ArticlePDF Available

DCHEFT Approach for Task Scheduling to Efficient Resource Allocation in Cloud Computing

Authors:

Abstract and Figures

Task scheduling is an important aspect to improve the utilization of resources in the Cloud Computing. This paper proposes a Divide and Conquer based approach for heterogeneous earliest finish time algorithm. The proposed system works in two phases. In the first phase it assigns the ranks to the incoming tasks with respect to size of it. In the second phase, we properly assign and manage the task to the virtual machine with the consideration of ideal time of respective virtual machine. This helps to get more effective resource utilization in Cloud Computing. The experimental results using Cybershake Scientific Workflow shows that the proposed Divide and Conquer HEFT performs better than HEFT in terms of task's finish time and response time. The result obtained by experimentally demonstrate that the proposed DCHEFT performance superiorly.
Content may be subject to copyright.
International Journal of Engineering and Applied Computer Science (IJEACS)
Volume: 02, Issue: 09, ISBN: 978-1-78808-411-6, November 2017
DOI:10.24032/ijeacs/0209/01
www.ijeacs.com
272
DCHEFT Approach for Task Scheduling to Efficient
Resource Allocation in Cloud Computing
Mahendra Bhatu Gawali
Department of IT, Thadomal Shahani Engineering
College, Bandra(W), University of Mumbai, Mumbai,
MS, India.
Subhash K. Shinde
CSE Dept, Lokmanya Tilak College of Engg.
Koparkhairane, Navi Mumbai
University of Mumbai, India.
Abstract- Task scheduling is an important aspect to improve the
utilization of resources in the Cloud Computing. This paper
proposes a Divide and Conquer based approach for
heterogeneous earliest finish time algorithm. The proposed
system works in two phases. In the first phase it assigns the ranks
to the incoming tasks with respect to size of it. In the second
phase, we properly assign and manage the task to the virtual
machine with the consideration of ideal time of respective virtual
machine. This helps to get more effective resource utilization in
Cloud Computing. The experimental results using Cybershake
Scientific Workflow shows that the proposed Divide and Conquer
HEFT performs better than HEFT in terms of task's finish time
and response time. The result obtained by experimentally
demonstrate that the proposed DCHEFT performance
superiorly.
Keywords- Cloud Computing; Task Scheduling; Resource
Allocation; Divide and Conquer; HEFT.
I. INTRODUCTION
The Cloud Computing is an emergent technology now a days,
which can support executing computationally heavy weight
application to infrastructure capital issues. The Cloud
Computing has overcome on the traditional way to deliver
service offline. The Cloud can serve the customer by
providing various types of services like SaaS, PaaS, IaaS [1].
The popularity of Cloud Computing has continuously
increasing with respect to advanced technology. Both
academia and industry are shifting their traditional
infrastructure setup to Cloud and being started to provide the
services in public, private and hybrid type of Cloud. Now
days, there are number of service providing companies are
available in the technology market which offers the Cloud
Services [2,3,4]. To choose the service provider is completely
depends upon the type of service user wants. As per the
requirement the user takes a decision about certain services to
be taken from certain service provider. More or less this
condition may applicable to all the Cloud Service User and
Cloud Service Provider. In the general different users may
submit their respective different task with various demands of
resources. To fulfil the demands of every user request with the
available resources is the challenge for Cloud Computing. It
may also effects on the performance of the Cloud Computing.
The objective function of this problem is to schedule the tasks
on resources and manage their executions so that execution
time for the task will decrease and resources also utilized
properly. Basically, task scheduling is quite complex and
important aspect. In this work we extend the heterogeneous
earliest finish time [5] methodology by introducing the divide
and conquer approach in it. Google Compute Engine used the
Cron tool for scheduling the tasks [6]. The major contribution
of this paper summarized as follows.
1) We modify the ranking algorithm which will assign the
ranks to the incoming Cloud requests.
2) Divide and Conquer methodology has been added into
heterogeneous earliest finish time for scheduling.
3) The experimental performance of the proposed solution
using Cloud Simulator.
