基于隨機學習的接入網(wǎng)服務功能鏈部署算法

陳前斌; 楊友超; 周鈺; 趙國繁; 唐倫

doi:10.11999/JEIT180310

基于隨機學習的接入網(wǎng)服務功能鏈部署算法

doi: 10.11999/JEIT180310 cstr: 32379.14.JEIT180310

重慶郵電大學通信與信息工程學院 ??重慶 ??400065

基金項目: 國家自然科學基金(61571073)

詳細信息

作者簡介:
陳前斌：男，1967年生，教授，博士生導師，主要研究方向為個人通信、多媒體信息處理與傳輸、下一代移動通信網(wǎng)絡

楊友超：男，1993年生，碩士生，研究方向為網(wǎng)絡虛擬化和切片資源分配

周鈺：男，1993年生，碩士生，研究方向為切片資源分配和深度學習

趙國繁：女，1993年生，碩士生，研究方向為5G網(wǎng)絡切片中的資源分配、可靠性

唐倫：男，1973年生，教授，博士生導師，主要研究方向為新一代無線通信網(wǎng)絡、異構(gòu)蜂窩網(wǎng)絡

通訊作者:
陳前斌　chenqb@cqupt.edu.cn

中圖分類號: TN929.5
計量
- 文章訪問數(shù): 1735
- HTML全文瀏覽量: 880
- PDF下載量: 63
- 被引次數(shù): 0
出版歷程
- 收稿日期: 2018-04-02
- 修回日期: 2018-09-03
- 網(wǎng)絡出版日期: 2018-09-12
- 刊出日期: 2019-02-01

Deployment Algorithm of Service Function Chain of Access Network Based on Stochastic Learning

School of Communication and Information Engineering, Chongqing University of Posts and Telecommunications, Chongqing 400065, China

Funds: The National Natural Science Foundation of China (61571073)

摘要

摘要:
針對5G云化接入網(wǎng)場景下物理網(wǎng)絡拓撲變化引起的高時延問題，讀文提出一種基于部分觀察馬爾可夫決策過程(POMDP)部分感知拓撲的接入網(wǎng)服務功能鏈(SFC)部署方案。該方案考慮在5G接入網(wǎng)C-RAN架構(gòu)下，通過心跳包觀測機制感知底層物理網(wǎng)絡拓撲變化，由于存在觀測誤差無法獲得全部真實的拓撲情況，因此采用基于POMDP的部分感知和隨機學習而自適應動態(tài)調(diào)整接入網(wǎng)切片的SFC的部署，優(yōu)化SFC在接入網(wǎng)側(cè)的時延。為了解決維度災問題，采用基于點的混合啟發(fā)式值迭代算法求解。仿真結(jié)果表明，該模型可以優(yōu)化部署接入網(wǎng)側(cè)的SFC，并提高接入網(wǎng)吞吐量和資源利用率。
- 網(wǎng)絡切片 /
- SFC動態(tài)部署 /
- 網(wǎng)絡拓撲感知 /
- 部分觀察馬爾可夫決策過程
Abstract:
To solve problem of the high delay caused by the change of physical network topology under the 5G access network C-RAN architecture, this paper proposes a scheme about dynamic deployment of Service Function Chain (SFC) in access network based on Partial Observation Markov Decision Process (POMDP). In this scheme, the system observes changes of the underlying physical network topology through the heartbeat packet observation mechanism. Due to the observation errors, it is impossible to obtain all the real topological conditions. Therefore, by the partial awareness and stochastic learning of POMDP, the system dynamically adjust the deployment of the SFC in the slice of the access network when topology changes, so as to optimize the delay. Finally, point-based hybrid heuristic value iteration algorithm is used to find SFC deployment strategy. The simulation results show that this model can support to optimize the deployment of SFC in the access network side and improve the access network’s throughput and resource utilization.
- Network slice /
- Service Function Chain (SFC) dynamic deployment /
- Network topology perception /
- Partial Observation Markov Decision Process (POMDP)

HTML全文

圖 1 系統(tǒng)模型

下載: 全尺寸圖片幻燈片

圖 2 接入網(wǎng)VNF部署方式

下載: 全尺寸圖片幻燈片

圖 3 3種SFC部署方案的吞吐量

下載: 全尺寸圖片幻燈片

圖 4 3種SFC資源分配算法方案的資源利用率

下載: 全尺寸圖片幻燈片

圖 5 3種POMDP求解算法對比

下載: 全尺寸圖片幻燈片

圖 6 3個切片的接入網(wǎng)VNF部署方式統(tǒng)計圖

下載: 全尺寸圖片幻燈片

表 1 算法1：更新探索信念點集合${{{B}_{\rm su}$

　(1) 用式(13)計算被擴點集${B^{{\rm pr}}$

　(2) for all ${} \in {B^{{\rm pr}}$ do

　(3) 用式(14)計算su$({})$

　(4) 用式(15)計算離${B_{{\rm su}}$最遠的后繼信念點${{}''}$

　(5) end for

　(6) 清空集合${V'}$的元素

　(7) for all ${} \in {B_{{\rm su}}$ do

　(8) 用式(17)計算下界向量${{{α}} _{}$并加入${V'}$中

　(9) end for

　(10) 將下界集合$\underline V $ 更新為${V'}$

　(11) for all ${} \in {B_{{\rm su}}$ do

　(12) ${V_{{{co}}} \leftarrow \{ {}|\exists s \in S,b(s) = 1\} $

　(13) ${v^0_{}\leftarrow \displaystyle\sum\limits_{{b'} \in {V_{\rm co}}} {v({{}'}) \cdot {}} $

