International Journal of Emerging Technology in Computer Science & Electronics (IJETCSE)
ISSN: 0976-1353
Volume 13 Issue 1 –MARCH 2015.
Abstract— Cloud computing is a subscription-based service where a network storage space and computation resources can be obtained. IaaS(Infrastructure as a Service) has become an important paradigm in cloud computing, it achieves economy of scale by multiplexing and therefore faces the challenge of scheduling task. The proposed scheduling approach, based on ACO (Ant Colony Optimization) schedules task from private cloud to external cloud when there is a peak demand. The system is modeled as a framework where a provider outsources its tasks to external cloud (Ecs) to meet the unpredictable situation in cloud environment. The approaches so far used rely on inter-cloud agreement that is required for the cloud federation. The proposed scheduler allocates the task to resource and adopts dynamic change in the cloud environment without any standardization also it provides optimized scheduling by which the factors of a good scheduler such as high throughput and reduced response time can be achieved and solves the issue of portability while migrating to different cloud. It was observed that the scheduling approach based on ACO yields better results and improves cloud provider’s profit by 0.26-11.58% compared with the existing methods.
Index Terms— Ant Colony Optimization (ACO), Hybrid
cloud, Infrastructure as a service (IaaS), task scheduling.
I. INTRODUCTION
Cloud Computing provides scalable resources dynamically as a service over internet in order to assure lots of monetary benefits to be scattered among its adopters. According to Gartner cloud computing is defined as “ a style
of computing in which scalable and elastic IT-enabled capabilities are delivered as a service to external customers using Internet technologies.”According to Forrester cloud
computing is defined as “a standardized IT capability
(services, software, or infrastructure) delivered via internet technologies in a pay-per-use, self-service way.”Cloud-based services integrate globally distributed
resources into seamless computing platforms.
Infrastructure as a Service (IaaS) becomes very popular as the foundation for higher level services such as Platform as a Service (PaaS) and Software as a Service (SaaS) [1]. IaaS providers such as Amazon EC2 and IBM Smart Cloud Enterprise [2], allow users to rent resources in the form of Virtual Machines (VMs). They can offer different VM types that are characterized by machine configuration, QoS[10] and pricing model.
A straightforward solution for a cloud provider is to over purchase the cloud resources such as memory,
bandwidth in advance, which is not cost-efficient [4]. An-other solution is Cloud Federation [3] that allows providers to trade their resources through federation agreement.
In this paradigm, providers can overcome their resource limitation by out-sourcing requests to other members in the federation. However, this federation is not easy to achieve at present, due to the lack of inter-operation standard and players’ motivation to federate [5].
To make an IaaS cloud itself elastic, A cloud resources allocation framework is proposed to allow it to utilize external clouds.
In this framework, an IaaS cloud has its own private cloud, and is able to outsource its tasks to other cloud providers called external clouds (ECs) when its local resources are not sufficient. An integer programming formulation is established, with the objective of maximizing the profit of the private cloud while at the same time guaranteeing QoS[10] to the user.
II. CHALLENGES AND REQUIREMENTS
There exists no generic model to represent various scenarios of task scheduling[16], especially when user’s requirements are vague and hard to encode through modeling languages. In particular, mapping QoS[10] requirements of applications to fine-grained resource level attributes is difficult. Modeling and quantifying non-functional requirements such as availability, is challenging.
Model parameterization i.e., finding suitable values for parameters in a proposed model, is a tedious task when the problem size is large. for example , for a multi-cloud scenario that includes n cloud providers and m VMs, m*n2 parameter assignments are needed in principle to express the VM migration overheads ignoring possible changes of VM sizes. Therefore, mechanisms that can help automatically capture those values are required.
The initial VM placement problem is typically formulated as a variant of the class constrained multiple-knapsack problem that is known to be NP-hard [16]. Thus to solve large-scale problem instances, tradeoffs between quality of solutions and execution time must be taken into account. This is a very important issue given the size of real life data centers.
Conflicting objectives. On energy-efficient scheduling, existing work focuses on certain aspects of QoS[10],
ACO Based Task Scheduling Algorithm for
Hybrid Cloud
M.Anuradha
#1and S.Selvakumar
*2#
PG Scholar, Computer Science, GKM College of Engineering and Technology, Chennai, India.