The remainder of this paper is organized as follows. The
section II will brief about the state-of-the- art in the Cloud
Computing. The task-scheduling problem is described in
section III. The section IV explains the divide and conquers
approach for heterogeneous earliest finish time. The section V
illustrates the experimental setup of the proposed DCHEFT in
addition with this its gives performance evaluation with
existing algorithms. Finally, the section VI concludes the
paper.
II. LITERATURE SURVEY
This section will brief about the state-of-the-art in the
various algorithms used to solve task scheduling issues in the
Cloud Computing.
Liu et al. have been designed a model for a programming,
which utilized the large scale data intensive batch applications
[7]. It can specify the data partitioning and the computation
task distribution, while the complexity of parallel programming
is hidden. Fallenbeck et al. present a dynamic approach to
create virtual clusters to deal with the conflict between parallel
and serial jobs [8]. In this approach, the job load is adjusted
automatically without running time prediction.
Mahendra Bhatu Gawali et al.
International Journal of Engineering and Applied Computer Science (IJEACS)
Volume: 02, Issue: 09, ISBN: 978-1-78808-411-6, November 2017
DOI:10.24032/ijeacs/0209/01
www.ijeacs.com
273
Wilde et al. proposed Swift, a scripting language for
distributed computing [9]. Swift focuses on the concurrent
execution, composition, and coordination of large scale
independent computational tasks. A workload balancing
mechanism with adaptive scheduling algorithms is
implemented in Swift, based on the availability of resources. A
dynamic scoring system is designed to provide an empirically
measured estimate of a site‟s ability to bear load, which is
similar to the feedback information mechanism proposed in our
design. However, the score in the Swift is decreased only when
the site fails to execute the job.
Junjie proposed a load balancing algorithm [10] for the
private Cloud using virtual machine to physical machine
mapping. The architecture of the algorithm contains a central
scheduling controller and a resource monitor. The scheduling
controller does all the work for calculating which resource is
able to take the task and then assigning the task to that specific
resource. Ren [11] presented a dynamic load balancing
algorithm for cloud computing based on an existing algorithm
called WLC (Weighted Least Connection). The WLC
algorithm assigns tasks to the node based on the number of
connections that exist for that node. This is done based on a
comparison of the SUM of connections of each node in the
Cloud and then the task is assigned to the node with least
number of connections. However, WLC does not take into
consideration the capabilities of each node such as processing
speed, storage capacity and bandwidth.
The paper in [12] proposes an algorithm called Load Balancing
Min-Min (LBMM). LBMM has a three level load balancing
framework. It uses the Opportunistic Load Balancing algorithm
(OLB) [13]. OLB is a static load balancing algorithm that has
the goal of keeping each node in the cloud busy. However,
OLB does not consider the execution time of the node. This
might cause the tasks to be processed in a slower manner and
will cause some bottlenecks since requests might be pending
waiting for nodes to be free. LBMM improves OLB by adding
a three layered architecture to the algorithm. The first level of
the LBMM architecture is the request manager which is
responsible for receiving the task.
III. TASK SCHEDULING PROBLEM
The Cloud Computing consists of various size tasks, a
collection of interconnected high end resources and a criterion
for a performance for scheduling. For this we have taken the
Cybershake Scientific Workflow [14] tasks as an input for
Cloud Computing system. Table 1 will elaborates the
Seismogram Synthesis tasks with its actual weight and
expected execution time. These tasks are computationally
heavy to execute. Especially, seismogram synthesis tasks are
consuming lot of computing resources to execute.
This application is represented by a directed acyclic graph,
G=(V,E), where V is the set of v tasks and E is the set of e
edges between tasks. This application's graph has bounded
with control-flow dependency constraint such that task ti
should complete its execution before task tj . In a given
application task graph if a task without any parent is call an
entry task and without having any child task is called as an exit
task.
We assume that the Cloud Computing data center consist of a
set of virtual machines configured by various computing
resources such as CPU, memory, bandwidth etc.
We assumed that eti,j gives the estimated execution time to
complete the task ti on Vmj .
When both ti and tj are scheduled on the same vitual
machine.
