idnits 2.17.1 draft-multi-dimensional-resource-aggregation-02.txt: Checking boilerplate required by RFC 5378 and the IETF Trust (see https://trustee.ietf.org/license-info): ---------------------------------------------------------------------------- No issues found here. Checking nits according to https://www.ietf.org/id-info/1id-guidelines.txt: ---------------------------------------------------------------------------- No issues found here. Checking nits according to https://www.ietf.org/id-info/checklist : ---------------------------------------------------------------------------- ** The document seems to lack a Security Considerations section. ** The document seems to lack an IANA Considerations section. (See Section 2.2 of https://www.ietf.org/id-info/checklist for how to handle the case when there are no actions for IANA.) ** There are 22 instances of too long lines in the document, the longest one being 7 characters in excess of 72. Miscellaneous warnings: ---------------------------------------------------------------------------- == The copyright year in the IETF Trust and authors Copyright Line does not match the current year -- The document date (October 29, 2019) is 1612 days in the past. Is this intentional? Checking references for intended status: Informational ---------------------------------------------------------------------------- -- Looks like a reference, but probably isn't: '1' on line 77 -- Looks like a reference, but probably isn't: '2' on line 77 -- Looks like a reference, but probably isn't: '3' on line 89 -- Looks like a reference, but probably isn't: '4' on line 99 -- Looks like a reference, but probably isn't: '5' on line 105 -- Looks like a reference, but probably isn't: '6' on line 111 -- Looks like a reference, but probably isn't: '7' on line 113 -- Looks like a reference, but probably isn't: '8' on line 132 -- Looks like a reference, but probably isn't: '9' on line 287 == Unused Reference: 'Ref1' is defined on line 379, but no explicit reference was found in the text == Unused Reference: 'Ref2' is defined on line 382, but no explicit reference was found in the text == Unused Reference: 'Ref3' is defined on line 386, but no explicit reference was found in the text == Unused Reference: 'Ref4' is defined on line 389, but no explicit reference was found in the text Summary: 3 errors (**), 0 flaws (~~), 5 warnings (==), 10 comments (--). Run idnits with the --verbose option for more detailed information about the items above. -------------------------------------------------------------------------------- 2 Cross Stratum Optimization Research Group H. Yang 3 Internet-Draft YQ. Liu 4 Intended status: Informational J. Zhang 5 Expires: May 1, 2020 Beijing University of Posts and Telecommunications 6 Y. Ao 7 Y. Qiuyan 8 Beijing University of Posts and Telecommunications. 9 October 29, 2019 11 Multi-dimensional Resource Aggregation in 5G Optical Fronthaul Networks 12 draft-multi-dimensional-resource-aggregation-02 14 Abstract 16 We propose a resource assignment scheme based on multi-dimensional 17 resource aggregation in 5G optical fronthaul networks. This new 18 scheme can suit to the higher demand of flexible resource allocation 19 of the fronthaul in the new 5G scenario. 21 Status of This Memo 23 This Internet-Draft is submitted in full conformance with the 24 provisions of BCP 78 and BCP 79. 26 Internet-Drafts are working documents of the Internet Engineering 27 Task Force (IETF). Note that other groups may also distribute 28 working documents as Internet-Drafts. The list of current Internet- 29 Drafts is at https://datatracker.ietf.org/drafts/current/. 31 Internet-Drafts are draft documents valid for a maximum of six months 32 and may be updated, replaced, or obsoleted by other documents at any 33 time. It is inappropriate to use Internet-Drafts as reference 34 material or to cite them other than as "work in progress." 36 This Internet-Draft will expire on May 1, 2020. 38 Copyright Notice 40 Copyright (c) 2019 IETF Trust and the persons identified as the 41 document authors. All rights reserved. 43 This document is subject to BCP 78 and the IETF Trust's Legal 44 Provisions Relating to IETF Documents 45 (https://trustee.ietf.org/license-info) in effect on the date of 46 publication of this document. Please review these documents 47 carefully, as they describe your rights and restrictions with respect 48 to this document. Code Components extracted from this document must 49 include Simplified BSD License text as described in Section 4.e of 50 the Trust Legal Provisions and are provided without warranty as 51 described in the Simplified BSD License. 53 Table of Contents 55 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . 2 56 2. 5G FRONTHAUL MODEL . . . . . . . . . . . . . . . . . . . . . 