520 Student Presentation Grid Sim Grid Modeling and

520 Student Presentation Grid. Sim – Grid Modeling and Simulation Toolkit 1

Outline • • Introduction GRACE Framework Nimrod-G Resource Broker Grid. Sim Grid Resource Scheduling Simulation Toolkit 2

Introduction • Inspired from electrical power Grid • Computational power Grid • Sharing, aggregation geographically distributed resources • Low level services(security, information, directory, resource management) • High level services(resource discovery, cost negotiation, resource selection) 3

GRACE Framework • Grid Architecture for Computational Economy • builds on the existing Grid systems • new services for resource management and trading and aggregation 4

A generic Grid architecture for computational economy 5

Commodity Market Models • GRP specify their service price • charge users according to the amount of resource consumed • GRP submit price specification to GTS • CPU cycles, storage, software, and network 6

Nimrod-G Resource Broker • economic-based resource management and scheduling algorithms • user-defined deadline and budget constraints • schedule optimizations and manages supply and demand 7

Architecture of Nimrod-G system • TFE – creation of jobs – job status • The scheduler – – resource discovery resource trading resource selection job assignment • Dispatcher – deploys jobs • Agents – setting up execution environment – transporting the code and data 8

The Flow actions in Nimrod-G 9

Economic-based Scheduling Algorithms • Cost Optimization – 1. Sort resources. – 2. assign jobs in turn, without exceeding the deadline. • Time Optimization – 1. For each resource, calculate the next completion time for an assigned job – 2. Sort resources by next completion time. – 3. Assign one job to the first resource for which the cost is within the budget. – 4. Repeat steps 1 -3 until done. 10

World-Wide Grid (WWG) Testbed 11

Parameter Sweep Application • Create a set of jobs with different parameters • Calculate the angular values from different degrees 12

Cost Optimisation Scheduling Australian peak time Australian off-peak time 13

Cost Optimisation Scheduling Australian peak time Australian off-peak time 14

Cost and Time Optimization Scheduling 15

Cost and Time Optimization Scheduling time optimization scheduling cost optimization scheduling 16

The need for Simulation Tools • • real testbed is not available expensive and time consuming limited to a few resources and domains testing scheduling algorithms for scalability and adaptability • scheduler performance is hard to trace 17

The core Entities of Grid. Sim 18

Event diagram among entities • internal event • external event • synchronous event • asynchronous event 19

Nimrod-G simulation 1. A set of Gridlets 2. Find resources and their costs, create resource list 3. Select scheduling policy based on user requirement 4. Dispatch Gridlets to resources 5. Submits Gridlets to resources 6. Updates runtime parameter to help predict job consumption rate 7. Repeat 3 -6 till finish all jobs or exceed deadline or budget 20

WWG testbed resources simulated using Grid. Sim 21

Cost-optimization scheduling 22

Cost-optimization scheduling 23

Time Optimization Scheduling 24

Comparing the Cost and Time Optimization Scheduling 25

Cost-Time Optimization Scheduling • 1. Sort resources in increasing order • 2. assign jobs in turn, without exceeding the deadline • 3. If more two or more resources have the same cost and capacity, schedule resources based on the Time Optimization algorithm 26

Cost-Time Optimization Scheduling 27

Cost vs. Cost-Time optimization scheduling 28

Cost vs. Cost-Time optimization scheduling 29

Cost vs. Cost-Time optimization scheduling 30

Conclusion • The Cost-Time optimization algorithm is more efficient than Cost optimization algorithm. • To choose Cost-Time optimization or to choose Time optimization algorithm depends on your deadline and budget. 31

Reference 1. 2. 3. 4. 5. R. Buyya, D. Abramson, and J. Giddy, A Case for Economy Grid Architecture for Service-Oriented Grid Computing, Proceedings of the International Parallel and Distributed Processing Symposium: 10 th IEEE International Heterogeneous Computing Workshop (HCW 2001), April 23, 2001, San Francisco, California, USA, IEEE CS Press, USA, 2001. R. Buyya, D. Abramson, and J. Giddy, Nimrod-G: An Architecture for a Resource Management and Scheduling System in a Global Computational Grid, The 4 th International Conference on High Performance Computing in Asia-Pacific Region (HPC Asia 2000), May 2000, Beijing, China, IEEE Computer Society Press, USA. R. Buyya, J. Giddy, D. Abramson, An Evaluation of Economy-based Resource Trading and Scheduling on Computational Power Grids for Parameter Sweep Applications, Proceedings of the 2 nd International Workshop on Active Middleware Services (AMS 2000), Kluwer Academic Press, August 1, 2000, Pittsburgh, USA. R. Buyya, M. Murshed, and D. Abramson, A Deadline and Budget Constrained Cost-Time Optimization Algorithm for Scheduling Task Farming Applications on Global Grids, Technical Report, Monash University, March 2002. http: //www. buyya. com/gridsim/ Rajkumar Buyya and Manzur Murshed, Grid. Sim: A Toolkit for the Modeling and Simulation of Distributed Resource Management and Scheduling for Grid Computing, The Journal of Concurrency and Computation: Practice and Experience (CCPE), Volume 14, Issue 13 -15, Wiley Press, Nov. -Dec. , 2002. 32