The various size of tasks are taking different time to complete
its execution on different virtual machine. So, let we describe
the task's earliest start time and earliest finish task. EST (ti ,
VMj ) and EFT (ti , VMj ) are the earliest start time of task on
VM and earliest finish time of task on VM respectively. The
earliest start time for entry task as,
EST(tentry, Vmj) = 0. (1)
For the next tasks in the application, the EST and EFT values
are computed in consideration of equ. (1). the subsequent
task's start time and finish time have been calculated by equ
(2) and equ. (3) respectively.
EST(ti , Vmj) = max{free[j], max (AFT(tm) )} (2)
EFT(ti , Vmj) = eti,j + EST (ti , Vmj) (3)
The scheduling policy must reduce the length of the waiting
queue of tasks for getting the resources. This will possible by
solving the objective equ . (4)
makespan = max (AFT(texit)) (4)
This objective function we achieve by our proposed system.
TABLE 1. CYBERSHAKE SEISMOGRAM SYNTHESIS TASKS
Size of Tasks (MB)
Time
62,69,51,663
39.06
69,47,76,323
38.49
58,57,63,637
36.27
53,68,97,326
32.29
67,05,35,542
62.25
40,67,28,38,798
96.91
45,23,96,996
45.60
50,27,64,231
28.67
62,41,88,532
24.56
42,65,77,006
31.05
51,58,32,878
54.87
68,14,99,417
23.99
44,14,51,516
26.46
IV. PROPOSED DCHEFT
We have developed DCHEFT for Cybershake Scientific
Workflow. The fig. 1 explains the architecture of proposed
DCHEFT with its mandatory components. Basically, the
proposed system has been divided into two main parts. Those
are ranking the incoming user's requests (task) and assign that
to the resources to minimize the finish time as well as
makespan.
The first part completely process the task before it actually
assign to the Cloud Computing resources. On the basis of
task's estimated execution time, its size and its control flow
dependency we assign the rank to every individual task.
Mahendra Bhatu Gawali et al.
International Journal of Engineering and Applied Computer Science (IJEACS)
Volume: 02, Issue: 09, ISBN: 978-1-78808-411-6, November 2017
DOI:10.24032/ijeacs/0209/01
www.ijeacs.com
274
We pass the ranked task for processing in waiting queue. As
soon as the virtual machine will free in an order to that we
assign the task to respective virtual machine. This has been
worked out in part two of proposed system architecture.
Figure 1. Proposed DCHEFT system architecture
A. Ranking the tasks by HEFT
Tasks are order by HEFT algorithm by assigning the first and
last ranking. The first rank of a task ti is calculated by equ. (5).
rankf (ti)= wi + maxtj e succ (ti) (ci,j + rankf (tj)) (5)
where, succ(ti) is the successors of task ti , ci,j is the average
communication cost of edge (i, j) and wi is the average
computation of task ti . The first task's which is ready to
schedule on virtual machine its first rank is equal to
rankf (texit)= wexit (6)
The last rank of a task ti is calculated as follows.
rankl (ti)= maxtj e predc (ti) (rankl (tj) + wj + ci,j ) (7)
where, pred(ti) is the predecessor of task ti . The last ranks are
computed recursively by traversing the task graph towards the
last task of an application which has been started from the first
task of the graph.
B. Proposed Divide and Conquer approach
In the HEFT the second phase is to schedule the ranked task
to virtual machine. While allocating the tasks to the virtual
machine the HEFT has considered the idle time of virtual
machines. An idle time is the difference between execution
start time and finish time of two tasks that were scheduled on
the same virtual machine.
However HEFT has some limitations. HEFT algorithm search
for an idle time and if idle time is less than the scheduled
task's execution time then task must have to wait until next
idle time. This HEFT affected on the waiting time of tasks.
This is the major motivation behind the work presented in this
paper. The proposed divide and conquer based HEFT
algorithm initially finds the idle time and schedule the task
without consideration whether idle time is less than task's
execution time. Fig. 2 explains the detail flow of the proposed
DCHEFT system.
