch s based on the subect degree of class node A, and then acheves the adaptve topology control of. Smulaton analyss verfy that when the node densty s not too large, even though the proposed scheme lowers the average receved average sgnal strength of users compared wth the conventonal scheme by %~%, t reduces the energy consumpton by 8.8%~42.%. Keywords - C-RAN, Complex network, Communty dvson, Topology control of I. INTRODUCTION Users random dstrbuton and moblty cause traffc s tdal phenomenon n wreless access network [], and the conventonal desgn of network deployment and control mechansm are for peak traffc. However, n most cases, the network state s not n peak condton, and thus causes unnecessary energy waste. Therefore, network topology vares wth traffc state as well as adaptve topology control becomes a meanngful task. Up to now, the man dea of researchng adaptve topology control problem s to close the base statons wth lower load, and utlze surroundng base statons to support network coverage by means of power compensaton [3][4], and thus conquer the problem of energy waste. Exstng technologes are mplemented n a dstrbuted network, and the base statons energy savng management requres a large amount of sgnalng nteracton (.e., load condtons and power condton) among base statons. Therefore, these knds of technologes have strong lmtaton, because t can only acheve local network optmzatons [], whch demonstrate that dstrbuted network s a serous restrcton factor for ts development. Fortunately, the Centralzed Random Access Network (C-RAN) archtecture provdes possblty for the mprovement of adaptve topology control technologes. Centralzed Processng, Collaboratve Rado, and Real-tme Cloud Infrastructure are the man characterstcs of C-RAN, and ths network structure has been consdered as the development drecton of future wreless network []. In C- RAN, the base band resource pool can reduce the sgnalng nteracton among base statons by means of centralzed management of multple base statons date. Therefore, t can acheve the adaptve topology control of the network, whch also means that the characterstcs of centralzed management can acheve overall optmzaton confguraton for the whole network traffc [8]. The complex network technology holds the characterstcs of self-organzaton, and t can be mplemented to the adaptve control of nonlnear complex system. The topology control n C-RAN assocates wth the Remote Rado Unt (), User Equpment (UE) and the channel between and UE whch has hgh node densty, complex connecton and obvous communty dvson [9], and ths s dentcal to the complex network []. Therefore, the applcaton of complex network for C-RAN to acheve the adaptve topology control becomes a meanngful research topc. II. COMPLEX NETWORK MODEL A. Establshment of complex network The detaled steps of buldng complex network model by usng wreless access network are shown as below: ), UE, and the channel between and UE are mapped as class node A, class node B, and sde respectvely, and based on ths to buld the complex network. We assume that the locaton of class node A not vares wth tme, and has two states: open and close states. The locaton of class node B vares wth tme. Furthermore, sde only exst n between class node A and B, and only one class node B connects wth one class node A, and thus has sde connecton. One class node A can connect wth multple class nodes B. In addton, If the s n open state, then ths s mapped as onlne state, otherwse offlne. 2) The duraton of communcaton servce whch s provded by to UE s defned as the weght of correspondng two nodes, f -th whch provdes the duraton of communcaton servce for -th UE s T, then the W=T weght of sde between node and node s. DOI.3/IJSSST.a.7.22.4 4. ISSN: 473-84x onlne, 473-83 prnt
3) The aggregaton of sde s weght between class node A and all the connected class node B are denoted as the subect degree of class node A (SDA), whch reflects the mportance degree of each nodes to the network, and the subect degree of -th s denoted as B B, N = W B. Communty Dvson The detaled steps of communty dvson for nodes are shown as below: ) If the SDA s hgher than the open threshold, then the correspondng class nodes A are denoted as set O. If the SDA s lower than the close threshold, then the correspondng class nodes A are denoted as set C. Furthermore, all the remanng class node A s denoted as set Z. 2) Accordng to the locaton of nodes, f class nodes A s close to the nodes n set O, then these class nodes A are added to the set O (f these nodes already exst n set O, then gnore t). Furthermore, the communtes whch nclude the adacent class nodes A n set O are merged one by one. 3) The communtes whch nclude the adacent class nodes A n set C are merged one by one. 4) The communtes whch nclude the remanng