Challenges in Dynamic Resource Allocation and Task Scheduling in Heterogeneous Clouds

International Journal on Recent and Innovation Trends in Computing and Communication ISSN: 2321-8169
Volume: 5 Issue: 7 93 – 97
_______________________________________________________________________________________________
93
IJRITCC | July 2017, Available @ http://www.ijritcc.org
_______________________________________________________________________________________
Challenges in Dynamic Resource Allocation and Task Scheduling in
Heterogeneous Clouds
Ms. Ragini Karwayun
Researcher
Mewar University
Rajasthan
raginik22 @gmail.com
Dr. H. S. Sharma
Pro Vice Chancellor
Mewar University
Rajasthan
profhssharma@yahoo.com
Dr. K. P. Yadav
Director
IIMT
Greater Noida
karunesh732@gmail.com
Abstract— Resource Allocation and Task scheduling are the most important key words in today’s dynamic cloud based
applications. Task scheduling involves assigning tasks to available processors with the aim of producing minimum execution time,
whereas resource allocation involves deciding on an allocation policy to allocate resources to various tasks so as to have
maximum resource utilization. Algorithms used for scheduling resources for virtual machines are designed for both homogeneous
and heterogeneous environments. Majority of the algorithms focus on processing ability often neglecting other features such as
network bandwidth and actual resource requirements. One of the major pitfalls in cloud computing is related to optimizing the
resources being allocated. Because of the uniqueness of the model, resource allocation is performed with the objective of
minimizing the costs associated with it. The other challenges of resource allocation are meeting customer demands and application
requirements. In this paper we will focus on the challenges faced in task scheduling and resource allocation in dynamic
heterogeneous clouds.
Keywords-Resource Allocation, Task Scheduling, distributed Heterogeneous Clouds
__________________________________________________*****_________________________________________________
I. INTRODUCTION
Internet based environment of cloud computing framework
provides opportunities for provisioning, consumption and
delivery of services such as information, software and
computing resources. In today’s world there are numerous
vendors that rent computing resources on-demand, bill on a
pay-as-you-go basis, and multiplex many users on the same
physical infrastructure. These cloud computing environments
provide an illusion of infinite computing resources to cloud
users so that they can increase or decrease their resource
consumption rate according to the demands.
In cloud computing environments, the two important
stakeholders: cloud providers and cloud users have different set
of goals and objectives often conflicting with each other. There
are providers who hold massive computing resources in their
large datacenters and rent resources out to users on a per-usage
basis. Then there are users who have applications with
fluctuating loads and lease resources from providers to run their
applications. Providers on one hand wish to maximize revenues
by achieving high resource utilization, while users wish to have
minimum expenses while meeting their performance
requirements[6].
The two aspects of cloud computing applications are based
on tradeoff between various characteristics depending on the
domain and for providing the service by the cloud service
provider and using the services by the cloud user, an initial
agreement called the Service Level Agreement (SLA) has to be
made between the cloud users and the cloud service provider.
[2] While resource allocation is made to the cloud user, SLA
violation should be avoided as much as possible or SLA
violation should be minimal without compromising Quality of
Service (QoS) parameters like performance, availability,
response time, security, reliability, throughput etc.
Performance: For some application demands, performance is
an important criteria. The system should perform well to
provide service to the cloud user.
Response Time: For interactive applications, response time is
an important factor. The system must respond well for these
kinds of applications.
Reliability: The system used should be reliable so that the
cloud user has no head ache of changing the system.
Availability: Whenever cloud resources are requested the
cloud service provider must be able to allocate within short
span.
Security: For critical applications like online transaction
applications, system us ed has to be secure. Otherwise it is not
safe to use such a kind of system.
Throughput: No. of applications run per unit time should be
more.
II. CHALLENGES IN RESOURCE ALLOCATION IN CLOUDS
In cloud computing, Resource Allocation is the process of
assigning available resources to the needed cloud applications
over the internet. Cloud services are starved if the resource
allocation is not managed properly. Resource provisioning
solves that problem by allowing the service providers to
properly manage the resources for each individual module.
Resource Allocation Strategy is all about integrating cloud
provider activities for utilizing and allocating available
resources within the limit of cloud environment so as to meet