3 57 3. Multi-dimensional RESOURCE aggregation ALGORITHM . . . . . . 5 58 3.1. SIMULATION AND RESULTS . . . . . . . . . . . . . . . . . 7 59 4. CONCLUSION . . . . . . . . . . . . . . . . . . . . . . . . . 8 60 5. Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . 9 61 6. Informative References . . . . . . . . . . . . . . . . . . . 9 62 Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . . 9 64 1. Introduction 66 With the development of computer technology, the application of 5G 67 technology has become more and more extensive. For its ultra-high 68 transmission rate and huge data capacity, 5G technology has made 69 great achievements in our daily life and work. The future 5G network 70 will integrate artificial intelligence, SDN, NFV, and cloud computing 71 technologies to adapt to more and more complex application scenarios. 73 The 5G network architecture is totally different from the 4G network. 74 The application of cloud technology has emerged in the 5G network 75 architecture. In the traditional C-RAN, all the base station 76 computing resources are aggregated into the BBU pool, and distributed 77 radio frequency signals are collected by RRH[1][2]. Parts of the 5G 78 network are centralized into several clouds according to their 79 separate functions which are controlled to form the "three clouds" 80 architecture of the 5G network. The access cloud supports multiple 81 wireless access modes, including converged centralized and 82 distributed. It is able to be adaptable in various backhaul links 83 and increase flexibility in the whole network. The control cloud is 84 used to achieve local and global session control and realize the 85 mobility management and QOS. It also builds an open interface for 86 business-oriented network capabilities. The transmit cloud improves 87 the reliability and reduces the latency of the whole network. It 88 also achieves efficient transmission of massive traffic data flow 89 under the control of the control cloud [3]. Moreover, compared with 90 the 4G network architecture, the 5G architecture separates the base 91 station processing unit, and reconstructs the BBU unit according to 92 the real-time nature of the processing content into two functional 93 entities which are CU and DU. The CU is mainly responsible for the 94 deployment of some core network functions sinking and edge 95 application services. The DU mainly handles the functions of the 96 physical layer and real-time requirements. The original BBU baseband 97 function is moved up to the AAU to reduce the transmission bandwidth 98 between the DU and the RRU. Centralized deployment of CUs can 99 facilitate flexible resource allocation [4]. 101 Based on the situation where the networking is dense, the resource 102 allocation is complex and diverse under the background of 5G network 103 and there are many allocation schemes which have been proposed. We 104 can use mobile cloud computing (MCC) technology to achieve joint 105 energy minimization [5]. From the perspective of cross-layer 106 resource allocation, we can consider this question as a mixed integer 107 nonlinear programming (MINLP), jointly consider elastic service 108 scaling, RRH selection and Combine beamforming, and optimize it with 109 a pruning algorithm. However, this greatly increases the complexity 110 of the algorithm and reduces the timeliness of resource allocation 111 [6]. Also, there is hybrid coordinated multi-point transmission 112 scheme (H-COMP) for downlink transmission between C-RAN and FUN-LLS 113 [7]. They can all improve the efficiency of resource allocation and 114 suggest the idea of ??joint scheduling, but they ignored the 115 separating and sinking 5G-RAN structure. 117 It becomes an important issue that we should use resources 118 efficiently as the 5G network architecture changes and the 119 application scenarios are more complex. In this paper, we have a 120 more detailed division of the resources in the 5G scenario. In the 121 second section, we define the functional model of 5G resource 122 allocation. In the third section, we propose a resource allocation 123 algorithm which adapts to the new requirements of the new scenario. 124 In the fourth section, we perform the simulation and obtain the 125 results. Finally, we will analyze the results and make out the 126 conclusions. 128 2. 5G FRONTHAUL MODEL 130 The 5G Wireless Access Network (RAN) is expected to increase the 131 number of access users while reducing latency to handle more and more 132 connected devices and data rates[8]. In the 5G RAN architecture, the 133 AAU (Active Antenna Processing Unit) includes some physical units of 134 the formal RRH, BBU, and transmits radio frequency signals to the DU. 135 The signal transmission of this part is defined as the transmission 136 in 5G fronthaul. Due to the separation of the BBU (base station 137 processing unit) in the 5G network, the CU which processes the 138 virtual resource and the DU which processes the physical layer 139 function are logically independent. So the resource transmission 140 between DU and AAU can be separately analyzed and optimized. 