![/24
16
3. Resource optimization
scheduling and allocation - SDN-based source scheduling
Task scheduling of cloud services scenarios
Each subtask can be divided into is different
Data transmission from the task can be performed
The processors are heterogeneous
The Available bandwidth between nodes changes with time
The task completion time includes task execution time and
data transmission time
(a)Total energy consumption
m: nodes
n: resources
x: subtask [0-1]
e: subtask EC
(b)Resource utilization
total: total time(trans+process)
h=1
→ 1− σ𝑗=1
𝑛
𝑡𝑜𝑡𝑎𝑙 − 𝑅𝐸𝑇𝑗 ÷ (𝑚 × 𝑡𝑜𝑡𝑎𝑙)
(c)Fitness function
w1+w2+w3=1](https://image.slidesharecdn.com/ppt-pdfresourceoptimizationschedulingandallocationforhierarchicaldistributedcloudservicesysteminsmar-211028094419/85/Paper-sharing_resource-optimization-scheduling-and-allocation-for-hierarchical-distributed-cloud-service-system-in-smart-city-17-320.jpg)
Paper sharing_resource optimization scheduling and allocation for hierarchical distributed cloud service system in smart city
- 1. Resource optimization scheduling and allocation for hierarchical distributed cloud service system in smart city From Future Generation Computer Systems 107 Author：Jing Li (2020) Presenter ：CHEN, YOU-SHENG 2021/10/28
- 2. JCR - Future Generation Computer Systems /24 1
- 3. Vocabularies 1/2 /24 2 P. English Chinese 247 VCE (Virtual Cloud Embedding) 虛擬雲端嵌入 247 PSO (particle swarm optimization) 粒子群最佳化 247 tenant resource 租用資源 247 ultra-high-density 超高密度 247 heterogeneous cells 異質蜂巢基地台 247 macro cells 大型基地台 247 seamless access 無縫存取 247 distribution capabilities 分配能力 247 cope with 應付 248 accelerate 促進 P. English Chinese 248 ubiquitous cellular network 無所不在的行動 網路 248 emerge one after another 相繼出現 248 spectrum 射頻頻譜 248 DEM (Digital Terrain Model) 數值高型模型 249 interference coordination 干擾協調 249 mutual assistance 互援 249 spread across 擴散到整個 250 Statistical multiplexing 統計式多工 250 SDN (software-defined networking) 軟體定義網路 251 fine-grained 细粒度的
- 4. Vocabularies 2/2 /24 3 P. English Chinese 251 fitness function 適應度函式 252 AMTS (adaptive multi- objective task scheduling) 自適多目標任務調度 252 FCPN (Fuzzy color Petri net) 模糊有色派翠網路 252 SPN (shortest job first) 最短程序優先 252 DPSN (distribution power system network) 分散式電力系統網路 252 time slots 時間空檔 252 Markov chain 馬可夫鏈 252 node propositions 節點定位 252 average throughput 平均吞吐量 P. English Chinese 252 spatial complexity 空間複雜度 253 prematurely falling 過早下降 253 fault tolerance 容錯 254 backlog 待辦事項 254 clearly dominant 占主導地位 254 amplitude of vibration change 震幅的震動變化
- 5. 1. Introduction 2. Smart City hierarchical distributed cloud service system for 5G 3. Resource optimization scheduling and allocation 4. Experimental simulation 5. Conclusion /24 4
- 7. 1. Introduction /24 6 5G mobile communications must be able to meet extremely high communication requirements 2018 2018 2017 S. Imtiaz, G.P. Koudouridis, H. Ghauch/ B. Zhang, X. Mao, J.L. Yu/ Mayoral, R. Munoz, R. Vilalta • System must deploy ultra-high-density heterogeneous cells • Full-area coverage can meet users’ seamless access to the network 2018 2017 2017 2015 A. An, G.D. Veciana/ G. Femenias, F. Riera-Palou, X. Mestre/ G.H.S. Carvalho, I. Woungang, A. Anpalagan/ A. Aissioui, A. Ksentini, A.M. Gueroui • 5G faces the dual challenge -Network data transmission -Data processing capabilities 2017 2016 2014 2014 2013 C. Li, Y.C. Liu, X. Yan/ C.I. Badoi, N.R. Prasad, R. Prasad/ Z. Cao, J. Lin, Y. Song/ T. Ma, Y. Chu, L. Zhao/ J. Li, M. Qiu, Z. Ming • Network devices are a relatively closed system. • Difficult to coordinate with the computing and storage resources of node