_______________________________________________________________________________________________
94
_______________________________________________________________________________________
the needs of the cloud application. It requires the type and
amount of resources needed by each application in order to
complete a user job. The order and time of allocation of
resources are also an input for an optimal strategy. An optimal
Resource allocation scheme should avoid the following criteria
as follows:
Resource contention situation arises when two applications try
to access the same resource at the same time.
Scarcity of resources arises when there are limited resources.
Resource fragmentation situation arises when the resources
are isolated. [There will be enough resources but not able to
allocate to the needed application.]
Over-provisioning of resources arises when the application
gets surplus resources than the demanded one.
Under-provisioning of resources occurs when the application
is assigned with fewer numbers of resources than the demand.
Cloud Users
Interfaces
Applications
Interfaces
Virtualized Resources
Cloud Providers
Cloud users estimation of resources demands to complete a job
before the estimated time may lead to an over-provisioning of
resources. Cloud providers’ allocation of resources may lead to
an under-provisioning of resources. From the cloud user’s
angle, the application requirement and Service Level
Agreement (SLA) are major inputs to allocation algorithms.
The offerings, resource status and available resources are the
inputs required from the provider’s side to manage and allocate
resources to host applications [1]. The outcome of any optimal
Resource allocation scheme must satisfy the QoS parameters
specified in SLA. Cloud providers while providing reliable
resources, faces a crucial problem in allocating and managing
resources dynamically across the applications.
Resource Allocation algorithms are available in two
categories – static and dynamic.
Static Allocation: The cloud user declares the required
resources beforehand. By doing so user is aware of the details
of required resources – types of resources and how many
instances of each type of resources are needed ahead of using
the system. But the drawback in this case is it leads to
underutilization or overutilization of resources depending on
the time the application is run. The striking feature of these
types of algorithms is that they use mostly fair allocation
strategies.
Dynamic Allocation: Cloud resources are not declared prior
to starting the application, but are requested by the cloud user
on the run or as and when the application needs. These
algorithms provide best utilization of resources avoiding
underutilization and overutilization of resources, as much as
possible. But nearly all dynamic resource allocation policies are
unfair. The requested resources might not be available when
requested on the fly. The service provider has to make
allocation from other participating cloud data center.
In most cases, the interaction between providers and users is
as follows.
1. A user sends a request for resources to a provider.
2. When the provider receives the request, it looks for
resources to satisfy the request and assigns the
resources to the requesting user, typically as a form of
virtual machines(VMs).
3. Then the user uses the assigned resources to run
applications and pays for the resources that are used.
4. When the user is done with the resources, they are
returned to the provider.
The most interesting aspect of the cloud computing
environment is that the major players in the field are often
different parties with their own interests. Typically, the goal of
providers is to generate as much revenue as possible with
minimum investment[6]. To achieve this they might want to
squash their computing resources; for example, by hosting as
many VMs as possible on each machine, thus wishing to
maximize resource utilization. However, placing too many
VMs on a single machine will cause VMs to interfere with
each other, resulting in degraded and/or unpredictable
performance which, in turn, affects the users. Thus, the
providers may evict existing VMs or reject resource requests
to maintain service quality, but it could make the environment
even more unpredictable. On the other hand, users want their
jobs done at minimal expense or, in other words, they wish to
have maximum performance at minimal cost. This involves
having proper resources that suit the workload requirements of
users' applications and utilize resources effectively. They also
have to take unpredictable resources into account when they
request resources and schedule applications.
Resource allocation in distributed heterogeneous clouds in
dynamic computing environment focuses on four fundamental
issues :
1. Resource modelling.
2. Resource allocation and treatment.
3. Resource offering and treatment.
4. Resource discovery and monitoring selection.
Resource allocation algorithms must have the complete
knowledge of the status of each resource in the computing
environment and use this knowledge to apply it intelligently to
allocate physical or virtual resources to applications as per
their requirements. [1] So the information that is required by
the resource allocation algorithms can be summed up as: cloud
resources, resource modelling, application requirements and
specifications and providers requirements.
Resource Modelling : Infinite computing resources are
available to be used on demand by the cloud users. The cloud
environment thru the use of resource allocation algorithms
tries to meet the demands of the users in a dynamic way,
allowing the multiplexing of physical resources, avoiding both
under provisioning and over provisioning of resources.
Moreover as it is a heterogeneous environment it is essential to
develop a proper resource model. All optimized allocation
algorithms are very strongly dependent on the resource

