2016 International Conference on Circuit, Power and Computing Technologies [ICCPCT]

Auction Based Dynamic Resource Allocation in

Cloud

E. Iniya Nehru Infant Smile Shyni. J RanjithBalakrishnan

Senior Technical Director, PG student, Assistant Professor,
National Informatics Center, TIFAC – CORE in PCT, TIFAC – CORE in PCT,
Chennai, India Velammal Engineering College, Velammal Engineering College,
Email: [email protected] Chennai, India Chennai, India
Email: [email protected] Email: [email protected]

Abstract—Cloud computing is one of the fastest emerging II. PROBLEMS IN RESOURCE ALLOCATION
technology. As we all know, cloud is a wide pool of resource
which provides resources based on the users request. Any service Cloud computing is mainly popular for flexibility which
can be provided as a service through cloud. As the cloud contains can be attained through scalability and availability of services
many resources, there may also wastage of resources. To reduce whenever in need. It aims to supply reliable, customized and
this wastage, cloud providers enter into auctioning of resources good QoS parameters in the dynamically changing computing
when the demand is high. Cloud computing includes distinct environments. To attain these goals, resource allocation is very
resources. Because of the complementary and supplementary much important. Proper resource allocation gains customer
effects between distinct assets, bidders have preferences not for satisfaction also.
just a single resource but also for a set of resources. Auctioning
for a bundle of resources is called combinatorial auction. A. Significance in resource allocation
Dynamic resource allocation in on demand for a bundle of Resource allocation in cloud is the process of assigning
resource is proposed by using CA – PROVISION algorithm and
the scenarios are simulated using Cloudsim, a simulator meant
cloud resources to the requested cloud users over the internet.
for cloud computing analysis. Resource allocation problems are [13]:
a) Resource contension: More users tries to access the same
Keywords—Cloud Computing, Auction, Bid, Dynamic Resource
resource thereby increasing traffic.
Allocation, Combinatorial auction and Cloudsim.
b) Resource scarcity: When there are no enough resources.
I. INTRODUCTION c) Resource fragmentation: It leads to isolation of resources
Cloud computing is the fastest emerging paradigm of to some users and they can’t be allocated to the persons in
distributed computing. It is a computing model and not need.
technology. It is an internet based computing where users have d) Over provisioning: Providing more resources than the
to pay as per their usage. Cloud computing has many need.
advantages like no infrastructure investment, storage e) Under provisioning: Providing less resources than the
capability, scalability, backup and recovery capabilities, requested amount.
anytime access with internet, cost effective, etc., Cloud To overcome these issues of resource allocation in cloud
provides many types of services that comes under the major computing, auction mechanism is used.
categories like infrastructure, software and platform. As cloud
computing reduces the installation costs, many of the users are B. Resource allocation by auction
switching over to cloud computing resources. Now a day’s Auction is a process of buying and selling goods by making
anything i.e. data, storage, monitoring, network etc., can be bids for the available resources. The highest bidder is provided
provided as a service via cloud computing which is denoted as with the resource. Auction comes into play when the demand is
XaaS (Anything-as-a-Service) [3]. high and resources are less. Both the buyer and the seller are
free to provide their valuation for the resource they are in need
Cloud computing is a virtualized environment which
and by iterative discussions, an agreement is made for the
provides services as Virtual Machine (VM) instances. These
payment of the resource. As in cloud computing, the wide
VM instances are distinct. Cloud users can choose which
pools of resource are not used by the users to its full utility.
configuration or amount of instances they want. As there are
Also in on demand situation, under utilization of resources can
many resources available, there may be some situations which
happen if we follow the FCFS (First Come First Serve) basis.
lead to underutilization of these resources. So, a proper
So, if we use auction mechanism for resource allocation,
resource allocation technique has to be done to increase the
utilization of resources as well as providers has economic
revenue of the cloud provider.
benefits and proper allocation reduces wastage of resources
[12].