Figure 2. Proposed System Flowchart
Algorithm: DCHEFT
Input- Task‟s Execution Time, Task ti
Output- Task‟s Output
1: Start
2: ti‟s Execution time is „x‟ then
3: If(x< idle_time_of_virtual_machine)
4: Assign „x‟ to that Virtual_Machine;
5: Else
6: Divide „x‟;
7: If (First part „x‟ <= idle_time)
8: Assign „x‟ to Virtual_machine;
9: Else
Wait for Ideal-Time;
10: For second part to end of the part of task
11: Do
12: find-out the idle-time of VM and
assign to the task;
13: enddo
14: endfor
15: Aggregate the execution of x= x`+x``+…+x`n;
16: End
Mahendra Bhatu Gawali et al.
International Journal of Engineering and Applied Computer Science (IJEACS)
Volume: 02, Issue: 09, ISBN: 978-1-78808-411-6, November 2017
DOI:10.24032/ijeacs/0209/01
www.ijeacs.com
275
V. EXPERIMENTAL SETUP
The proposed DCHEFT approach work is experimented on
Cloud Simulator [15], which gives the real-time environment
scenario of Cloud Computing. Datacenter Information has
been listed in Table 2. Tables 3 consist of configuration for
Datacenter which includes allocation policy, architecture, OS,
hyper visor, scheduling and monitoring interval, threshold
value etc. Host in the Datacenter used to show the amount of
provisional RAM, bandwidth, storage capacity, power,
processing element etc. of given task which process by
Datacenter . Table 4 explains the host configuration details.
Configuration details of customized simulation setup are given
in Table 5 and it consist of general information of Datacenters
like number of Datacenters, number of host, number of
processing units, capacity etc. Every Datacenter component
instantiates a generalized application provisioning component
that implement a set of policies for allocating bandwidth,
memory and storage devices to hosts and virtual machines.
Table 6 holds information related to storage area network
capacity, latency and bandwidth.
TABLE 2: DATACENTER INFORMATION
Sr. No.
Information
Contains
1
Number of Datacenter
1
2
Number of Host
1
3
Number of Processing Units
4
4
Processing capacity (MIPS)
9600
5
Storage Capacity
11 TB
6
Total Amount of RAM
40 GB
TABLE 3: DATACENTER CONFIGURATION DETAILS
Sr. No.
Information
Contains
1
Allocation Policy
SDMCOA
2
Architecture
X86
3
Operating system
Linux
4
Hypervisor
Xen
5
Upper threshold
0.8
6
Lower threshold
0.2
7
VM Migration
Enabled
8
Monitoring Interval
180
TABLE 4: HOST CONFIGURATION DETAILS
Sr. No.
Information
Contains
1
RAM
40 GB
2
Bandwidth
10,00,000
3
Operating System
Linux
4
Hypervisor
Xen
TABLE 5: CUSTOMER CONFIGURATION DETAILS
Sr. No.
Information
Contains
1
Users
1
2
Cloudlets sent per minutes
50
3
Avg. Length of Cloudlet
50,000
4
Avg. Cloudlet file Size
500 Bytes
5
Avg. Cloudlet output size
500 Bytes
TABLE 6: CUSTOMER CONFIGURATION DETAILS
Sr. No.
Information
Contains
1
Number of VMs
20
2
Avg. Image Size
1000 Bytes
3
Avg. RAM
512 MB
4
Avg. Bandwidth
1,00,000 Mbps
5
Procedure Element
1
6
Priority
1
7
Scheduling Priority
Dynamic Workload
VI. RESULT AND DISCUSSION
This section will brief about the performance of proposed
novel DCHEFT approach.
Let, we evaluate our proposed DCHEFT approach with
existing BATS [16] and Heuristic Approach [17], SDMCOA
[18] on the given Cybershake Seismogram Synthesis tasks.
Evaluation of proposed DCHEFT system is based on two key
factors i.e. turnaround time and response time.
A. Evaluation of Turn Around Time
This is one of the major performance factor to check the
evaluation of the system. Basically, it is the span of total time
taken between the submission of a request (task) for execution
to the complete the same. Specifically, turnaround time is
based up on the programming/ software logic. We compare
our proposed DCHEFT system with BATS, Heuristic and
SDMCOA. We found that our system is works fine as
compared with existing systems which has been shown result
in Table 7 and Fig. 3.