_______________________________________________________________________________________________
95
_______________________________________________________________________________________
modelling scheme used. Computing and network resources
can be represented by using RDF (Resource Description
Framework) and NDL ( Network Description Language). It
must be kept in mind that these tools must be capable of
representing resources, networks and applications virtually in
terms of properties and functionalities[7].
Another important point of consideration is the granularity of
resource description. More the details used in describing the
resources better will be the allocation being more flexible in
usage of resources , on the other hand more details will make
the resource selection and allocation module complex and
harder to handle.
One additional point of challenge in resource allocation in
heterogeneous environment is the interoperability in clouds
inorder to provide seamless flow of data between clouds,
across clouds and their local applications. Layered
architectural design and use of open APIs are practical options
for interoperability.
There are two types of heterogeneities faced by
interoperability in clouds [8]. Vertical is intra-cloud whereas
Horizontal is inter-cloud( different vendors) heterogeneities.
Vertical heterogeneity can be addressed with the help of
middleware and standardization. OVF (Open Virtualization
Format) can be used for managing virtual machines across the
system. Horizontal heterogeneity, as clouds from different
vendors are involved, is very difficult to obtain. As each
vendor may use different level of abstraction to represent the
resources, it is challenging to define a method of interaction
between clouds. This issue can be addressed with the help of
high level of granularity in the modelling but has a serious
drawback of losing information. But it is advantageous to have
horizontal interoperability as there are numerous instances
when a user’s request cannot be addressed by a cloud provider
due to unavailability of resources locally. In such cases, the
provider can borrow resources from another vendor by
dynamically negotiating.
Resource Offering and Treatment: After modelling of
resources, the providers offer interfaces (middleware) that on
one hand handles resources at a lower end and on the other
hand deal with the requirements of the applications at a higher
level. It is important to emphasize on the fact that there should
be a clear level of abstraction between the resource modelling
and the way they are offered to the end user. For example, a
provider can model the resources independently such as speed
of processor or size of memory but offer them as a collection
of items , such as a VM.
Since it is important to have a generic solution for clouds so
that they may support as many applications as possible,
resource offering becomes a very complex process. Providers
must handle major issues like – how to have a good trade off
between the granularity of resource modelling and an easy
generic solution, and how can one define the generic level of
the cloud?
The Resource Allocation Scheme must ensure that all
requirements must be met with the set of available resources
as much as possible. Moreover the requirements of the cloud
users must be specified properly in the SLA and it is essential
that both parties must adhere to it. The list of the requirements
specified in the SLA may include but not limited to the
following:
i. Network bandwidth requirement.
ii. Computational capability (processor speed and
memory) requirement.
iii. Topology of the nodes which help the cloud users to
define inter-node relationships and restrictions on
uplink and downlink communications.
iv. Jurisdiction of physical location and methods of
handling data by applications. Location of data may
be restricted within a country or a continent by
copyright laws.
v. The node proximity is another requirement which
may be seen as a constraint imposed on the physical
distance or delay value (maximum or minimum)
between nodes.
vi. The application interaction defines the types of
configuration applications will use to exchange
information with each other.
vii. The cloud user should also be able to define
scalability rules which gives the specifications for the
growth of applications and how will and how many
resources will they consume.
Authors in [9] describes the methods how the VMs will be
deployed by the cloud users based on the thresholds of the
above specified metrics.
Resource Discovery and Monitoring: Providers need to find
appropriate resources that comply with the user’s request. The
most daunting task is to fine suitable resources within the
boundaries (physical/geographical) in a heterogeneous cloud.
In addition to this the costing factor of inter domain traffic also
need to be take care of.
In [8] a very simple implementation of the resource discovery
service is described for the network virtualization scenario that
uses a discovery framework with an advertisement process.
Cloud users (brokers) use it to discover and match available
resources from different providers. It consists of distributed
repositories responsible for storing resource descriptions and
states. Considering that one of the key features of cloud
computing is its capability of acquiring and releasing
resources on demand [10], resource monitoring should be
continuous, and should help with allocation and reallocation
decisions as part of overall resource usage optimization. It
should be carefully analyzed to find an acceptable trade-off
between the amount of control overhead and the frequency of
resource information refreshing.
The resource monitoring may be passive or active. It is passive
when one or more entities collect information by continuously
sending polling messages to nodes asking for information or
do this on demand when necessary. On the other hand, the
monitoring is active when nodes are autonomous and may
decide when to send asynchronously state information to some
central entity. Both the alternatives can be simultaneously be
used to improve the monitoring solution. To achieve this it is
necessary to synchronize updates in repositories to maintain
consistency and validity of state information.
Resource Selection and Optimization: Once the information
regarding cloud resource availability is collected, a set of
appropriate candidates may be highlighted. In the next phase
the resource selection process needs to find a configuration
that fulfils all requirements and optimizes the usage of the
infrastructure. In virtual networks, for example, the resource
selection mechanisms main objective is to find the best