978-1-5090-1277-0/16/$31.00 ©2016 IEEE

 2016 International Conference on Circuit, Power and Computing Technologies [ICCPCT]
Auction environment includes auctioneer, broker, cloud approximation algorithms for auctions. This involves the
provider and user. Auctioneer may be the broker sometimes. sequence of greedy procedures for allocation of resources
This is depicted in the Fig. 1 below [1]. which favors only to the cloud provider’s profit. Daniel
Lehmann et al.[6] proposed greedy method for winner
determination for single minded auctions. Bidders are
prioritized based on the norms in descending order and no two
bidders have the same norm. Deadlock condition is also
checked after each allocation.
Truthful character of the algorithm is explained by
AhuvaMualem et al.[2]This provides the idea that, whenever
the value of the resource increases, the usage cost for that
resource also increases. This deals with the dynamic pricing
mechanism of resources. SharrukhZaman and Daniel Grosu[9]
proposed the CA – LP and CA – GREEDY mechanisms of
combinatorial auction to find out the winner and for dynamic
allocation. This algorithms uses fixed pricing mechanisms as it
follows static allocation of resources. CA – LP mechanism
involves solving of auction by linear programming. CA –
GREEDY algorithm uses is the extended version of greedy
allocation as it favors only the providers. Recent works of
Figure 1: Simple auction environment in cloud them[10] reveals the CA – PROVISION algorithm which is
Many existing techniques in auction use static allocation the density based dynamic resource allocation. Our proposed
and pricing. The problems in static allocation are: work eliminates the dummy user concept and reverses price
methods in CA – PROVISION algorithm. Implementation is
i. Contracts in resource allocation always favor the vendors. done by using Cloudsim simulator.
There may be some situations when the instances of users
and vendors are mismatched. IV. COMBINATORIAL AUCTION BASED DYNAMIC RESOURCE
ii. It is very difficult to maintain the SLA in fixed pricing. ALLOCATION
Combinatorial auction allows the users to bid for a bundle
These problems can be overcome by dynamic resource
allocation. Dynamic resource allocation provides economic of resources or packages. Bidding amount is concerned for the
benefit to the provider and the client has to pay only for the whole package.
usage of the resources. Truthfulness can be attained. In auction
winner determination is the major issue as it is NP – hard
problem.
Through virtualization, cloud provider can be able
configure the VMs based on the user’s preferences. Any
combinations of VM’s can also be obtained with the available
resources with the provider. Instead of single resource, auction
for this bundle of resources or combinations is called
combinatorial auctions [7].
Organization of this paper is given by:In Section II,
problems in resource allocation is discussed. Section III deals
with related works. Section IV explains the combinatorial
auction algorithm and Section V provides the simulation
results. In Section VI future enhancements of this work are
discussed.
III. RELATED WORKS
VM provisioning by auction is experimented by many
researchers. Dynamic provisioning of resources was done by
Quiroz et al. through decentralized online clustering algorithm
in the grid environments. This allows users to choose their
own combinations of resources. Dynamically changing
environment is the base problem. Our proposed system deals
with the current demands from the users [4]. Combinatorial
auction based protocol for resource allocation in grids is well Figure 2: Combinatorial auction
explained by Anubhav Das and Daniel Grosu[5].Their work
explains about the winner determination problem through
 2016 International Conference on Circuit, Power and Computing Technologies [ICCPCT]
In some cases, cloud provider itself provides some default request, vj is the amount the user wishes to pay for the
packages like in Amazon [3] and users can also have the requested resource.
preferences for the packages of their combinations. Winner determination is done based on the calculation of
A sample combinatorial auction mechanism is depicted in densities of bids. Density of the bid is calculated by using the
Fig. 2 [5]. Allowing the users to express their preferences formula,
increases the provider’s repudiation and revenue.
Combinatorial auction mechanism is mainly meant for
improving economic efficiency and system performance [11]. wherevjrepresents the amount the user wishes to pay for the
In this work, when the resource contention or the resource and sj is given by the number of instances of resources
competition for the resource increases, the cloud manager will allocated and the weights of the resources. The bid density is
call for the auction of resources among the users. Users how much amount the user wishes to pay for single unit of
receive the notification and then start to send bids for the resource in need.
resources. This combinatorial auction based mechanism called
CA – PROVISION provides the information about the The densities of the users are sorted in descending order.
allocation of available resources to the requested users, Users with the highest bid are assigned the winner and the
amount to be paid by the users, set of VM instances allocated winners are sorted in the list based on their densities.
Provisioning of resources is done on the basis of this sorted
to each user and this mechanism mainly follows the all or
order in the all or none basis. After allocation of resources for a
none format. Either the set of requested resources are allocated
user the environment is checked for any deadlock condition. If
or none of them is allocated to the users. Users can request for it is found then the rollback operation is implemented. Once all
their own combinations of VMs through the bids. the resources are provided as VM instances to the users, the
payment calculation phase starts based on the usage of the
resources. The users are constantly monitored and once their
Collection of bids consumption is over they are directed to the payment gateway.
Cost for owned resources which are idle is also considered.
Cost calculation for both idle and consumed resources is
Calculating bid densities calculated by

Sorting of bids as per bid densities

Where CRand CI represents the cost of resource at run time
and idle time respectively.
As this algorithm follows dynamic allocation, same
Winner determination resource can be allocated to the other user when it is not in use
by the owned user. This increases the provider’s revenue and
repudiation.This CA – PROVISION algorithm is truthful as it
calculates the amount only based on the users usage only.
Allocation of resources
V. EXPERIMENTAL RESULTS
CloudSim simulator is used for the simulation of cloud
environment. It supports modeling and simulation of large
Pricing scale cloud computing data centers and supports simulation of
virtualized hosts [8].
Figure 3: Flow chart for CA – PROVISION algorithm