142 According to the 5G fronthaul network architecture, resources can be 143 divided into three levels: DU resources, AAU resources, and 144 transmission resources. Thus we can optimize resources allocation in 145 these three levels .From the view of form, the transmission resource 146 and the computing resource span the transmission layer and the DU 147 processing layer in the horizontal direction. In terms of the 148 capacity ability, the multi-layer structure and networking are 149 working in the vertical direction, which is shown in Figure.1. Based 150 on this virtual mode, a 5G fronthaul network functional architecture 151 can be proposed. According to the classified resource types, the DU 152 controller DC, the AAU controller AC, and the transmission controller 153 TC are respectively used to control each part. 155 The AC (AAU controller) is used to control the allocation of AAU 156 resources. It can acquire and manage virtual radio resources and 157 perform radio frequency allocation on them. The DC (DU controller) 158 is used to control and obtain the DU resource information through 159 external triggers and interact with the TC. The TC (transfer 160 controller) is used to control the transmission resource. When the 161 service request arrives, the TC performs the resource estimation 162 algorithm on the DU, the AAU, and the transmission resource, and 163 performs resource allocation according to the algorithm result. (As 164 is demonstrated in Figure.2). 166 ----------------------------------------- 167 | ---------- | 168 | | AAU | | 169 | ---------- | 170 | | | 171 | ---------- | 172 | | WDM | | 173 | ---------- | 174 | | | 175 | ------ ---------- ------- | 176 | | DU |--| TRAMSFER |--| DU | | 177 | ------ ---------- ------- | 178 | | 179 ----------------------------------------- 181 Fig.1 5G network architecture 183 ----------------------------------------------------------------------------- 184 | AAU ----------- ------------- ----------- | 185 | | AAU |-------| AAU |-------| AAU | | 186 | CONTROLLER |ALLOCATION | | MONITORING | | MODEL | | 187 | ----------- ------------- ----------- | 188 | | | 189 --------------------|-------------------------------------------------------- 190 --------------------|-------------------------------------------------------- 191 | TRANSFER ----------- ------------- ----------- | 192 | | TRANSFER |-------| PCE+ |--------| DBM | | 193 | CONTROLLER | CONTROL | | OPENFLOW | | | | 194 | ----------- ------------- | | | 195 | | | | | | 196 | ----------- ------------- | | | 197 | | CSO | | RAA |--------| | | 198 | ----------- ------------- ----------- | 199 --------------------|-------------------------------------------------------- 200 --------------------|-------------------------------------------------------- 201 | DU ----------- ------------- ----------- | 202 | CONTROLLER |CSO AGENT |-------|DU MONITORING|--------| DU MODEL | | 203 | ----------- ------------- ----------- | 204 ----------------------------------------------------------------------------- 206 Fig.2 5G function model 208 3. Multi-dimensional RESOURCE aggregation ALGORITHM 210 Considering the resource allocation in the 5G application scenario, 211 we use AAU, DU, and transmission resources to optimize multi-layer 212 resources. Compared with the traditional situation where only one 213 resource model optimization is considered to evaluate resource 214 utilization, the resource allocation scheme in 4G context is no 215 longer applicable to 5G technology scenarios. Based on the proposed 216 functional architecture, we design a resource allocation algorithm 217 for 5G scenarios. 219 First, the node is defined and expressed as G (A, A', R, R', T, T', 220 C) according to the functional architecture mentioned above. Here, A 221 = {a1, a2, ... an} and A' = {a1', a2', ... an'} represent a 222 collection of AAU transmission nodes. R = {r1, r2, ... rn} and R' = 223 {r1', r2', ... rn'} represent a bidirectional transmission link group 224 between A and A'. T = {t1, t2, ... tn} and T' = {t1',t2', ... tn'} 225 represent the set of spectra on each link. Also, A, A', R, R', T, 226 T', and C represent the number of all types of nodes. For DU 227 resources, two time -varying- processing parameters are used to 228 describe and represent the case of resource utilization, including 229 the resource storage rate U0 and CPU memory usage U1. In addition, 230 the transmission layer parameters include the candidate path hop 231 count H and the weight W of each link occupied bandwidth. The AU 232 processing layer parameters include the symbol rate Br and the radio 233 frequency Fr. DU is used to provide storage capacity and computing 234 resources. 