The document discusses meeting deadlines for scientific workflows running on public clouds. It proposes replicating tasks to utilize idle resources and budgets to minimize workflow execution time while still meeting deadlines. The proposed approach models workflows as directed acyclic graphs and replicates tasks to reduce the impact of performance variations in public cloud resources. Existing work focuses on either minimizing time ignoring deadlines and budgets or minimizing cost while meeting deadlines.

1. Meeting Deadlines of Scientific
Workflows in
Public Clouds with Tasks Replication
S.R.MUGUNTHAN SUBMITTED BY
ASSISTANT PROFESSOR(SG)&HOD/CSE B.POORNIMA
MECSE II YEAR
08-09-2014 SVS COLLEGE OF ENGINEERING
1

2. Outline
• Objective
• General architecture of workflow system
• Issues in workflow
• Literature survey
• Proposed work
• System model
• Existing work
• References
08-09-2014 SVS COLLEGE OF ENGINEERING 2

3. Objective
• To reduce the impact of performance variation of public cloud
resources in the deadlines of workflow
• Deadline constrained workflow –Its delivers the result before
the deadline meets.
• To minimize the workflow execution time by ignoring
deadline and budgets.
• To use idle time of provisioned resources and budgets surplus
to replicate task.
08-09-2014 SVS COLLEGE OF ENGINEERING 3

4. General Architecture of Cloud
Workflow Systems
08-09-2014 SVS COLLEGE OF ENGINEERING 4

5. Issues in workflow
• Evaluate the performance of their implementations.
• Extremely valuable for the development and comparison of
workflow management systems.
• Characterizations of five scientific workflows:
Montage: astronomy
CyberShake: earthquake science
Epigenomics: biology
LIGO Inspiral AnalysisWorkflow: gravitational physics
SIPHT: biology
08-09-2014 SVS COLLEGE OF ENGINEERING 5

6. Literature Support
• Deadline-constrained workflow scheduling algorithms
for Infrastructure as a Service Clouds(2013)
• Multiple QoS Constrained Scheduling Strategy of
MultipleWorkflows for Cloud Computing(2009)
08-09-2014 SVS COLLEGE OF ENGINEERING 6

7. Deadline-constrained workflow scheduling algorithms
for Infrastructure as a Service Clouds(2013)
• In this paper use PCP algorithm for the Cloud environment and
propose two workflow scheduling algorithms.
• Which aims to minimize the cost of workflow execution while
meeting a user defined deadline.
• One-phase algorithm which is called IaaS Cloud Partial
Critical Paths (IC-PCP)
• Two-phase algorithm which is called IaaS Cloud Partial
Critical Paths with Deadline Distribution (IC-PCPD)
08-09-2014 SVS COLLEGE OF ENGINEERING 7

8. The IC-PCP Scheduling Algorithm
1: procedure ScheduleWorkflow(G(T , E), D)
2: determine available computation services
3: add tentry, texit and their corresponding dependencies to G
4: compute EST (ti), EFT (ti) and LFT (ti) for each task in G
5: AST(tentry) ← 0, AST(texit ) ←D
6: mark tentry and texit as assigned
7: call AssignParents(texit )
8: end procedure
08-09-2014 SVS COLLEGE OF ENGINEERING 8

9. Deadline-constrained workflow scheduling algorithms
for Infrastructure as a Service Clouds(2013)
Advantage
• The new algorithms consider the main features of the
current commercial Clouds such as on-demand resource
provisioning, homogeneous networks, and the pay-as-you-
go pricing model.
Disadvantage
• In accuracy of the estimated execution and transmission
times.
08-09-2014 SVS COLLEGE OF ENGINEERING 9

