由於機器學習的技術不斷的進步，任務對於雲端群集的需求逐漸增加，這些資源密集型的任務需要更多的計算資源來支持，更多的計算資源和任務就表示對於資源分配要有更好的處理。對於阿里巴巴所提出的模擬器，它是利用貪婪演算法進行任務排程，因為貪婪演算法存在著一些缺點，因此本論文提出以Min-Min演算法和Max-Min演算法以動態任務分配的方式來改善貪婪演算法的缺點。另外，所提出之排程演算法因具有高時間複雜度，這對於大規模分散式系統的任務排程造成了挑戰，所以提出了將任務已更小的單位切割任務資料集後再進行排程。實驗結果表明，Min-Min演算和Max-Min演算法之於阿里巴巴所提出的模擬器結果在任務的 Makespan 有良好的表現，而在切割任務後的結果來看也有良好的表現。

Due to the continuous advancements in machine learning technology, there is a growing demand for tasks in cloud clusters. These resource-intensive tasks require more computational resources to support them. More tasks and computational resources imply a need for improved resource allocation. In the case of the simulator proposed by Alibaba, it utilizes a greedy algorithm for task scheduling. However, since greedy algorithms have their limitations, this paper suggests enhancing the drawbacks of the greedy algorithm by employing the Min-Min and Max-Min algorithms with dynamic task allocation.

Furthermore, the scheduling algorithms introduced here present a challenge due to their high time complexity, especially in the context of task scheduling in large-scale distributed systems. Therefore, it is proposed to break down the task dataset into smaller units for scheduling. Experimental results indicate that both the Min Min and Max-Min algorithms perform well in terms of the Makespan of tasks in Alibaba's simulator. Moreover, when considering the results after task partitioning, these algorithms still demonstrate excellent performance.

[1] Akhilesh , J., & Arvind , U. (2019). Cloud scheduling using improved hyper heuristic framework. In R. Kamal, M. Henshaw, & P. Nair (Ed.), InteInternational Conference on Advanced Computing Networking and Informatics (pp. 127-133). Singapore: Springer Singapore.
[2] Alibaba. (2022, March 7). Alibaba Cluster Trace/cluster-trace-gpu-v2020. Retrieved from github.com: https://github.com/alibaba/clusterdata
[3] Arun, K. S., Michael , S., & Zhiyong , Z. (2018). Chapter 10 - Metaheuristic Algorithms: A Comprehensive Review. In A.-B. Mohamed , A.-F. Laila , & S. K. Arun , Computational Intelligence for Multimedia Big Data on the Cloud with Engineering Applications (pp. 185-231). London: Academic Press.
[4] Chauhan, S., & Joshi, R. (2010). A weighted mean time Min-Min Max-Min selective scheduling strategy for independent tasks on Grid. 2010 IEEE 2nd International Advance Computing Conference (IACC), (pp. 4-9). Patiala, India.
[5] Fard, H., Prodan, R., & Wolf, F. (2020). Dynamic Multi-objective Scheduling of Microservices in the Cloud. 2020 IEEE/ACM 13th International Conference on Utility and Cloud Computing (UCC), (pp. 386-393). Leicester, UK.
[6] freeCodeCamp. (2019, 11 19). Greedy Algorithms Explained with Examples. Retrieved from freeCodeCamp.org: https://www.freecodecamp.org/news/what-is-a-greedy-algorithm/
[7] Gaurang , P., Rutvik, M., & Upendra, B. (2015). Enhanced Load Balanced Min-min Algorithm for Static Meta Task Scheduling in Cloud Computing. 3rd International Conference on Recent Trends in Computing 2015 (ICRTC-2015), 57, pp. 545-553. Ghaziabad, India.
[8] Huang, J., Xiao, C., & Wu, W. (2020). RLSK: A Job Scheduler for Federated Kubernetes Clusters based on Reinforcement Learning. 2020 IEEE International Conference on Cloud Engineering (IC2E), (pp. 116-123). Sydney, NSW, Australia.
[9] Jørgen, B.-J., Gregory, G., & Anders , Y. (2004, 11 15). When the greedy algorithm fails. Discrete Optimization, 1(2), pp. 121-127.
[10] Leo , B., Jerome , F., Charles , S. J., & Richard , A. (1984). Classification and Regression Trees. New York: Routledge.
[11] Lohi, S., & Tiwari, N. (2020). A high performance machine learning algorithm TspINA; scheduling multifariousness destined tasks by better efficiency. 2020 Fourth World Conference on Smart Trends in Systems, Security and Sustainability (WorldS4), (pp. 603-607). London, UK.
[12] Mala , K., & Sarbjeet, S. (2015, 11 01). A review of metaheuristic scheduling techniques in cloud computing. Egyptian Informatics Journal, 16, pp. 275-295.
[13] Min-Yi , T., Ping-Fang , C., Yen-Jan, C., & Wei-Jen, W. (2011). Heuristic scheduling strategies for linear-dependent and independent jobs on heterogeneous grids. Grid and Distributed Computing: International Conference (pp. 496--505). Jeju Island, Korea: Springer.
[14] Mokhtari, A., Hossen, M., Jamshidi, P., & Salehi, M. (2022). FELARE: Fair Scheduling of Machine Learning Tasks on Heterogeneous Edge Systems. 2022 IEEE 15th International Conference on Cloud Computing (CLOUD), (pp. 459-468). Barcelona, Spain.
[15] Qizhen, W., Wencong, X., Yinghao, Y., Wei, W., Cheng, W., Jian, H., . . . Yu, D. (2022). MLaaS in the wild: Workload analysis and scheduling in Large-Scale heterogeneous GPU clusters. 19th USENIX Symposium on Networked Systems Design and Implementation (NSDI 22) (pp. 945--960). Renton, WA: USENIX Association.
[16] Sarraiuand, S., & Bien, F. (2020). Predictive Technique Of Task Scheduling For BigData In Cloud. 2020 International Conference on Emerging Smart Computing and Informatics (ESCI), (pp. 11-16). Pune, India.
[17] Vashishth, V., Chhabra, A., & Sood, A. (2017). A predictive approach to task scheduling for Big Data in cloud environments using classification algorithms. 2017 7th International Conference on Cloud Computing, Data Science & Engineering - Confluence, (pp. 188-192). Noida, India.
[18] Xiaotang , W., Minghe , H., & Jianhua, S. (2012). Study on Resources Scheduling Based on ACO Allgorithm and PSO Algorithm in Cloud Computing. 2012 11th International Symposium on Distributed Computing and Applications to Business, Engineering & Science, (pp. 219-222). Guilin, China.
[19] Xiumin, Z., Gongxuan, Z., Jin, S., Junlong, Z., Tongquan, W., & Shiyan, H. (2019, 4 1). Minimizing cost and makespan for workflow scheduling in cloud using fuzzy dominance sort based HEFT. Future Generation Computer Systems, 93, pp. 278-289.
[20] Yihui, F., Zhi, L., Yunjian , Z., Tatiana, J., Yidi , W., Yang, Z., . . . Tao, G. (2021). Scaling large production clusters with partitioned synchronization. 2021 USENIX Annual Technical Conference (USENIX ATC 21) (pp. 81--97). USENIX Association.