International Journal of Emerging Technology in Computer Science & Electronics (IJETCSE)
ISSN: 0976
however they commonly overlook the energy aspect that may conflict with other QoS requirements
Continuous optimization[16]. Given the dynamic nature of clouds, resource allocation need to be renewed regularly for performance reasons, failure, etc., It is challenging to efficiently decide when and how to reconfigure the cloud in order to dynamically adapt to the changes.
III. SYSTEM ARCHITECTURE
Consumer layer is the point where the customers interact
with the cloud service provider. The cloud user considered in the proposed system is thick client. In addition to that there can be a thin client and mobile client.
In general consumer layer comprises various cloud user.
Presentation layer allows to implement and design user
interface through which the cloud user’s interact with service provider it also allows the management of user inter which is required in systems where more than one user interact.
Fig1. Component view of proposed system
Business layer is the main partition of private cloud where
actual processing to achieve an objective will be carried out. It collects information from the user as request for performing business process. Interface layer
for the interaction with the public cloud. The interaction includes collecting pricing models from various external clouds.
The user interface allows number of use
cloud service provider. The users are of thin client, thick client or mobile client. They are supposed to submit their task which needs to be executed by utilizing the cloud resource. The users of cloud can be allowed to store their dat
and run their application without having any local resources or data centers.
The important issue is how to handle multiple request in a user session. The resource handler collects task from different cloud users and allows them to be scheduled execution and acts as a base for providing services in the cloud environment.
International Journal of Emerging Technology in Computer Science & Electronics (IJETCSE)
ISSN: 0976-1353
Volume 13 Issue 1 –MARCH 2015.
however they commonly overlook the energy-efficiency aspect that may conflict with other QoS requirements
Given the dynamic nature of clouds, resource allocation need to be renewed regularly for performance reasons, failure, etc., It is challenging to efficiently decide when and how to reconfigure the cloud in
adapt to the changes.
RCHITECTURE
is the point where the customers interact with the cloud service provider. The cloud user considered in the proposed system is thick client. In addition to that there
In general consumer layer comprises various cloud user. to implement and design user interface through which the cloud user’s interact with service provider it also allows the management of user interaction systems where more than one user
Fig1. Component view of proposed system
is the main partition of private cloud where actual processing to achieve an objective will be carried out. ion from the user as request for
Interface layer is responsible
for the interaction with the public cloud. The interaction includes collecting pricing models from various external allows number of users to interact with cloud service provider. The users are of thin client, thick client or mobile client. They are supposed to submit their task which needs to be executed by utilizing the cloud resource. The users of cloud can be allowed to store their data, develop and run their application without having any local resources The important issue is how to handle multiple request in a collects task from different cloud users and allows them to be scheduled for execution and acts as a base for providing services in the
The Scheduler component allows to allocate suitable resources to user task by collecting details about the resource availability. It performs effective scheduling with the available cloud resources and in case of peak demand the tasks are being outsourced to the external cloud where the data center is identified by means of pricing models.
Resource pool in cloud environment allows the cloud
service provider to serve multiple
resources of a cloud environment includes storage, bandwidth and cpu. By providing processing, storage and network components as a resource pool the cloud computing environment allows the consumer to feel free to manage and control those resources.
Resource monitor provides information to the scheduler
about the availability of resources. The scheduler interacts and collects data about resources for the effective utilization of resource so that no resource will be set as idle in the service provider.
The Cloud interface enables services to move between different providers and allows clients to easily switch between providers based on business objectives.
IV. SOLUTION FRAMEWORK
Consider[18] CP={CP1,CP2,…
providers. Assume CP1 is the private cloud and CP are external clouds. VM={VM
VM types and A={a1,a2,…,aw
required to be scheduled for each VM and each application has a runtime rj and a task set {t
An Integer programming model is formulated to solve this problem. The objective of integer programming formulation is to maximize the profit of private cloud on the premise of guaranteeing QOS. To formulate this problem, problem parameters and decision variables
Table II, respectively.
The problem can be formulated as the following IP model. Maximum Profit achieved by the cloud service provider is
∑ ∑ ∑ ∑
Subject to ∑ 1, ∀ 1,2, … ,
N Number of cloud providers I Number of VM types W Number of applications Pv Price of the vth VM
Ckv Cost of the vth VM type in CP
Rj Runtime of each task in the j Tj Number of tasks in the j
Bjv If bjv=1,the jth application use VM type
VMv;otherwise, it does not use this type.