- 8. 1. Introduction /24 7 ◼ The HDCSN (hierarchical distributed cloud service network) model is an abstract model that meets the basic requirements of the industry for the 5G network system architecture ◼ Author proposes a resource graph scheduling algorithm(AMTS, Adaptive Multi-Objective Task Scheduling) based on PSO (particle swarm optimization) algorithm to accelerate the optimization of Smart City system resource AI warning system • AI model requires high number of training samples and test sample • Little change cannot solve the accuracy • Use spatial analysis functions Resource management system • Multiple isolated systems in the data center • Difficult to apply various distributed file system
- 9. /24 8 2. Smart City hierarchical distributed cloud service system for 5G
- 10. /24 9 2. Smart City hierarchical distributed cloud service system for 5G 2017 2013 2013 S. Li, Z. Zhang, T. Robertazzi/ Y. Laili, T. Fei, Z. Lin/ J.T. Tsai, J.C. Fang, J.H. Chou 5G signals enable users to use the entire system anytime and anywhere. Seamless access is to the network 2013 B. Addis, D. Ardagna, B. Panicucci Because of the rapid development, 5G systems must be able to achieve a large number of M2M(or D2D) communications, making communication smooth and convenient 2017 A. Afram, F. Janabi-Shari fi, A.S. Fung, et al. Real-time risk system forecast results do not reflect the advantages of DEM modeling, and the application time. The span is short, only one forecast result is listed, and its application effect needs further testing HDCSN was proposed and this model consists of three levels HDCSN Model Access Cloud Core Cloud Distributed Micro Cloud
- 11. /24 10 2. Smart City hierarchical distributed cloud service system for 5G - Access Cloud This layer provides access to wireless networks C-RAN technology is used to perform centralized processing and allocation of resources • RRH mainly performs receiving and transmitting of wireless signals • BBU and RRH connected by high-speed fiber links (remote) RRHs C Radio Remote Heads BBUs (Base Band Units) Advantage Energy efficiency has been improved Cooperation mechanisms are easy to implement Cell deployment can be performed Optimize the allocation of resources (ex: spectrum、power) Construction cost low, can fixed cost 70% less Easy to upgrade and update the system 圖為4G天線設備（左1、2）及5G整合天線及RRU的AAS 設備（右1）
- 12. /24 11 2. Smart City hierarchical distributed cloud service system for 5G - Distributed Micro Cloud A Smart City distributed local micro cloud system is formed basis on C-RAN Advantage Equipped with an appropriate calculate virtual resources Reduce the data load of the mobile communication network More effectively perform mobile Internet Reduces the delay for mobile users to obtain data Data sharing and mutual assistance can be realized Predict and perceive the user’s migration direction in advance The micro cloud can also provide high quality local application service
- 13. /24 12 2. Smart City hierarchical distributed cloud service system for 5G - Core Cloud The providers need to efficiently connect lots of physical computing entities and network devices to form one or several large data center that are spread across multiple locations 2019 2017 W. Wei, X. Xia, W. Marcin, et al./ M. Barshan, H. Moe ns, S. Latre Researches on resource deployment and resource virtualization 2019 Yougang Sun, Haiy an Qiang, Junqi Xu, Guobin Lin Studies how mobile terminals can efficiently manage resources through flexible applications in the cloud
- 14. /24 13 3. Resource optimization scheduling and allocation