　(14) for all $ < {b_i},{v_i} > \in {B_{{\rm su}} - {V_{{\rm co}}$ do

　(15) $c({b_i}) \leftarrow \mathop {\min }\limits_{s \in S} \frac{{b(s)}}{{{b_i}(s)}}$

　(16) $f({b_i}) \leftarrow {v_i} - \sum\limits_{{'} \in {V_{{\rm co}}}} {v({{'}){b_i}(s)} $

　(17) end for

　(18) $v \leftarrow {v^0_} + \mathop {\displaystyle\min }_i c({b_i})f({b_i})$并將點值對$ < {},v > $加入上界集合$\mathop { V}\limits\!\!\!\!^{\displaystyle{-} } $

　(19) end for

下載: 導出CSV

表 2 算法2：基于${{{B}_{{\rm su}}$更新值函數(shù)向量集${{{Γ} _{{{t +}}1}$

　(1) for all ${} \in {B_{{{\rm su}}$ do

　(2) 向量集合${{{Γ} _{t + 1,\chi }} \leftarrow \varnothing $

　(3) for all $a \in A$ do

　(4) 向量${{{Γ} _{t + 1,\beta }} \leftarrow 0$

　(5) for all $z \in Z$ do

　(6) 用式(18)計算${{Γ} _{t + 1}^{a,z}$

　(7) ${{{Γ} _{t + 1,\beta }} \leftarrow \mathop {\arg \max }_{{α} \in {{Γ} _{t + 1}^{a,z}}} {} \cdot {{α}} + {{Γ} _1^a$

　(8) end for

　(9) 將向量${{{Γ} _{t + 1,\beta }}$加入集合${{{Γ} _{t + 1,\chi }}$中

　(10) end for

　(11) 將${{{Γ} _{t + 1,\chi }}$中與${}$相乘最大的向量加入${{{Γ} _{t + 1}}$

　(12) end for

下載: 導出CSV

參考文獻(16)

SHARMA S, MILLER R, and FRANCINI A. A cloud-native approach to 5G network slicing[J]. IEEE Communications Magazine, 2017, 55(8): 120–127. doi: 10.1109/MCOM.2017.1600942

ZHANG Haijun, LIU Na, and CHU Xiaoli. Network slicing based 5G and future mobile networks: Mobility, resource management, and challenge[J]. IEEE Communications Magazine, 2017, 55(8): 138–145. doi: 10.1109/MCOM.217.1600940

KATSALIS K, NIKAEIN N, and SCHILLER E. Network slices toward 5G communications: Slicing the LTE network[J]. IEEE Communications Magazine, 2017, 55(8): 146–154. doi: 10.1109/MCOM.2017.1600936

FOUKAS X, PATOUNAS G, and ELMOKASHFI A. Network slicing in 5G: Survey and challenges[J]. IEEE Communications Magazine, 2017, 55(5): 94–100. doi: 10.1109/MCOM.2017.1600951

LI Xin and SAMAKA M. Network slicing for 5G: Challenges and opportunities[J]. IEEE Internet Computing, 2017, 21(5): 20–27. doi: 10.1109/MIC.2017.3481355

MIJUMBI R, SERRAT J, and GORRICHO J L. Network function virtualization: state-of-the-art and research challenges[J]. IEEE Communications Surveys Tutorials, 2017, 18(1): 236–262. doi: 10.1109/COMST.2015.2477041

GIL J H and BOTERO J F. Resource allocation in NFV: A comprehensive survey[J]. IEEE Transactions on Network and Service Management, 2016, 13(3): 518–532. doi: 10.1109/TNSM.2016.2598420

HUANG Huawei and SONG Guo. Service chaining for hybrid network function[J]. IEEE Transactions on Cloud Computing, 2017. doi: 10.1109/TCC.2017.2721401

QU Long, ASSI C, and SHABAN K. Delay-aware scheduling and resource optimization with network function virtualization[J]. IEEE Transactions on Communications, 2016, 64(9): 3746–3758. doi: 10.1109/TCOMM.2016.2580150

MAHMOOD A M, AL-YASIRI A, and ALANI O Y K. A new processing approach for reducing computational complexity in cloud-RAN mobile networks[J]. IEEE Access, 2018, 6: 6927–6946. doi: 10.1109/ACCESS.2017.2782763

CHIH I. RAN revolution with NGFI (xhaul) for 5G[J]. Journal of Lightwave Technology, 2018, 36(2): 541–550. doi: 10.1109/JLT.2017.2764

ZHANG Nan, LIU Yafeng, and FARMANBAR H. Network slicing for service-oriented networks under resource constraints[J]. IEEE Journal on Selected Areas in Communications, 2017, 35(11): 2512–2521. doi: 10.1109/JSAC.2017.2760147

HAYASHIBARA N, DEFAGO X, and YARED R. The φ accrual failure detector[C]. IEEE International Symposium on Reliable Distributed Systems, Florianpolis, Brazil, 2014: 66–78.

劉峰. 基于部分可觀察馬爾科夫決策過程的序列規(guī)劃問題的研究[D]. [博士論文], 南京大學, 2015.

LIU Feng. A study of sequence planning based on partially observable markov decision process[D]. [Ph.D. dissertation], Nanjing University, 2015.

CILDEN E and POLAT F. Toward generalization of automated temporal abstraction to partially observable reinforcement learning[J]. IEEE Transactions on Cybernetics, 2017, 45(8): 1414–1425. doi: 10.1109/TCYB.2014.2352038

ZHENG Qiang, ZHENG Kan, ZHANG Haijun, et al. Delay-optimal virtualized radio resource scheduling in software-defined vehicular networks via stochastic learning[J]. IEEE Transactions on Vehicular Technology, 2016, 65(10): 7857–7867. doi: 10.1109/TVT.2016.2538461

施引文獻

資源附件(0)

訪問統(tǒng)計