_______________________________________________________________________________________________
96
_______________________________________________________________________________________
mapping of the virtual networks on the substrate network with
respect to the constraints [11]. Selection of a suitable solution
from a set is a very complex task due to the dynamically
changing environment , very complex algorithm, and all the
other different requirements relevant to the provider.
Optimization algorithms may be used for resource selection.
Optimization strategies used may vary , from simple and well-
known techniques such as simple heuristics with thresholds or
linear programming to newer, more complex ones, such as
Lyapunov optimization [12]. In addition, artificial intelligence
algorithms, biologically inspired ones (e.g., ant colony
behaviour), and game theory may also be applied in this
scenario. Authors in [13] define a system called Volley to
automatically migrate data across geo-distributed datacentres.
This solution uses an iterative optimization algorithm based on
weighted spherical means [13]. Resource selection strategies
fall into a priori and a posterior classes. In the a priori case,
the first allocation solution is an optimal one. To achieve this
goal, the strategy should consider all variables influencing the
allocation. For example, considering VM instances being
allocated, the optimization strategy should figure out the
problem, presenting a solution (or a set of possibilities) that
satisfies all constraints and meets the goals (e.g., minimization
of reallocations) in an optimal manner. In an a posteriori case,
once an initial allocation that can be a suboptimal solution is
made, the provider should manage its resources in a
continuous way in order to improve this solution. If necessary,
decisions such as to add or reallocate resources should be
made in order to optimize the system utilization or comply
with cloud users’ requirements. Since resource utilization and
provisioning are dynamic and changing all the time, it is
important that any a posteriori optimization strategy quickly
reach an optimal allocation level, as a result of a few
configuration trials. Furthermore, it should also be able to
optimize the old ones, readjusting them according to new
demand. In this case, the optimization strategy may also fit
with the definition of a priori and dynamic classification
Thus we have described the four challenges that should be
faced for defining any resource allocation scheme. The first
two challenges need to be taken care of in conception phase
and the next two in operational phase. When designing a
distributed heterogeneous cloud the provider should choose
the nature of its offering such as service, infrastructure and
platform as a service (SaaS, IaaS and PaaS).
III. CHALLENGES IN TASK SCHEDULING IN CLOUDS
The main objective of task scheduling is to schedule the
arriving tasks onto available processors with the aim of
producing minimum schedule length and without violating the
precedence constraints. It is often seen that the different
sections of the application task need different types of
computations. So for single machine architecture it is
impossible to satisfy all the computational requirements of
such applications. So it is essential to schedule different tasks
of such applications to different suitable processors across the
distributed heterogeneous computing resources. This system