Table I. Problem parameters
International Journal of Emerging Technology in Computer Science & Electronics (IJETCSE)
component allows to allocate suitable resources to user task by collecting details about the resource availability. It performs effective scheduling with the vailable cloud resources and in case of peak demand the tasks are being outsourced to the external cloud where the data center is identified by means of pricing models.
in cloud environment allows the cloud service provider to serve multiple cloud consumers. The resources of a cloud environment includes storage, bandwidth and cpu. By providing processing, storage and network components as a resource pool the cloud computing environment allows the consumer to feel free to manage and provides information to the scheduler about the availability of resources. The scheduler interacts and collects data about resources for the effective utilization of resource so that no resource will be set as idle in the enables services to move between different providers and allows clients to easily switch between providers based on business objectives.
SOLUTION FRAMEWORK
,…,CPn} be the set of cloud is the private cloud and CP2,…,CPn are external clouds. VM={VM1,VM2,…,VMI} be the set of
w} be the set of applications required to be scheduled for each VM and each application has a runtime rj and a task set {tj1,tj2,…,tjTj}.
An Integer programming model is formulated to solve this problem. The objective of integer programming formulation is to maximize the profit of private cloud on the premise of guaranteeing QOS. To formulate this problem, problem cision variables are defined in Table I and The problem can be formulated as the following IP model. Maximum Profit achieved by the cloud service provider is
∑ ∑ ∑
∈ 1,2, … , ! , ∈
Number of cloud providers Number of VM types Number of applications
VM type in CPk
Runtime of each task in the jth application
Number of tasks in the jth application
application use VM type ;otherwise, it does not use this type.
International Journal of Emerging Technology in Computer Science & Electronics (IJETCSE)
ISSN: 0976
Table II. Decision variables
The first term of objective function represents the income of CP1 and the second one means its cost
specified in the formulation guarantees that each task is allocated to exactly one cloud provider. The problem formulated as an IP model for task allocation
problems using a mathematical programming approach will take a large amount of computational time
V. ANTCOLONYOPTIMIZATION Ant colony optimization is a probabilistic te
in problems which deal with finding better paths through graphs. Artificial or simulated agents locate optimal solutions by moving through a parametric space representing all possible solutions.
Natural agents lay down pheromones directing e
to food source while exploring the environment. Similarly, ants record their positions and the quality of their solutions while migrating for resources.
So that in later iterations more ants locate better solution by knowing the history of their predecessors. Here, simulated agents are considered as user’s task and the
cloud resources such CPU, Memory, etc.
Fig 2. Class diagram of ACO based scheduling In the proposed system, when there is a peak demand or massive unpredictable requests in the cloud environment cloud provider can outsource its requests which it cannot handle to one or more public clouds.
In order to identify the suitable and optimal resources for the submitted user task, Ant colony based scheduling is proposed. The tasks are blind and migrated towards public cloud because of the insufficiency of provider.
ACO provides a suitable and best solution in this problem instance like the original ants searching for food sources based on the pheromone trail. ACO broker is implemented in CloudSim[3], a simulator for cloud environment.
Yjlk Binary decision variable, such that y
is allocated to CPk; otherwise y
International Journal of Emerging Technology in Computer Science & Electronics (IJETCSE)
ISSN: 0976-1353
Volume 13 Issue 1 –MARCH 2015.
Decision variables
The first term of objective function represents the income of CP1 and the second one means its cost[17]. Constraint guarantees that each task is vider. The problem is task allocation. Solving such problems using a mathematical programming approach will take a large amount of computational time[18].
OPTIMIZATION
is a probabilistic technique useful in problems which deal with finding better paths through graphs. Artificial or simulated agents locate optimal solutions by moving through a parametric space representing all Natural agents lay down pheromones directing each other e environment. Similarly, their positions and the quality of their solutions So that in later iterations more ants locate better solution by knowing the history of their predecessors. Here, simulated and the food source is
d scheduling In the proposed system, when there is a peak demand or massive unpredictable requests in the cloud environment the cloud provider can outsource its requests which it cannot le and optimal resources for the submitted user task, Ant colony based scheduling is proposed. The tasks are blind and migrated towards public cloud because of the insufficiency of provider.
ACO provides a suitable and best solution in this problem e like the original ants searching for food sources broker is implemented in , a simulator for cloud environment.