- 15. /24 14 3. Resource optimization scheduling and allocation Smart City cloud service need a unified resource allocation system A cloud service cluster resource scheduling should have the following Quantified and the quality of service is ensured, monitor and quantify the resource usage status Scheduling is to coordinate real-time information, avoid task failure and other resource waste Centralized strategy center, implement a controllable and programmable software- defined cloud center
- 16. /24 15 3. Resource optimization scheduling and allocation - SDN-based source scheduling Using resource vector graph model can be applied to diverse resource demand occasions through dimensionality reduction and nested expansion of resource vector • If echo node link and switch resources with a cluster (independent) • If resource node can be a vector containing various resources • If continuing to fine-grained, computing resources are SDN(Software-defined networking) that enables dynamic, programmatically efficient network configuration in order to improve network performance and monitoring
- 17. /24 16 3. Resource optimization scheduling and allocation - SDN-based source scheduling Task scheduling of cloud services scenarios Each subtask can be divided into is different Data transmission from the task can be performed The processors are heterogeneous The Available bandwidth between nodes changes with time The task completion time includes task execution time and data transmission time (a)Total energy consumption m: nodes n: resources x: subtask [0-1] e: subtask EC (b)Resource utilization total: total time(trans+process) h=1 → 1− σ𝑗=1 𝑛 𝑡𝑜𝑡𝑎𝑙 − 𝑅𝐸𝑇𝑗 ÷ (𝑚 × 𝑡𝑜𝑡𝑎𝑙) (c)Fitness function w1+w2+w3=1
- 18. /24 17 3. Resource optimization scheduling and allocation - SDN-based source scheduling Particle’s position and velocity formula v: velocity r: rand(0-1) c:self pos pb:self best gb:group best Round 1 Round 2
- 20. /24 19 4. Experimental simulation CloudSim platform is used to simulate the task scheduling algorithm in the cloud computing environment, and Genetic Algorithm(GA) is adopted to do the comparison test Experimental parameters are set as follows： T1 = 20 ( Round1 tasks) T2 = 150 ( Round2 tasks) N1 = 5 ( Round1 resource) N2 = 20 ( Round2 resource) Runs 5 times and the average value is taken as the comparison MIPS is the unit of Pj and means The execution speed of resource node j. The parameter α represents the zooming factor in GA, and β indicates the crossover factor in GA.
- 21. /24 20 4. Experimental simulation The task completion time of the PSO-based AMTS algorithm at each iteration is less than that of GA It means that the PSO-based AMTS algorithm can converge quickly than GA under the two kinds of deployment especially in the initial stage of experiments We can get the conclusion that PSO-based AMTS algorithm can ensure complete tasks more quickly
- 22. /24 21 4. Experimental simulation The average energy consumption of PSO-based AMTS algorithm is less than that of GA under the two kinds of deployment
- 23. /24 22 4. Experimental simulation It represents that PSO-based AMTS algorithm consumes less energy than GA under the two scenarios
- 25. 5. Conclusion /24 24 3-level distributed cloud service network model (HDCSN) of “access cloud + distributed micro cloud + background core cloud’’ is constructed The performance parameters of the algorithm are verified by simulation and comparison analysis, and we found： • The PSO-based AMTS algorithm is proposed to optimize the response speed and resource efficiency of multi-node collaborative scheduling (maximize resource utilization and minimize task completion time, energy consumption) • The algorithm is an effective scheduling algorithm, and can obtain better optimal solutions
- 26. Declaration： There are some problems in this paper. I have tried my best to explain that the experimental data are not from the data of this paper.