helps in executing computationally intensive applications with
different computational requirements. But the performance of
such a system is highly dependent on the scheduling algorithm
used.
We assume that the physical locations of VMs are determined
by the provider. It is assumed that, the responsibility to
determine when to request how many resources of which type,
and how to schedule the user's application on the allocated
resources, lies with the user[6].
In a heterogeneous environment, it is important to schedule a
job on its preferred resources to achieve high performance. In
addition, it is not straightforward to provide fairness among
jobs when there are multiple jobs. Moreover, the capacity of
different machine types needs to be considered when a user
makes a resource request.
Task Scheduling algorithms are static as well as dynamic in
nature. As static algorithms are more fair, stable and easy to
use, they are more commonly used. Hadoop also uses static
algorithm for task scheduling [3].
Performance metrics for task scheduling algorithms are
schedule length, speedup, efficiency and time complexity :
Schedule length : is the maximum finish time of the exit task
in scheduled DAG ( Directed Acyclic Graph).
Speedup: of a schedule is defined as the ratio of the schedule
length obtained by assigning all tasks to the fastest processor,
to the scheduled length of application.
Efficiency: is the speedup divided by the number of
processors used. It is the measure of the effective percentage
of processor time is doing useful computation.
Time complexity: is the amount of time taken to assign every
task to specific processor according to specific priority.
Once a set of resources such as virtual machines are allocated
by the cloud driver ,the analytics engine uses the resources that
are heterogeneous and shared among multiple jobs. In this
section, we consider challenges in job scheduling on a shared,
heterogeneous cluster, to provide good performance while
guaranteeing fairness.
Share and Fairness
In a shared cluster, providing fairness is one of the most
important features that the analytics engine should support.
There are different ways to define fairness, but one method
might be having each job receive equal (or weighted) share of
computing resources at any given moment. In that sense,
Hadoop Fair Scheduler [4] takes the number of slots assigned
to a job as a metric of share, and fairness is provided by
assigning same number of slots to each job.
In a heterogeneous cluster, as all the slots are different, the
number of slots might not be an appropriate metric of the
share. Moreover, even on the same slot, the computation speed
varies depending on jobs. The performance variance on
different resources is also important to consider to improve
overall performance of the data analytics cluster; if jobs are
assigned to slots irrespective of their preferences will not only
make the job run slow, but also may prevent other jobs that
prefer the slot from utilizing it.
Progress Share
Progress share refers to how much progress each job is making
with assigned resources (or slots in Hadoop) compared to the
case of running the job on the entire cluster without sharing;
therefore, it is between 0 (no progress at all) and 1 (all
available resources are occupied).
Scheduler
To calculate the Progress share of each job, the analytics
engine scheduler should be aware of the per-slot computing