VI. EXPERIMENTALRESULTS
Experiments were conducted using CloudS
cloud simulator. Initially user traffic were generated for the allocation of requests in private cloud. Round robin based scheduling approach were used for the scheduling in the private cloud.
In order to show insufficiency of resources in private cloud more number of unpredictable requests were generated .ACO based scheduling is performed among the public cloud for the suitable and optimal allocation of user requests.
As a result of simulation we observed that the proposed system improves the profit of cloud provider b
user who could not understand where their actual tasks are executed. And at the same time it provides QoS to the user where no user requests are rejected irrespective of the situation.
Fig 3. Graph depicting normal workload
At normal workload the number of tasks varies gradually with time . the Fig 3 depicts a typical graph with increase in requirement of cloud resources.
number of tasks to be handled by a cloud service provider gradually increase based on the req
user traffic will be predictable and predetermined. The virtual machines are identified based on the requirements of the user.
Fig 4. Graph representing peak load
0 5 10 15 20 25 30 0 1 2
Arrival of Task at normal
condition
0 5 10 15 20 25 30 35 0 2 4Unpredictable tasks
Binary decision variable, such that yjlk=1 if tjI
otherwise yjlk=0
International Journal of Emerging Technology in Computer Science & Electronics (IJETCSE)
RESULTSANDANALYSIS periments were conducted using CloudSim[3], the
or. Initially user traffic were generated for the allocation of requests in private cloud. Round robin based scheduling approach were used for the scheduling in the In order to show insufficiency of resources in private cloud unpredictable requests were generated .ACO based scheduling is performed among the public cloud for the suitable and optimal allocation of user requests.
As a result of simulation we observed that the proposed system improves the profit of cloud provider by satisfying the user who could not understand where their actual tasks are executed. And at the same time it provides QoS to the user where no user requests are rejected irrespective of the
Fig 3. Graph depicting normal workload
workload the number of tasks varies gradually with time . the Fig 3 depicts a typical graph with increase in requirement of cloud resources. At usual workload the number of tasks to be handled by a cloud service provider gradually increase based on the requirements of the user. The user traffic will be predictable and predetermined. The virtual machines are identified based on the requirements of the user.
Fig 4. Graph representing peak load
3 4
Arrival of Task at normal
condition
No of tasks
6 8
Unpredictable tasks
International Journal of Emerging Technology in Computer Science & Electronics (IJETCSE)
ISSN: 0976-1353
Volume 13 Issue 1 –MARCH 2015.
The proposed system addresses the issue of scheduling task in peak load. Where as the arrival of user requests are massive and unpredictable. In Fig 4, in a typical cloud environment the number of task are more at time 4 and hence the requirement for cloud resources will be maximum.
The provider has to satisfy the user request by providing sufficient resources for the execution of task in cloud environment. It is necessary that the request of the user should not be rejected, if it occurs the user may switch over to other provider hence causes performance degradation.
Some of the scenarios involved in the execution are tabulated below.
Name CPU Memory
Small 1 1.8
Large 5 2.6
Table III. VM Instance types
Application Cloud resource Parameter Value Resources Number Number of tasks 10 batches CPU 512 VM instance types 2 Memory 1024GB
Runtime 1hour
Throughput 5 batches Table IV. Problem Instance
Table V. Comparison with other algorithms Experimental results of existing system [18] outperforms the standard PSO[6] for a given problem instance is given by
We observed that the proposed system outperforms the existing system .since the existing system has a possibility of trapping into local minima it achieves a solution which is not
actual best solution hence the provision of resource is not profitable for the provider. The user who requires the resource with RAM capacity of 100GB may provided with the resource with 120GB or 90GB which is not actual but nearer.
ACO outperforms the existing system by providing a better utilization of resources based on the value in the pheromone table and it randomly chooses the task in the intention that every task is important.
Fig 5.Performance of ACO with SLPSO
The proposed system increases the profit of IaaS provider by outsourcing task to external cloud when there is a peak demand. Since the resources are scalable when external cloud datacenter are included. The system also provides better service to the cloud user.
VII. RELATEDWORK
In cloud computing[1], scheduling of user task in cloud resource is different from one another. Scheduling of task in cloud environment is not the same as in traditional scheduling methods. Hence, task scheduling in cloud computing has focus and different methods have been proposed by cloud researchers.