TABLE 7: TURN AROUND TIME COMPARISON IN MS
Tasks
DCHEFT
SDMCOA
Heuristic
BATS
Task 3
2405.79
2613.79
2832.94
3599.29
Task 5
2405.07
2613.07
2914.42
3599.29
Task 7
2403.02
2611.02
2913.87
3599.29
Task 9
2398.93
2606.93
2911.75
3599.29
Task 11
2498.78
2636.78
2907.67
3599.29
Task 14
2348.36
2556.36
2772.11
3599.29
Task 16
2346.44
2554.44
2857.89
3599.29
Task 18
2329.48
2537.48
2855.97
3599.29
Task 20
2325.36
2533.36
2833.36
3599.29
Task 22
2332.06
2540.06
2834.72
3599.29
Task 24
2355.70
2563.70
2841.49
3599.29
Task 26
2324.86
2532.86
2832.86
3599.29
Task 28
2327.37
2535.37
2833.96
3599.29
Mahendra Bhatu Gawali et al.
International Journal of Engineering and Applied Computer Science (IJEACS)
Volume: 02, Issue: 09, ISBN: 978-1-78808-411-6, November 2017
DOI:10.24032/ijeacs/0209/01
www.ijeacs.com
276
Figure 3 TAT Comparison in ms
B. Evaluation of Response Time
This is another major performance factor to check the
evaluation of the system. Response time is the time taken from
the issuance of a task to the commence of a response to that
task. We compare our proposed DCHEFT system with BATS,
Heuristic and SDMCOA. We found that our system is works
fine as compared with existing systems which has been shown
result in Table 8 and Fig. 4.
TABLE 7: RESPONSE TIME COMPARISON IN MS
Tasks
DCHEFT
SDMCOA
Heuristic
BATS
Task 3
2.63
2.83
2.83
5.1
Task 5
2.63
2.83
2.91
5.1
Task 7
2.63
2.83
2.9
5.1
Task 9
2.59
2.83
2.91
5.1
Task 11
2.59
2.83
2.90
5.1
Task 14
2.63
2.77
2.77
5.1
Task 16
2.54
2.77
2.85
5.1
Task 18
2.54
2.77
2.85
5.1
Task 20
2.46
2.77
2.83
5.1
Task 22
2.46
2.77
2.83
5.1
Task 24
2.63
2.77
2.84
5.1
Task 26
2.63
2.77
2.83
5.1
Task 28
2.63
2.77
2.83
5.1
Figure 4 RT Comparison in ms
VII. CONCLUSION
This paper describes a proposed Divide and Conquer
Heterogeneous Earliest Finish Time Algorithm for task
scheduling to efficiently managed the resources in Cloud
Computing. To utilize the resources ideal time we used the
concept of Divide and Conquer. Because of this methodology
the task's waiting time is drastically reduce while the utilization
of resources have been increased which has been proved by
experimentally. The results from various simulations using
Cybershake Scientific Seismogram tasks as an input shows that
the DCHEFT approach performs better than SDMCOA,
Heuristic and BATS existing approaches. In future the tasks
waiting time needs to be reduced when the availability of
resources are less.
REFERENCES
[1] Zhang Qi, Lu Cheng, and Raouf Boutaba, "Cloud computing: state-of-
the-art and research challenges", Journal of internet services and
applications, no. 1 (2010): pp 7-18.J. Clerk Maxwell, A Treatise on
Electricity and Magnetism, 3rd ed., vol. 2. Oxford: Clarendon, 1892,
pp.6873.
[2] Amazon AWS, http://aws.amazon.com/. Accessed online 02 June 2017.
K. Elissa, “Title of paper if known,” unpublished.
[3] Microsoft cloud, http://www.microsoft.com/en-us/cloud/.Accessed
online 02 June 2017..
[4] IBM cloud, http://www.ibm.com/ibm/cloud/. Accessed online 02 June
2017..
[5] Copcuoglu, Haluk, Salim Hariri, and Min-you Wu. "Performance-
effective and low-complexity task scheduling for heterogeneous
computing", IEEE transactions on parallel and distributed systems 13.3
T(2002): 260-274.