- 27. Resources 1. Resource optimization scheduling and allocation for hierarchical distributed cloud service system in smart city/ Future Generation Computer Systems 107(2020) 247–255/ Jing Li / Download form https://www.sciencedirect.com/science/article/pii/S0167739X1932028X 2. AMTS: Adaptive multi-objective task scheduling strategy in cloud computing/ China Communications 13(2016) 162-171/ Hua He; Guangquan Xu; Shanchen Pang; Zenghua Zhao/ https://ieeexplore.ieee.org/document/7464133 3. PPT template- Tulsa Powerpoint and Keynote Template/ Download form https://www.templatesppt.com/tulsa- presentation-template#preview 4. P6. Tuan Nguyen,小型基地台確保5G連線能力，擷取圖1：異質的無線網路基礎架構, EET Taiwan from https://www.eettaiwan.com/20180420ta31-small-cell-networks-and-the-evolution-of-5g/ 5. P7,P14,P15 Microsoft Stock images (royalty-free images) 6. P10.中華電信5G新機房首公開！ 「C-RAN」更省電、故障率低(天線設備圖) https://news.ltn.com.tw/news/life/breakingnews/3289109 7. P17. PSO粒子移動示意圖 https://dotblogs.com.tw/dragon229/2013/01/10/87127
- 28. Extended learning 1. 小型基地台確保5G連線能力 https://www.eettaiwan.com/20180420ta31-small-cell-networks-and-the-evolution-of-5g 2. 粒子群最佳化 https://zh.m.wikipedia.org/wiki/粒子群优化 3. 以Python實作粒子群演算法 https://medium.com/qiubingcheng/以python實作粒子群演算法-particle-swarm- optimization-pso-f0d0404c443b 4. PSO 粒子群演算法 https://dotblogs.com.tw/dragon229/2013/01/10/87127 5. GA 基因演算法 https://dotblogs.com.tw/dragon229/2013/01/03/86692 6. Cloud-RAN架構導入新興異質網路 https://archive.eettaiwan.com/www.eettaiwan.com/ART_8800701517_675327_TA_505d784c.HTM 7. 與行動網路相輔相成 Cloud RAN滿足5G寬頻服務 https://www.2cm.com.tw/2cm/zh- tw/tech/F2F99C400C694C059F860E36DAEA779B 8. 中華電信5G關鍵基礎建設首次曝光! 靠C-RAN機房集中管理多個基地臺，提高網路穩定度 https://www.ithome.com.tw/news/139961
- 29. Extended learning 1. 解讀TD-SCDMA標準的背後意涵 https://www.sogi.com.tw/articles/解讀TD_SCDMA標準的背後意涵/6095131 2. 至強融核(MIC) https://zh.m.wikipedia.org/wiki/至强融核 3. 軟體定義網路5G通訊網路重要核心技術 https://blog.twnic.tw/2021/02/25/17185/ 4. Cloud RAN: 雲端無線接取網路與應用課程單元：Remote Radio Heads (RRH) http://140.117.164.12/data/C- RAN/Session05-Remote%20Radio%20Head%20(RRH).pptx 5. 什麼是SLA服務品質保障協議？（一）https://medium.com/@mukung.fung/什麼是sla服務品質保障協議- 91bdc9f273c4 6. 三分鐘帶你入門瞭解openstack的Nova專案 https://iter01.com/551548.html 7. 馬可夫鏈 https://zh.m.wikipedia.org/wiki/马尔可夫链 8. 佩特里網 https://zh.m.wikipedia.org/wiki/佩特里網
- 30. Extended learning 1. AMTS: Adaptive multi-objective task scheduling strategy in cloud computing https://ieeexplore.ieee.org/document/7464133 2. VCE-PSO: Virtual Cloud Embedding through a Meta-heuristic Approach https://ieeexplore.ieee.org/document/6832157 3. Development of automated operating procedure system using fuzzy colored petri nets for nuclear power plants https://www.sciencedirect.com/science/article/pii/S0306454903003177 4. A quality control method for nuclear instrumentation and control systems based on software safety prediction https://www.semanticscholar.org/paper/A-quality-control-method-for-nuclear-and-control-on-Son- Seong/25d89d90d451e819992abd27fee1d36459f449cb 5. 結合停限電管理系統與模糊彩色派翠網路於配電系統饋線規劃之研究研究成果報告(精簡版) http://ir.lib.kuas.edu.tw/retrieve/8869/952221E151058.pdf