The issue of scheduling task in cloud environment includes proposal for effective utilization of cloud resources and providing better service to the cloud user[19]. Some of the work that have been done in the past are discussed below. In[12], Ruber etal., proposed task outsourcing in terms of heavy load for increasing the utilization of datacenter
But it failed to specify which workloads to outsource and where to schedule.
In[1], Ciu etal., proposed a resource reservation strategy in cloud environment which includes previous log records for reservation. The system considers only the single cloud and also rejects the user requests when resources are not sufficient.
Scheduling approaches based on genetic algorithm[5] considers independent and dividable tasks. The approach minimizes the maximum completion time of all tasks but does not considers the case of resource limitation also it suggests to use the resources as few as possible.
0 2000 4000 6000 8000 10000 12000 (1 0 ,5 ) (3 0 ,1 0 ) (5 0 ,1 5 ) (7 0 ,2 0 ) (9 0 ,2 0 ) Profit Task,Resource ACO SLPSO Algorithm Average
runtime Average throughput
SPO 4.4 500 bps
SLPSO 4.6 550 bps
Round Robin 4.9 570 bps
International Journal of Emerging Technology in Computer Science & Electronics (IJETCSE)
ISSN: 0976-1353
Volume 13 Issue 1 –MARCH 2015.
In[17], Zuo etal., suggested an approach based on “SLPSO” which has its origin from Swarm intelligence technique but which implements particle swarm optimization repeatedly for attaining self-adaptivity. Whereas, there is a possibility of trapping into local optima.
Most of the work done in cloud environment for task scheduling concentrates either on user benefit or on the provider profit. Both of which are important for an effective system.
The proposed system based on ACO increases the provider’s profit, user benefit, promotes for better migration among cloud and perfectly adapts to dynamic environment.
VIII. CONCLUSION AND FUTURE WORK
The proposed system increases the profit of cloud service provider by outsourcing task to external cloud when its resources are not enough. It allows the user’s to choose the resource configuration by providing a better interface so that all the users are satisfied with the guaranteed QoS. The work mainly focuses on the profit of provider and better QoS to user without any standardization or agreement required in cloud federation. The use of ACO outperforms the existing SLPSO(Self-adaptive learning PSO) and provides optimized scheduling there by it increases the response time and throughput of the proposed system. In future, the work can be improved with other swarm intelligence technique in hybrid with ACO for the betterment of the system.
ACKNOWLEDGMENT
I thank the almighty for giving me this thinking and innovativeness in the research field and my family especially my son for his patience that I spent most of his time with me, in this work. I am very grateful to my guide Dr.S.Selvakumar for providing constant support throughout this work.
REFERENCES
[1] Cui [1] Cui Lin, Shiyong Lu, “Scheduling Scientific Workflows Elastically for Cloud Computing” in IEEE 4th International Conference on Cloud Computing, 2011.
[2] Dar-Tzen Peng, Kang G. Shin and Tarek F. Abdelzaher, “ Assignment and Scheduling Communicating Periodic Tasks in Distributed Real-Time Systems”, IEEE Transactions On Software Engineering, Vol. 23, No. 12, pp. 745-758, 1997.
[3] Dr. Rajkumar Buyya, “CloudSim: a toolkit for modelling and simulation of cloud computing environments and evaluation of resource provisioning algorithm”, published online 24 august in Wiley Online Library 2010, pp. 23-50.
[4] J. M. Hyman, J. M. Hyman, A. A. Lazar, A. A. Lazar, G. Pacifici, and G. Pacifici.“Real-time scheduling with quality of service constraints” IEEE Journal on Selected Areas in Communications,, pp:1052–1063, 1991.
[5] K. Thirupathi Rao, P. Sai Kiran, L.S.S Reddy, V. Krishna Reddy, B. Thirumala Rao, “Genetic Algorithm For Energy Efficient Placement Of Virtual Machines In Cloud Environment”, in proc IEEE International Conference on Future Information
Technology (IEEE ICFIT 2010), China, pp:V2-213 - V2-217. [6] Kennedy, J.; Eberhart, R., “Particle Swarm Optimization”, IEEE
International Conference on Neural Networks, Perth, WA, Australia, Nov.1995, pp.1942-1948.
[7] L. M. Gambardella and M. Dorigo. “An ant colony system hybridized with a new local search for the sequential ordering problem” INFORMS Journal on Computing, 12(3):237–255, 2000.