[6] Google Compute Engine, https://cloud.google.com/solutions/reliable-
task-scheduling-compute-engine. Accessed Online 04 June 2017.
[7] H. Liu, D. Orban, GridBatch: cloud computing for large-scale data-
intensive batch applications, in: IEEE International Symposium on
Cluster Computing and the Grid, pp. 295305, 2008.
Turnaround time in ms
Tasks
Response time in ms
Tasks
Mahendra Bhatu Gawali et al.
International Journal of Engineering and Applied Computer Science (IJEACS)
Volume: 02, Issue: 09, ISBN: 978-1-78808-411-6, November 2017
DOI:10.24032/ijeacs/0209/01
www.ijeacs.com
277
[8] Niels Fallenbeck, Hans-Joachim Picht, Matthew Smith, and Bernd
Freisleben. 2006. Xen and the Art of Cluster Scheduling. In Proceedings
of the 2nd International Workshop on Virtualization Technology in
Distributed Computing (VTDC '06). IEEE Computer Society,
Washington, DC, USA, 4-.
DOI=http://dx.doi.org/10.1109/VTDC.2006.18.
[9] M. Wilde, M. Hategan, J.M. Wozniak, B. Clifford, D.S. Katz, I. Foster,
Swift: a language for distributed parallel scripting, Parallel Computing
37 (9) (2011) 633652.
[10] Ni, J., Y. Huang, Z. Luan, J. Zhang and D. Qian, "Virtual machine
mapping policy based on load balancing in private cloud environment,"
in proc. International Conference on Cloud and Service Computing
(CSC), IEEE, pp: 292-295, December 2011.
[11] Ren, X., R. Lin and H. Zou, "A dynamic load balancing strategy for
cloud computing platform based on exponential smoothing forecast" in
proc. International Conference on. Cloud Computing and Intelligent
Systems (CCIS), IEEE, pp: 220-224, September 2011.
[12] Wang, S-C., K-Q. Yan, W-P. Liao and S-S. Wang, "Towards a load
balancing in a three-level cloud computing network", in proc. 3rd
International Conference on. Computer Science and Information
Technology (ICCSIT), IEEE, Vol. 1, pp:108-113, July 2010.
[13] Sang, A., X. Wang, M. Madihian and RD. Gitlin, "Coordinated load
balancing, handoff/cell-site selection, and scheduling in multi-cell
packet data systems," in Wireless Networks, Vol. 14, No. 1, pp: 103-
120, January 2008.
[14] Pegasus Scientific Workflow.
https://confluence.pegasus.isi.edu/display/pegasus/ WorkflowGenerator.
Access Online. 04 June 2017.
[15] Calheiros, R.N., Ranjan, R., Beloglazov, A., De Rose, C.A. and Buyya,
R., 2011. CloudSim: a toolkit for modeling and simulation of cloud
computing environments and evaluation of resource provisioning
algorithms. Software: Practice and experience, 41(1), pp.23-50, 2011.
[16] Mahendra Bhatu Gawali and Subhash K. Shinde. "Implementation of
IDEA, BATS, ARIMA and queuing model for task scheduling in cloud
computing." In Eco-friendly Computing and Communication Systems
(ICECCS), 2016 Fifth International Conference on, pp. 7-12. IEEE,
2016. DOI: 10.1109/Eco-friendly.2016.7893233.
[17] Mahendra Bhatu Gawali and Subhash K. Shinde (Under Review) “Task
Scheduling and Resource Allocation in Cloud Computing by Heuristic
Approach”. Springer, Journal of Cloud Computing, Advances, Systems
and Applications. DOI 10.1186/s13677-017.
[18] Mahendra Bhatu Gawali and Subhash K. Shinde, "Standard Deviation
Based Modified Cuckoo Optimization Algorithm for Task Scheduling to
Efficient Resource Allocation in Cloud Computing", Journal of
Advances in Information Technology, Vol. 8, No. 4, pp. 210-218,
November, 2017. doi: 10.12720/jait.8.4.210-218.
AUTHOR PROFILE
Mahendra Bhatu Gawali received his BE degree
in 2008 and M.E. degree in 2013 from North
Maharashtra University, Jalgaon, MS, India.