_______________________________________________________________________________________________
97
_______________________________________________________________________________________
rate(CR). To that end, each job should go through two phases:
calibration and normal. Calibration phase starts when a task is
submitted. In this phase, by measuring the completion time of
these tasks, the scheduler can determine the CR. Once the
scheduler knows the CR, the task enters the normal phase.
During this phase, when a slot becomes available, a job of
which the share is less than its minimum or fair share is
selected. However, if there is another job with a significantly
higher computing rate on the slot, the scheduler chooses that
job to improve overall performance. This is similar to the
―delay scheduling‖ [14] mechanism in Hadoop fair scheduler.
By using progress share, the scheduler can make an
appropriate decision. As a result, the cluster is better utilized.
In addition, each job receives a fair amount of computing
resources.
IV. CONCLUSION
The cloud environment provides heterogeneous hardware and
resource demands; therefore, it is important to exploit these
features to make a data analytics cluster in the cloud efficient.
In this paper, we presented the issues concerning the cloud
users and providers to allocate resources in a cost-effective
manner, and discussed a scheduling issues that should be
taken care of to provide good performance and fairness
simultaneously in heterogeneous cluster.
Beyond data analytics systems, many systems running in the
cloud involve multilevel scheduling. Resource allocation is
done at infrastructure level and job scheduling is performed at
application level. Mesos [15] takes this approach to provide
resource sharing and isolation across distributed applications.
It adopts Dominant Resource Fairness(DRF) [6] as an example
of resource allocation policy and leaves application level
scheduling to each application.
RFERENCES
[1] Patricia Takako Endo et al. :Resource allocation for
distributed cloud :Concept and Research
challenges(IEEE,2011),pp.42-46 .
[2] Lizhewang,JieTao,Kunze Castellanos ,A.C,Kramer
,D.,Karl,w, ‖High Performance Computing and
Communications‖,IEEE International Conference
HPCC,2008,pp.825-830.
[3] Hadoop. http://hadoop.apache.org.
[4] Hadoop fair scheduler. http://hadoop.apache.org/
common/docs/r0.20.1/fair_scheduler.html.
[5] I. Houidi et al., ―Virtual Resource Description and Clustering
for Virtual Network Discovery,‖ Proc. IEEE ICC Wksp.
Network of the Future, Dresden,bGermany, June 2009.
[6] Gunho Lee et al., ― Heterogeneity aware Resource Allocation
and Scheduling in Clouds‖, Yahoo Research.
[7] N. M. K. Chowdhury and R. Boutaba, ―A Survey of Network
Virtualization,‖ Computer Networks: Int’l. J. Comp. and
Telecommun. Networking, Apr.2010, pp. 862–76.
[8] A. Sheth, and A. Ranabahu, ―Semantic Modeling for Cloud
Computing, Part I,‖ IEEE Computer Society — Semantics &
Services, 2010.
[9] C. Chapman et al., ―Software Architecture Definition for On-
Demand Cloud Provisioning,‖ Proc. 19th ACM Int’l. Symp.
High Performance Distrib. Computing, 2010, pp. 61–72.
[10] Q. Zhang, L. Cheng, and R. Boutaba. ―Cloud Computing:
State-of-the-Art and Research Challenges,‖ J. Internet
Services and Apps., Springer, 2010, pp. 7–18.
[11] I. Houidi, W. Louati, and D. Zeghlache, ―A Distributed
Virtual Network Mapping Algorithm,‖ Proc. IEEE ICC, 2008,
pp. 5634–40.
[12] R. Urgaonkar et al., ―Dynamic Resource Allocation and
Power Managementin Virtualized Data Centers,‖ IEEE
NOMS), 2010.
[13] S. Agarwal et al., ―Volley: Automated Data Placement for
Geo-Distributed Cloud Services,‖ Proc. 7th USENIX Conf.
Networked Sys. Design and Implementation, 2010.
[14] M. Zaharia et al. Delay scheduling: A simple technique for
achieving locality and fairness in cluster scheduling. In
EuroSys, 2010.
[15] B. Hindman, A. Konwinski, M. Zaharia, A. Ghodsi, D.
Joseph, R. Katz, S. Shenker, and I. Stoica.Mesos: A platform
for fine-grained resource sharingin the data center. In NSDI,
2011
[16] A. Ghodsi, M. Zaharia, B. Hindman, A. Konwinski, S.
Shenker, and I. Stoica. Dominant resource fairness: Fair
allocation of multiple resource types. In NSDI, 2011.

Challenges in Dynamic Resource Allocation and Task Scheduling in Heterogeneous Clouds

Recommended

More Related Content

What's hot (17)

Similar to Challenges in Dynamic Resource Allocation and Task Scheduling in Heterogeneous Clouds (20)

More from rahulmonikasharma (20)

Recently uploaded (20)

Challenges in Dynamic Resource Allocation and Task Scheduling in Heterogeneous Clouds