[8] M. Dorigo and T. St¨utzle. “Ant Colony Optimization” MIT Press, Cambridge, MA,2004.
[9] M. Dorigo, V. Maniezzo, and A. Colorni, “The ant system: an autocatalytic optimizing process “, Technical Report TR91-016, Politecnico di Milano (1991).
[10] Meng Xu, Lizhen Cui, Haiyang Wang, Yanbing Bi, “A Multiple QoS Constrained Scheduling Strategy of Multiple Workflows for Cloud Computing”, in IEEE International Symposium on Parallel and Distributed Processing, Chengdu 10-12 Aug,2009, pp 629-634. [11] Strategy of Multiple Workflows for Cloud Computing”, in IEEE
International Symposium on Parallel and Distributed Processing, Chengdu 10-12 Aug, 2009, pp 629-634.
[12] Syed Tauhid Zuheri1, Tamanna Shamrin2 and Rusia Tanbin3, Firoj Mahmud4, “An Efficient Load Balancing Approach in Cloud Environment by using Round Robin Algorithm”, International Journal of Artificial and Mechatronics, volume 1, issue 5, 2013, pp 96-99. [13] T. D. Braun, H. J. Siegel, N. Beck, L. L. B¨ol¨oni, M. Maheswaran, A.
I. Reuther, J. P.Robertson, M. D. Theys, B. Yao, D.Hensgen,andR.F.Freund.“A comparison of eleven static heuristics for mapping a class of independent tasks onto heterogeneous distributed computing systems” Journal of Parallel and Distributed Computing, 61:810–837, June 2001.
[14] Tzu-Chiang Chiang, Po-Yin Chang, and Yueh-Min Huang, “Multi-Processor Tasks with Resource and Timing Constraints Using Particle Swarm Optimization”, IJCSNS International Journal of Computer Science and Network Security, Vol.6 No.4, pp. 71-77, 2006. [15] V. Krishna Reddy, B. Thirumal Rao, L.S.S. Reddy, P. Sai Kiran “Research Issues in Cloud Computing” Global Journal of Computer Science and Technology, Volume 11, Issue 11, July 2011
[16] Ruben Van Den Bossche, Kurt Vanmechelen and Jan Broeckhove,”Cost-Optimal Scheduling in Hybrid IaaS Clouds for Deadline Constraied Workloads”,IEEE 3rd International Conference on Cloud Computing.2010.
[17] Xingquan Zuo, Member, IEEE, Guoxiang Zhang, and Wei Tan, “Self-Adaptive Learning PSO-Based Deadline Constrained Task Scheduling for Hybrid IaaS Cloud.”, IEEE Transactions On Automation science and engineering, Vol.11,No.2,April 2014. [18] Zhangjun Wu, Xiao Liu, Zhiwei Ni, Dong Yuan, Yun Yang, “A
Market-Oriented Hierarchical Scheduling Strategy in Cloud Workflow Systems” in JSC2010.
M.Anuradha, completed B.E., in Computer
science and Engineering and worked as a Lecturer for 5 years Now doing PG in GKM College of Engineering and Technology, Chennai. Attracted towards “Grid computing” after attended an FDP conducted by Dr.Tamaraiselvi, Dean, MIT Campus, analyzed certain papers in Grid. As a successor of Grid got interest in Cloud the result is the submitted work. Future ambition is to do research in Cloud computing and has an idea to create “Virtual Laboratory” for the betterment of rural engineering students.
S.Selvakumar received Doctor of
Philosophy in Computer Science and Engineering from the Anna University. He has over 20 years experience in institutions and Organizations. Currently he is a Professor and Head of CSE, GKM College of Engineering & Technology. He is a recognized Supervisor for Ph.D of Anna University, Chennai and guiding 8 Ph.D Scholars. He has long been interested in Software Engineering. He has carried out various AICTE sponsored Short Term Programs and worked on various Projects. He is a Senior Member in the Computer Society of India, Member in IEEE, ACM, the Institution of Engineers and the ISTE. He has published over 20 papers in International Journals and Conferences and reviewed journals and conferences. He is Authorized Trainer PSP/TSP by Software Engineers Institute (SEI), Carnegie Mellon University. He carried out training programs for Intergra Microsystems Ltd, Tata Consultancy Services.