236 We denote a request as SRi(S, B, U0, U1) according to its attributes. 237 B denotes the bandwidth. The resource allocation algorithm selects 238 the corresponding path and DU according to the state parameters 239 acquired by the DU, the states of the AC, and the TC. In order to 240 comprehensively consider the resource scheduling of all the three 241 levels of DU, AAU, and transport layer, a resource allocation factor 242 ?? is used to jointly allocate the resources of these three 243 dimensions. For the DU layer, two parameters U0 and U1 are used to 244 describe the current resource usage of the DU part, and a 245 normalization factor ?? is used to coordinate the storage utilization 246 and CPU usage in the DU layer, which is shown in formula (1). In the 247 case of the transport layer, the traffic weights W and the candidate 248 path hop count H are used to indicate the load balance of the 249 transmission link. For the bearer link, the larger the traffic 250 weight is, the smaller the link redundancy of the barer space is. 251 Therefore, the traffic should be selected. A link with a small 252 weight is better as expressed in formula (2). For the AAU layer, the 253 radio frequency spectrum resources and symbol rate occupancy should 254 be considered. Considering the symbol parameter Fr and the radio 255 frequency parameter Br, since the radio frequency is negatively 256 correlated with the carrying capacity, the AAU layer resource is 257 represented by the formula (3). DU parameters, transmission 258 parameters, and AAU parameters are represented by fa, fb, and fc, 259 respectively. 261 The nodes with the smallest processing function in the DU, AAU, and 262 transmission layer are respectively represented as Fa, Fb, Fc. And 263 the two resource coordination factors of ?? and ?? are combined to 264 perform multi-layer resources which are normalized by Fa, Fb, and Fc. 265 The normalization process is expressed as equation (4). When the 266 minimum value is obtained according to ??, the most appropriate path 267 and node are selected, and corresponding resource allocation is 268 performed. 270 The relevant algorithm flowchart is given in Figure3. First, we 271 obtain the relevant resource utilization of each layer of the input 272 service request SRi. Then we use it to calculate the parameters of 273 each layer and get the resource allocation parameter ??. And 274 additionally, we compare all the parameters and find the minute one. 275 Finally, find the path and node corresponding to the min ?? and 276 perform radio frequency allocation. 278 3.1. SIMULATION AND RESULTS 280 In order to test the optimization of the resource allocation of the 281 scheme and verify its efficiency, we also made several comparisons 282 between the proposed algorithm and the traditional one. The 283 traditional way for resource allocation optimizes the processing of 284 spectrum resources based on the virtualization of network functions. 285 It combines both the centralized and distributed elements. It can 286 also independently develop centralized control platforms, such as 287 virtualization and sectioning of network[9]. They use network 288 throughput as the optimization goal and consider the use of only one 289 certain resource in a single way. They do not refine the resources 290 according to the difference of user services and the architecture of 291 network development. 293 Based on the software test platform, we build a simulation model. We 294 use the Open vSwitch proxy controller to control the interaction 295 between the nodes. In the 5G fronthaul, the heavy traffic load is 296 from 40 Erlang to 150 Erlang. For the proposed model and the 297 Openflow-based control platform, three virtual machine deployment 298 planes are used: the TC server supports the interaction between AC 299 and DC. The DC server is used to acquire and supervise the DU 300 computing resources. The AC server obtains the radio distribution. 