10. Multiple QoS Constrained Scheduling Strategy of
Multiple Workflows for Cloud Computing(2009)
• In this paper introduce a Multiple QoS Constrained Scheduling
Strategy of Multi-Workflows (MQMW) to address the problem.
• The strategy started at any time and QoS requirements are taken into
account .
• First, cloud provides services for multi-users. So the scheduling
strategy must provide different QoS requirements to different users.
• Second, there will be many workflow instances on the cloud
platform at the same time.
08-09-2014 SVS COLLEGE OF ENGINEERING 10

11. Multiple QoS Constrained Scheduling Strategy of
Multiple Workflows for Cloud Computing(2009)
Advantage
• Used to develop multiple workflow with different QoS
requirements.
• Increase the effect of total makespan and cost of workflow
greatly.
Disadvantage
• QoS constrained not include the parameters of reliability and
availability to the workflow.
08-09-2014 SVS COLLEGE OF ENGINEERING 11

12. Proposed Work
• To increase the performances variation of the resources on the
softdeadline of workflow application, here use an algorithm
that uses idle time of provisioned resources.
• Its meet and reduces the total execution time of applications as
the budget available for replication increases.
• The workflow model is extensively applied in diverse areas
such as astronomy, bioinformatics, and physics.
08-09-2014 SVS COLLEGE OF ENGINEERING 12

13. Proposed Work(Cont..)
• Scientific workflows are described as direct acyclic graphs
(DAGs)whose nodes represent tasks and vertices represent
dependencies among tasks.
• To being able to schedule the workflow in such a way that it
completes before its deadline.
• The workflow scheduler needs an estimation and run time of
the applications.
08-09-2014 SVS COLLEGE OF ENGINEERING 13

14. System Model
• A scientific workflow application is modeled as a Direct
Acyclic Graph (DAG) G=(T,ET).
• Dependencies are denoted in the form of Ei,j=(ti,tj),ti,tj€ T.
• Task ti is a parent task of tj and tj is a child task of ti.
• Each workflow G has a soft deadline dl(G) associated to it.
• The problem addressed in this paper consists in the execution
of a workflow G in the cloud on or before dl(G).
• For this problem to be solved, two subproblems have to be
solved , namely provisioning and scheduling.
08-09-2014 SVS COLLEGE OF ENGINEERING 14

15. Existing work
• Existing research in execution of scientific workflows in
Clouds either try to minimize the workflow execution time
ignoring deadlines and budgets.
• Also focus on the minimization of cost while trying to meet
the application deadline.
08-09-2014 SVS COLLEGE OF ENGINEERING 15

16. References
• M. Xu, L. Cui, H. Wang, and Y. Bi, ‘‘AMultiple QoS
Constrained Scheduling Strategy of Multiple Workflows for
Cloud Computing,’’ in Proc. Int’l Symp. ISPA, 2009, pp. 629-
634.
• Saeid ,Mahmoud , Dick H.J,” Deadline-constrained workflow
scheduling algorithms for Infrastructure as a Service Clouds”
in proc. Journal In Future Generation Computer System
29(2013) 158-169.
• G. Juve, A. Chervenak, E. Deelman, S. Bharathi, G. Mehta,
and K. Vahi, ‘‘Characterizing and Profiling Scientific
Workflows,’’ Future Gener. Comput. Syst., vol. 29, no. 3, pp.
682-692, Mar. 2013
08-09-2014 SVS COLLEGE OF ENGINEERING 16

17. • J. Yu, R. Buyya, and K. Ramamohanarao, ‘‘Workflow
Scheduling Algorithms for Grid Computing,’’ in
Metaheuristics for Scheduling in Distributed Computing
Environments, F. Xhafa and A.Abraham, Eds. New York, NY,
USA: Springer-Verlag, 2008
• Y.-K. Kwok and I. Ahmad, ‘‘Static Scheduling Algorithms
for Allocating Directed Task Graphs to Multiprocessors,’’
ACM Comput. Surveys, vol. 31, no. 4, pp. 406-471, Dec.
1999.
08-09-2014 SVS COLLEGE OF ENGINEERING 17

18. Thank You
08-09-2014 SVS COLLEGE OF ENGINEERING 18