class nodes A are merged to one communty. ) Mappng the communty dvson results of complex network back to the wreless access network, and then we can get the communty dvson results of and UE n each cell. III. TOPOLOGY CONTROL BASED ON COMPLEX NETWORK A. Topology Control of Hgh Subect Degree s Communty We buld the topology optmzaton model for the communty whch has hgh subect degree s. m nq= n = n S. T. N S (,..., m) S where denotes the symbol of closed, n s the number of closed, and m denote the symbol and number of whch has hgh subect degree respectvely, N N s the relatons between and. f, then s adacent to. S = and = N, f, then s not adacent S denote opened and closed, to respectvely. The detaled steps of solvng ths optmzaton model are shown as below: ) The mapped of SDA whch s larger than the open threshold are marked as H n complex network, the closed n the set s marked as watng open, and then creates open combnaton of all the watng open.. 2) Check each open combnaton whether t satsfes that the closed whch s adacent to the wth H mark s n = NS (,..., m) opened,.e.,. 3) Calculate the number of opened whch satsfes n Q= S = the above condtons,.e.,, thus the combnaton wth least number of opened s the optmzaton combnaton. B. Topology Control of Low Subect Degree s Communty We buld the topology optmzaton model for the communty whch has low subect degree s, whch s denotes as: m nq= m = k M S. T. S P N B S ( =,..., n;,..., m) S k k k where,, n, and m denote the symbol of watng close, UE symbol, number of UE, and number of low subect degree n the communty, respectvely; k and M denote the symbol and number of n the network, P respectvely; s the afflaton between Ue P = and, P = and Ue denote that belongs or not N belongs to, respectvely; k s the relatons between and k, N= N= k k and denote that s or s not B Ue k adacent to ; k denotes that whether can swtch to B = B = Ue k, k and k denote that can or cannot swtch to S= S= k ; and denote that do not close or close. The steps of solvng ths optmzaton model are shown as below: ) Generate the close combnaton of all the watng close for all the n the communty, and check whether the user n the closed can swtch to the adacent opened n each of the close combnaton,.e., M k S P N B S ( =,..., n;,..., m) k k k 2) Calculate the number of opened n each of = combnaton whch satsfy the above condtons,.e.,, and then the combnaton whch has least Q value s selected as the optmal topology result. m Q= S IV. SIMULATION AND PERFORMANCE ANALYSIS RESULTS A. Smulaton Envronment In the smulaton, we assume that the number of s 37, and all the ntalzed s n open state. The number of UE s 3, where % of users are randomly dstrbuted n the -th, -th, 7-th, 3-th, -th, -th cell, and the DOI.3/IJSSST.a.7.22.4 4.2 ISSN: 473-84x onlne, 473-83 prnt
remanng users are dstrbuted n the others cell. The moblty of UE follows unform moton, where % of users follow the normal dstrbuton wth mean.2km/h and varance., and the movement drecton has a degree wth X axs, and follows unform dstrbuton from -4 degree to - degree. % of users follow the normal dstrbuton wth mean.2km/h and varance., and the movement drecton has a degree wth X axs, and follows unform dstrbuton from 3 degree to 4 degree. The remanng users follow normal dstrbuton wth mean.2km/h and varance., and the movement drecton has a degree wth X axs, and follows unform dstrbuton from degree to degree. The detaled smulaton parameters are lsted n table. TABLE I SIMULATION PARAMETERS Parameters Values Number of 37 Nr. of UE 3 Carrer Frequency 2/GHz Transmsson Power of 4dBm Cell Radus 2m Open Threshold Close Threshold 4 Least Receved Power -44.dBm Occuped Bandwdth 8KHz PSD of Gaussan Whte Nose -9 n/w/hz B. Smulaton Performance ) Adaptablty Analyss Fgures - are the topology n dfferent tme, where colored denotes that s n open state, noncolored denotes close state. Y axs ( Unt: km ). -. - 3 3 4 7 3 时刻 T= 8 37 9 2 2 2 9 3 22 4 23 2 24 Y axs ( Unt: km ) Y axs ( Unt: km ) Yaxs ( Unt: km ) 时刻 T=2 8 37. 7 9 2 3 2 2 9 3 3 22 4 4 -. 3 23 2 24 - -. - -... Fgure 2. User dstrbuton and state when T=2.. -. - 3 3 4 7 3 时刻 T=3 8 37 9 2 2 2 9 3 22 4 23 2 24 -. - -... Fgure 3. User dstrbuton and state when T=3. 时刻 T=4 8 37. 7 9 2 3 2 2 9 3 3 22 4 4 -. 3 23 2 24 - -. - -... Fgure 4. User dstrbuton and state when T=4. -. - -... Fgure. User dstrbuton and state when T=. DOI.3/IJSSST.a.7.22.4 4.3 ISSN: 473-84x onlne, 473-83 prnt
