1. An Advance Algorithm for Task Management On Activity Based Costing in Cloud Computing By : Ashutosh Ingole Sumit Chavan Rajesh Singh Sinhgad Institute of Technology,LONAVALA 19 th Apr 2012

2. Contents Introduction Cloud Computing Architecture Problem Statement Problems Overall Scenario Flow of Algorithm Cost calculation and Scheduling Feedback Mechanism Advantages Conclusion and References 19 th Apr 2012 Sinhgad Institute of Technology,LONAVALA

3. Introduction 19 th Apr 2012 What is Cloud Computing? Shared computing resources Shared computing resources As opposed to local servers and devices As opposed to local servers and devices Made up of Grid Infrastructure Made up of Grid Infrastructure Scalable Scalable Virtualization Virtualization Web applications Web applications Specialized raw computing services Specialized raw computing services Technical Key Points User interaction interface: How users of cloud interface with the cloud User interaction interface: How users of cloud interface with the cloud Services catalog: Services a user can request Services catalog: Services a user can request System management: Manages the resources available System management: Manages the resources available Provisioning tool: Carves out the systems from the cloud to deliver on the requested service Provisioning tool: Carves out the systems from the cloud to deliver on the requested service Monitoring and metering: Tracks the usage of the cloud (optional) Monitoring and metering: Tracks the usage of the cloud (optional) Servers: Virtual or physical servers managed by System management Servers: Virtual or physical servers managed by System management

4. Cloud Computing Architecture 19 th Apr 2012 Sinhgad Institute of Technology,LONAVALA

5. Problem Statement Sinhgad Institute of Technology,LONAVALA To reduce the total time required for task scheduling in Cloud Computing using Activity Based Costing Algorithm 19 th Apr 2012

6. Keywords Sinhgad Institute of Technology,LONAVALA Cloud Computing Cloud Computing Task Management Task Management Activity Based Costing Activity Based Costing Advance Algorithm Advance Algorithm 19 th Apr 2012

7. Overall Scenario $4 $1 $2 Dynamic Priority Allocator Queue 1 Queue 2 Queue 3 Cost Calculation Decide Priority According To Cost Virtual Machines Scheduling Server 19 th Apr 2012 Sinhgad Institute of Technology,LONAVALA

8. Flow of Algorithm 19 th Apr 2012 Sinhgad Institute of Technology,LONAVALA

9. Dependencies and DAG Sinhgad Institute of Technology,LONAVALA Get the Activity dependencies form Task divider module. Get the Activity dependencies form Task divider module. Calculate the DAG from the given information. Calculate the DAG from the given information. According to the DAG fill up the Task queue. According to the DAG fill up the Task queue. An Activity only starts after all its immediate predecessors finish. An Activity only starts after all its immediate predecessors finish. Activities with no immediate predecessor are entry-activities, and activities without immediate successors are exit-activities. Activities with no immediate predecessor are entry-activities, and activities without immediate successors are exit-activities. T1 T3 T2 T4 T1T3T2T4 19 th Apr 2012

10. Cost calculation of Task Sinhgad Institute of Technology,LONAVALA Parameters considered Resource cost Resource cost Arrival Time Arrival Time Predicted Execution Time Predicted Execution Time Dependency (DAG) Dependency (DAG) 19 th Apr 2012

11. Continued… (Formulae) Sinhgad Institute of Technology,LONAVALA 19 th Apr 2012 1)Resource Cost Evaluation 2)Estimated Execution Time Where, T ij : estimated execution time for i th activity of j th task Tj: estimated execution time for complete j th task K j : total number of activities in j th task

12. Continued… (Formulae) Sinhgad Institute of Technology,LONAVALA Total cost of Activity = 19 th Apr 2012 ∑ ( ∑ Resource cost for a single activity) * execution time Total cost of Task = ∑ (Activity Cost)

13. Concept of Waiting Queue Sinhgad Institute of Technology,LONAVALA In case of dependent activities, if a certain activity say A11 needs server1 and its dependent activity say A12 needs to be on server2, then without complete execution of A11, we cant forward A12 from ready queue to server2. In such a case, we also cant keep a2 in ready queue for a long time and hence we use waiting queue to store these kind of dependent activities Waiting Queue Ready Queue 19 th Apr 2012 A22A34A52 A62 …. A51 A61…. A12A31 A32 A33

14. Feedback and Update Sinhgad Institute of Technology,LONAVALA When we get feedback from server that a particular activity is completed then we can schedule its dependent activity from waiting queue to its destination server. When we get feedback from server that a particular activity is completed then we can schedule its dependent activity from waiting queue to its destination server. After that each resource table and universal resource table gets updated. After that each resource table and universal resource table gets updated. 19 th Apr 2012 Waiting Queue A22A34A52 A62 ….

15. Failure Cases Sinhgad Institute of Technology,LONAVALA Simulation parameters Average bounded slowdown and average per-processor slowdown Average waiting time & average weighted waiting time Network Error Computational Error Power Failure Queue Overflow 19 th Apr 2012

16. Conclusion Sinhgad Institute of Technology,LONAVALA  An ABC-system of low complexity, such as a system with a small number of cost drivers, is not only easier to handle but also easier to understand.  It seems to be an interesting next step to analyze how ABC decision rules perform when several approximations apply.  Thus ABC reduces the scheduling cost and reduces delays. 19 th Apr 2012

17. References Sinhgad Institute of Technology,LONAVALA 19 th Apr 2012 Sandeep Tayal, University School of Information Technology, Guru Gobind Singh, Indraprastha University, Delhi- 10006, India “Tasks Scheduling optimization for the Cloud Computing System2011, IJAEST,2011 Archana Ganapathi, Yanpei Chen, Armando Fox, Randy Katz, David Patterson Computer Science Division, University of California at Berkeley,” Statistics-Driven Workload Modeling for the Cloud”, IEEE 2011 QI CAO, ZHI-BO WEI, WEN-MAO GONG International School of Software Wuhan University Wuhan, China, “An Optimized Algorithm for Task Scheduling Based On Activity Based Costing in Cloud Computing”, IEEE 2009 Jinhua Hu, Jianhua Gu, Guofei Sun, Tianhai Zhao, NPU HPC Center Xi‟an, China “A Scheduling Strategy on Load Balancing of Virtual Machine Resources in Cloud Computing Environment”, IEEE 2010 Jiahui Jin, Junzhou Luo, Aibo Song, Fang Dong and Runqun Xiong School of Computer Science and Engineering, Southeast University Nanjing, P.R. China “BAR: An Efficient Data Locality Driven Task Scheduling Algorithm for Cloud Computing”, IEEE 2011 GAN Guo-ning, HUANG Ting-Iei, GAO Shuai School of Computer science and engineering Guilin University of Electronic Technology Guilin, China “Genetic Simulated Annealing Algorithm for Task Scheduling based on Cloud Computing Environment”, IEEE 2010 [1] [2] [3] [4] [5] [6]

18. Any Questions…??? 19 th Apr 2012 Sinhgad Institute of Technology,LONAVALA

19. Thank You…!!! 19 th Apr 2012 Sinhgad Institute of Technology,LONAVALA