301 On the established platform, the optimization of the proposed 302 solution is demonstrated by testing the resource occupancy rate and 303 path provision latency of the server. Based on the proposed resource 304 allocation algorithm, the preset weight ?? is set to 50%, so that the 305 CPU occupancy rate and the resource storage rate occupy the same 306 proportion, and then the preset weights ??, ?? are set to 33.33%, so 307 that the resources occupy of the three layers can gain the same 308 weight. And the CPU storage rate occupied by each service is 309 randomly allocated between 0 and 1%. When the request reaches, the 310 best path and node will be calculated according to the formula, and 311 the corresponding RF resources will be provided. Then we obtain the 312 relevant indicators and compare them. In order to get the 313 optimization of time precision, we compared the path provision 314 latency between our way and the traditional way. And the results are 315 shown in Figure.4 where GES represents the scheme above and CSO 316 represents the traditional one. What??s more, in order to obtain the 317 resource utilization of the proposed method, we also compared the 318 resource occupation rate between those two ways. The experiments 319 proved that the scheme we proposed could improve the efficiency of 320 resource allocation. The path provision latency is lower and the 321 resource occupation rate is higher. It means that this solution has 322 many advantages for 5G fronthaul resource allocation and can improve 323 the flexibility of the whole network. 325 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 326 | | path provision | 327 | Traffic load +-+-+-+-+-+-+-+-+-+-+-+-+-+ 328 | | CSO | GES | 329 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 330 | 40 | 29.1 | 25.7 | 331 | 60 | 32.7 | 27.3 | 332 | 80 | 35.1 | 28.8 | 333 | 100 | 36.6 | 32.7 | 334 | 120 | 42.5 | 38.0 | 335 | 140 | 49.4 | 43.3 | 336 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 338 Tab.1 path provision of two strategies 340 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 341 | |resource occupation rate | 342 | | path provision | 343 | Traffic load +-+-+-+-+-+-+-+-+-+-+-+-+-+ 344 | | CSO | GES | 345 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 346 | 40 | 0.05 | 0.06 | 347 | 60 | 0.11 | 0.14 | 348 | 80 | 0.19 | 0.23 | 349 | 100 | 0.32 | 0.37 | 350 | 120 | 0.40 | 0.50 | 351 | 140 | 0.51 | 0.58 | 352 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 354 Tab.2 resource occupation rate of two strategies 356 4. CONCLUSION 358 In summary, this paper considers the resource allocation requirements 359 in the 5G technology scenario. According to the changes of the 5G 360 network architecture and the multiple use of resources, we 361 redistribute the resources and propose the corresponding functional 362 models. It is used to adopt a resource allocation algorithm to 363 optimize the resource allocation of each layer and realize the joint 364 deploy and utilization of multi-layer resources. In the traditional 365 resource allocation model, we used to consider the utilization of 366 only one certain type of resource. This solution realizes the global 367 deployment of 5G fronthaul resources, which is able to improve the 368 flexibility of the 5G fronthaul network. 370 5. Acknowledgments 372 This work has been supported in part by NSFC project (61501049), 373 Fundamental Research Funds for the Central Universities (2018XKJC06) 374 and State Key Laboratory of Information Photonics and Optical 375 Communications (BUPT), P. R. China (No. IPOC2017ZT11). 377 6. Informative References 379 [Ref1] Yang, Hui., Zhang, Jie., and Yongli. Zhao, "CSO: Cross 380 Stratum Optimization for Optical as a Service", Aug 2015. 382 [Ref2] Yang, H. and Jie. Zhang, "Experimental demonstration of 383 multi-dimensional resources integration for service 384 provisioning in cloud radio over fiber network", 2016. 386 [Ref3] Yao, Li., "Joint Optimization of BBU Pool Allocation and 387 Selection for C-RAN Networks", 2018. 389 [Ref4] Ramon, Casellas., "Control, Management, and Orchestration 390 of Optical Networks: Evolution, Trends, and Challenges", 391 2018. 393 Authors' Addresses 395 Hui Yang 396 Beijing University of Posts and Telecommunications 397 No.10,Xitucheng Road,Haidian District 398 Beijing 100876 399 P.R.China 401 Phone: +8613466774108 402 Email: yang.hui.y@126.com 403 URI: http://www.bupt.edu.cn/ 405 Yiqian Liu 406 Beijing University of Posts and Telecommunications 407 No.10,Xitucheng Road,Haidian District 408 Beijing 100876 409 P.R.China 411 Phone: +8613177087617 412 Email: 497706153@qq.com 413 URI: http://www.bupt.edu.cn/ 414 Jie Zhang 415 Beijing University of Posts and Telecommunications 416 No.10,Xitucheng Road,Haidian District 417 Beijing 100876 418 P.R.China 420 Phone: +8613911060930 421 Email: lgr24@bupt.edu.cn 422 URI: http://www.bupt.edu.cn/ 424 Ao Yu 425 Beijing University of Posts and Telecommunications. 426 No.10,Xitucheng Road,Haidian District 427 Beijing 100876 428 P.R.China 430 Email: yuaoupc@163.com 431 URI: http://www.bupt.edu.cn/ 433 Qiuyan Yao 434 Beijing University of Posts and Telecommunications. 435 No.10,Xitucheng Road,Haidian District 436 Beijing 100876 437 P.R.China 439 Email: yqy86716@126.com 440 URI: http://www.bupt.edu.cn/