Y axs ( Unt: km ) Y axs ( Unt: km ). -. - 3 3 4 7 3 时刻 T= 8 2 4 2 37 9 2 -. - -... Fgure. User dstrbuton and state when T=. 8 37. 7 9 2 3 2 2 9 3 3 22 4 4 -. 3 23 2 24 - -. - -... 3 时刻 T= Fgure. User dstrbuton and state when T=. From these fgures, we can see that topology vares wth users movement, and t can adaptvely adust topology based on the users dstrbuton. When users are less, and then close partal to reduce energy consumpton. When users are more, and then open partal to guarantee servce qualty. For example, users are less when T= n the -th and 3-th cell, then the s n close state, however, wth the bottom rght movement of massve users, the users n -th and 3-th cell becomes more and more, then the should be opened partally. When T=2, -th cell s s opened, when T=3, users are movng down, therefore, 3-th cell s opened. The frst 4 moment n -th cell are all larger, therefore, s n open state. When T= and T=, the users becomes less n -th cell, therefore, s n close state. Fnally, the above explanaton proves that our proposed topology control scheme can adaptvely adust s state to the varaton of network. 2) Energy Consumpton and Sgnal Strength Analyss From fgure 7, we can see that the energy consumpton of s 7dbm when the conventonal scheme s adopted. However, under the proposed scheme, the maxmum value of energy consumpton s 38 dbm, whch declnes 8.8% compared wth the conventonal scheme; the mnmum value s 97 dbm, whch declnes 42.% compared wth the conventonal scheme. 9 24 2 22 23 Fgure 8 shows that the average receved sgnal strength of users (ARSSU) s -.dbm~-.3dbm when we utlze the conventonal scheme. However, the ARSSU s -.9dbm~-.7dbm when the proposed scheme s utlzed. The mnmum dfference value s.4dbm,.e., the ARSSU reduces by.2%. The maxmum dfference value s.dbm,.e., the ARSSU declnes by 4.8%. Energy consumpton of (dbm) ARSSU (dbm) 8 7 4 3 2 Conventonal Scheme 9 2 3 4 Tme Fgure 7. Comparson of energy consumpton of. -.2 -.4 -. -.8 - -.2 -.4 -. -.8 Conventonal Scheme - 2 3 4 Tme Fgure 8. Comparson of ARSSU. Furthermore, we change the user densty n secton 3., and analyze the energy consumpton of and ARSSU whch vares wth the users densty. We assume that the number of users s, 2,, 3,, 4, 4 and, and the number of RRR keeps the same wth the secton 3.. From the fgure 9, we can observe that the energy consumpton decreases by 4.3% when the number of users s, and the energy consumpton decreases by 8.9% when the number of users s compared wth the conventonal scheme. Fgure shows that the ARSSU decreases by 4.8% when the number of users s, and the ARSSU decreases by.% when the number of users s. DOI.3/IJSSST.a.7.22.4 4.4 ISSN: 473-84x onlne, 473-83 prnt
Energy Consumpton of (dbm) 8 7 4 3 2 Conventonal Scheme be merged together accordng to the subect degree of to the whole network, and then we can get the communty dvson results. Fnally, the proposed scheme acheves the adaptve topology control by obtanng the least number of, whch s based on the optmzaton theory. Smulaton results shows that even though the average receved sgnal strength declnes by %-%, the energy consumpton of can be decreased by 8.8%-42.%. Therefore, the energy savng performance of our proposed schemes s better than the conventon scheme,.e., results to sgnfcant reductons of energy consumpton of. ARSSU (dbm) 9 2 3 4 4 Number of Users Fgure 9. Energy consumpton of vs. Number of UE. - -. - -. - Conventonal Scheme -. 2 3 4 4 Number of Users Fgure. ARSSU vs. Number of UE V. CONCLUSION In ths paper, we buld the complex network model by analyzng the characterstcs of C-RAN, and propose that, UE, and the channel between and UE are mapped as class node A, class node B, and sde respectvely, and then we can establsh the complex network. Furthermore, on ths bass, the smlar sngle communty can REFERENCES [] Hubn Jn, Janguang Guo, and Bn L, Research of Tdal plannng n C-RAN, Communcaton World,pp. 39-4,. [2] Csco Vsual Networkng Index: Global Moble Data Traffc Forecast Update, 23-, Csco, San Jose, CA, USA, 23. [3] Guyng Wu, Hgh energy dynamc topology research n next generaton wreless moble HetNet Network, Chengdu:Unversty Of Electronc Scence And Technology Of Chna,23. [4] Peng Yu, Energy savng management scheme n wreless communcaton network, BeJng:Beng Unversty of Posts and Telecommuncatons, 23. [] Jnsong Wu, Sun Ran and Hong gang Zhang, Green Communcaton, CRC Press, 23. [] Chna Moble Research Insttute, Centralzed resource pool, Collaboratve processng, Real-tme cloud green wreless access network archtecture, 2. [7] Xaoyun Wang, C-RAN: Evoluton of, Chna Communcaton, vol. 3, pp. 7-2, 2. [8] C. M. R, Insttute, C-ran: The road towards green ran, Avalable:labs.chnamoble.com/cran,2. [9] Palla G, Dereny I, Farkas I, Vcsek T, Uncoverng the overlappng communty structures of complex networks n nature and socety, Nature, vol. 4, No. 743, pp. 84 88,. [] Jax DI, KeYU, Xaofe WU, Xnyu ZHANG, Chunyng XU, A complex network vew on traffc flows of moble nternet, IEEE Beng Secton. Proceedngs of 23 th IEEE Internatonal Conference on Broadband Network & Multmeda Technology. IEEE Beng Secton, vol.,23. DOI.3/IJSSST.a.7.22.4 4. ISSN: 473-84x onlne, 473-83 prnt