Currently he is pursuing his Ph.D. at Thadomal
Shahani Engineering College, Bandra(W),
University of Mumbai, Mumbai, India. He
focuses on Task Scheduling and Resource
Allocation in Cloud Computing.
Subhash K. Shinde is working as a Professor in
the Department of Computer Engineering at
Lokmanya Tilak College of Engineering, Navi
Mumbai, India. He received his Ph.D. from
Swami Ramanand Teertha Marathwada
University, India in 2012. He has published more
than 40 research papers in the field of Web
Mining, Frequent Pattern Discovery and
Integration of domain knowledge in web
personalized recommendations in the reputed
journals and conferences.
© 2017 by the author(s); licensee Empirical Research Press Ltd. United Kingdom. [In association with
Independent Publishing Network, U.K.]. This is an open access article distributed under the terms and conditions
of the Creative Commons by Attribution (CC-BY) license. (http://creativecommons.org/licenses/by/4.0/).
ResearchGate has not been able to resolve any citations for this publication.
Article
Full-text available
We investigate a wireless system of multiple cells, each having a downlink shared channel in support of high-speed packet data services. In practice, such a system consists of hierarchically organized entities including a central server, Base Stations (BSs), and Mobile Stations (MSs). Our goal is to improve global resource utilization and reduce regional congestion given asymmetric arrivals and departures of mobile users, a goal requiring load balancing among multiple cells. For this purpose, we propose a scalable cross-layer framework to coordinate packet-level scheduling, call-level cell-site selection and handoff, and system-level cell coverage based on load, throughput, and channel measurements. In this framework, an opportunistic scheduling algorithm—the weighted Alpha-Rule—exploits the gain of multiuser diversity in each cell independently, trading aggregate (mean) downlink throughput for fairness and minimum rate guarantees among MSs. Each MS adapts to its channel dynamics and the load fluctuations in neighboring cells, in accordance with MSs’ mobility or their arrival and departure, by initiating load-aware handoff and cell-site selection. The central server adjusts schedulers of all cells to coordinate their coverage by prompting cell breathing or distributed MS handoffs. Across the whole system, BSs and MSs constantly monitor their load, throughput, or channel quality in order to facilitate the overall system coordination. Our specific contributions in such a framework are highlighted by the minimum-rate guaranteed weighted Alpha-Rule scheduling, the load-aware MS handoff/cell-site selection, and the Media Access Control (MAC)-layer cell breathing. Our evaluations show that the proposed framework can improve global resource utilization and load balancing, resulting in a smaller blocking rate of MS arrivals without extra resources while the aggregate throughput remains roughly the same or improved at the hot-spots. Our simulation tests also show that the coordinated system is robust to dynamic load fluctuations and is scalable to both the system dimension and the size of MS population.
Article
Full-text available
Cloud computing has recently emerged as a new paradigm for hosting and delivering services over the Internet. Cloud computing is attractive to business owners as it eliminates the requirement for users to plan ahead for provisioning, and allows enterprises to start from the small and increase resources only when there is a rise in service demand. However, despite the fact that cloud computing offers huge opportunities to the IT industry, the development of cloud computing technology is currently at its infancy, with many issues still to be addressed. In this paper, we present a survey of cloud computing, highlighting its key concepts, architectural principles, state-of-the-art implementation as well as research challenges. The aim of this paper is to provide a better understanding of the design challenges of cloud computing and identify important research directions in this increasingly important area. KeywordsCloud computing-Data centers-Virtualization
Conference Paper
Full-text available
In shared use clusters, scheduling systems must schedule both serial and parallel jobs in a fair manner, while at the same time optimizing overall cluster efficiency. Since serial and parallel jobs conflict considerably, scheduling both types of jobs concurrently is a difficult task. Two major strategies are in common use: partitioning the cluster (thus avoiding the problem) and reservation combined with backfilling. Both have major drawbacks in overall performance, ease of use and fairness depending on the particular configuration, which can lead to heated debates between the users of the different types of jobs. In this paper, we introduce an approach to create dynamic virtual cluster partitions using para-virtualization techniques, to deal with the conflicts between parallel and serial jobs. The system dynamically adjusts to different types of job loads and offers easy and transparent use and configuration to both users and administrators, without resorting to complicated runtime prediction or backfilling algorithms. A proof-of-concept implementation based on the Sun Grid Engine scheduling system and Xen is presented.