This algorithm is divided into three major phases:

Collecting the bids from the users, winner determination by
density calculation and payment. A cloud provider might
consist of n different types of VM instances and its weight
based on computing power is indicated by w. Cloud providers
decides the weights for the resources during the configuration
of the resources itself. Users can able to choose any type of
VM instances they want and bid for the resource in
combinations as

This indicates the user j is requesting the bid containing the Figure 4: Density Calculation
m different types of resources they are in need. In this bid The major usage of Cloudsim in this project is Cloudsim
contains packages for implementing auctions. In this
 2016 International Conference on Circuit, Power and Computing Technologies [ICCPCT]
simulation, five bidders and brokers are taken and the and heterogeneity should be addressed without affecting the
algorithm is applied and results are studied. Results shown performance and availability.
that the provider’s revenue is increased to certain considerable
extend than the normal auctions. REFERENCES

[1] Abinandan S. Prasad and Shrisha Rao, “A Mechanism Design Approach

to Resource Procurement in Cloud Computing”, Proc. IEEE
Transactions on Computers, pp. 17-30, 2014.
[2] Ahuva Mualem and Noam Nisan,“Truthful approximation mechanisms
for restricted combinatorial auctions,”Proc. 18th Nat’l conf. Artificial
Intelligence, pp. 379-384,2002.
[3] Amazon, “Amazon Elastic Compute Cloud (Amazon EC2),”
http://aws.amazon.com/ec2/,2013.
[4] A. Quiroz, H. Kim, M. Parashar, N. Gnanasambandam, and N. Sharma,
“Towards Autonomic Workload Provisioing for Enterprise Grids and
Clouds,” Proc. IEEE/ACM 10th Int’l Conf. Grid Computing, pp. 50-57,
2009.
[5] A. Das and D. Grosu, “Combinatorial Auction-Based Protocols for
Resource Allocation in Grids,” Proc. 19th Int’l Parallel and Distributed
Processing Symp., Sixth Workshop Parallel and Distributed Scientific
and Eng. Computing, 2005.
[6] D. Lehmann, L. I. OÇallaghan, and Y. Shoham, “Truth Revelation in
Figure 5: Output for the simulation Approximately Efficient Combinatorial Auctions,” J. the ACM, vol. 49,
no. 5, pp. 577-602, 2002.
Client’s request traffic is also reduced due to the combinations [7] P. Cramton, Y. Shoham, and R. Steinberg, Combinatorial Auctions. MIT
of resource packages rather than single resource request. Press, 2005.
Dynamic resource allocation increases the cloud users and [8] Rodrigo N. Calheiros, Rajiv Ranjan, Anton Beloglazov, Cesar A. F. De
provider’s repudiation. Underutilization of resources is Rose and Rajkumar Buyya, “CloudSim: A toolkit for modeling and
minimized. Further implementations on increasing the number simulation of cloud computing environments and evaluation of resource
provisioning algorithms,” Proc. Wiley Online Library, 2010.
of bidders and datacenters are under progress.
[9] S. Zaman and D. Grosu, “Combinatorial Auction-Based Allocation of
Virtual Machine Instances in Clouds,” Proc. IEEE Second Int’l Conf.
VI. CONCLUSION AND FUTURE WORK
Cloud Comp. Technology and Science, pp. 127-134, 2010.
Results indicated that this CA – PROVISION algorithm can [10] S. Zaman and D. Grosu, “A Combinatorial Auction-Based Mechanism
generate better revenue than the other algorithms as it is based for Dynamic VM Provisioning and Allocation in Clouds,” Proc. IEEE
Transactions on Cloud Comp., pp. 129-141, 2013.
on dynamic allocation of resources. Provider’s repudiation
[11] Sven de Vries and Rakesh Vohra, “Combinatorial Auctions: A survey”.
also increased due to the truthfulness of the algorithm. Further
[12] Youwen Lan, Weiqin Tong, Zongheng Liu, Yan Hou, “Multi-Unit
processes are under execution. Efficient pricing mechanisms Continuous Double Auction Based Resource Allocation Method,” Int’l
need to be studied in the future as there may occur some Conf. Intelligent Control and Information Processing, pp. 773-777,
fraudulent cases as the cloud environment is mainly based on 2012.
the internet. Auction along with VM migration techniques are [13] V. Vinothina, Dr. R. Sridaran and Dr. Padmavathi Ganapathi, “A Survey
to be studied further. In this paper resource allocation by on Resource Allocation Startegies in Cloud Computing,” Proc. Int’l
Journal of Advanced Computer Science and Application, pp. 97-104,
auction is mainly concentrated. But when implementing in real 2012.
time, the major cloud computing issues like security, privacy