Conference Paper
Network bandwidth and hardware technology are developing rapidly, resulting in the vigorous development of the Internet. A new concept, cloud computing, uses low-power hosts to achieve high reliability. The cloud computing, an Internet-based development in which dynamically scalable and often virtualized resources are provided as a service over the Internet has become a significant issue. The cloud computing refers to a class of systems and applications that employ distributed resources to perform a function in a decentralized manner. Cloud computing is to utilize the computing resources (service nodes) on the network to facilitate the execution of complicated tasks that require large-scale computation. Thus, the selecting nodes for executing a task in the cloud computing must be considered, and to exploit the effectiveness of the resources, they have to be properly selected according to the properties of the task. However, in this study, a two-phase scheduling algorithm under a three-level cloud computing network is advanced. The proposed scheduling algorithm combines OLB (Opportunistic Load Balancing) and LBMM (Load Balance Min-Min) scheduling algorithms that can utilize more better executing efficiency and maintain the load balancing of system.
Article
Cloud computing is a recent advancement wherein IT infrastructure and applications are provided as ‘services’ to end-users under a usage-based payment model. It can leverage virtualized services even on the fly based on requirements (workload patterns and QoS) varying with time. The application services hosted under Cloud computing model have complex provisioning, composition, configuration, and deployment requirements. Evaluating the performance of Cloud provisioning policies, application workload models, and resources performance models in a repeatable manner under varying system and user configurations and requirements is difficult to achieve. To overcome this challenge, we propose CloudSim: an extensible simulation toolkit that enables modeling and simulation of Cloud computing systems and application provisioning environments. The CloudSim toolkit supports both system and behavior modeling of Cloud system components such as data centers, virtual machines (VMs) and resource provisioning policies. It implements generic application provisioning techniques that can be extended with ease and limited effort. Currently, it supports modeling and simulation of Cloud computing environments consisting of both single and inter-networked clouds (federation of clouds). Moreover, it exposes custom interfaces for implementing policies and provisioning techniques for allocation of VMs under inter-networked Cloud computing scenarios. Several researchers from organizations, such as HP Labs in U.S.A., are using CloudSim in their investigation on Cloud resource provisioning and energy-efficient management of data center resources. The usefulness of CloudSim is demonstrated by a case study involving dynamic provisioning of application services in the hybrid federated clouds environment. The result of this case study proves that the federated Cloud computing model significantly improves the application QoS requirements under fluctuating resource and service demand patterns.
Article
Scientists, engineers, and statisticians must execute domain-specific application programs many times on large collections of file-based data. This activity requires complex orchestration and data management as data is passed to, from, and among application invocations. Distributed and parallel computing resources can accelerate such processing, but their use further increases programming complexity. The Swift parallel scripting language reduces these complexities by making file system structures accessible via language constructs and by allowing ordinary application programs to be composed into powerful parallel scripts that can efficiently utilize parallel and distributed resources. We present Swift’s implicitly parallel and deterministic programming model, which applies external applications to file collections using a functional style that abstracts and simplifies distributed parallel execution.
Conference Paper
To be competitive, enterprises are collecting and analyzing increasingly large amount of data in order to derive business insights. However, there are at least two challenges to meet the increasing demand. First, the growth in the amount of data far outpaces the computation power growth of a uniprocessor. The growing gap between the supply and demand of computation power forces Enterprises to parallelize their application code. Unfortunately, parallel programming is both time-consuming and error-prone. Second, the emerging Cloud Computing paradigm imposes constraints on the underlying infrastructure, which forces enterprises to rethink their application architecture. We propose the GridBatch system, which aims at solving large-scale data-intensive batch problems under the Cloud infrastructure constraints. GridBatch is a programming model and associated library that hides the complexity of parallel programming, yet it gives the users complete control on how data are partitioned and how computation is distributed so that applications can have the highest performance possible. Through a real client example, we show that GridBatch achieves high performance in Amazon's EC2 computing Cloud.