1. A SPECIAL TALK ON CLOUD COMPUTING : BASIC CONCEPT
REGARDING LOAD BALANCING AND VM SCHEDULING
USING
SIMULATION TOOLKIT
Presented By:
Soumen Santra
Assistant Professor
DEPT. OF MCA
Monday, May 11, 2020 Soumen Santra 1
3. Introduction
The cloud architecture utilizes a three-dimensional model that contents several
processing tiers: Infrastructure as a Service (IaaS), Platform as a Service (PaaS),
Software as a Service (SaaS), etc.
It is necessary to implement static as well as dynamic computing.
Load balancing is a challenge for a distributed system.
All the load balancing algorithm are not always practically feasible or cost efficient.
Our aim is to provide an approach for load balancing .
Monday, May 11, 2020 Soumen Santra 3
4. Cloud Component
Cloudlets :It relates with cloud-jobs which is given by the Clients or End users .End
users interact with the clients to manage information related to the cloud.
Clients: They generally fall into three categories as given in
ď‚· Mobile: Windows Mobile Smartphone etc.
 Thin: They don’t do any computation work. Servers do all the works for them. Thin
clients don’t have any internal memory.
ď‚· Thick: These use different browsers to connect to the Internet cloud.
Datacenter: It is nothing but a collection of servers hosting different applications. A end
user connects to the Datacenter to subscribe different applications.
Distributed Server: It is the host which is the part of a cloud which are present
throughout the Internet hosting different applications. Content
Monday, May 11, 2020 Soumen Santra 4
5. Cont..
Virtual Machines (VM): It is a system which processes the cloud job. The parameters or attributes
are image size, VM memory, MIPS rate, bandwidth, number of cpus and virtual machine monitor.
Monday, May 11, 2020 Soumen Santra 5
6. Types of Clouds
Based on the domain or environment in which clouds are used, clouds can be divided into
3 categories :
Public Clouds
Private Clouds
Hybrid Clouds (combination of both private and public clouds) .
Dynamically delivers complete package of services on user demand.
These services broadly classified into three types:
Infrastructure as a Service (IaaS) - Twitter etc.
Platform as a Service (PaaS) - Amazon's Elastic Compute Cloud (EC2), Facebook etc.
Software as a Service (SaaS) – Web Mail , MS Office Live etc.
Monday, May 11, 2020 Soumen Santra 6
7. Load Balancing
Load balancing in Cloud environment are based on current running phase of the system.
It can be done in two different ways: distributed and non-distributed.
In load balancing a common objective, known as cooperative, improve the throughput of the
system, overall response time, etc. [4].
Distributed dynamic load balancing can affect the whole system.
Node needs to interchange status information with every other node in the system.
Types of Load scheduling algorithms:
Symmetric: Allocation by the scheduler to every VM with respect of incoming jobs.
Asymmetric: Allocation by the scheduler to every VM with respect of incoming jobs.
Static: It doesn't depend on the current state of the system.
Dynamic: Decisions on load scheduling are based on current state of the system.
NOTE : FCFS, RR, SJF etc.
Monday, May 11, 2020 Soumen Santra 7
8. Cont..
• Load balancing is the process
of distributing the load among
various machines of a
distributed system.
• Virtual machines (VM) to
improve resource utilization
and job response time while
some of the vm’s are heavily
loaded while other vm’s are
idle or doing very little.
• Load balancing ensures that
all the processor in the system
or every vm in the network
does approximately the equal
amount of work (MIPS rate).
Monday, May 11, 2020 Soumen Santra 8
9. Basic Strategies for VM Migration in Load
Balancing
Policies or Strategies in load scheduling :
Transfer strategy: It selects a job for transferring from a local node to a remote node [6].
Selection strategy: It specifies the processors, to exchange with proper matching.
Location strategy: Find out the location of remote node.
Information strategy: It responsible for collecting information about the nodes in the
system.
Communication strategy: It communicates over federated network among various
locations.
Monday, May 11, 2020 Soumen Santra 9
10. An Approach on Load Balancing Due to
Probabilistic VM Migration Policy:
1. Initialize: Initialize P, the probability vector: P [i] for each VM where i is the number of
Cloud job (cloudlet), i = 1, 2,…., n .
2. For each Host:
(a) Generate: Generate VM each individual by sampling the space using the
probability Vector P.
(b) Evaluate: Evaluate each computing jobs for each VM, assign B, a set of
MaxCPUUtilized with MinCPUAllocation VMs’ to the best individual found.
(c) Update P: Update the vector P based on the best individual B:
P [i+1] = (1 - α) P [i] + α B [i]
where α is the pre-defined learning rate.
(d) Return to step 2a
where α is the probability vector assign to 10% . VM migration treated as
10% and remaining 90% are allocated for next incoming customer.
Monday, May 11, 2020 Soumen Santra 10
11. Different Approach of Load Balancing
• Dynamic load balancing algorithms follow distributed nature.
• The nodes in the system needs to interact with every other node.
• Distributed system affect by each node to interchange status information with
every other node.
• Centralized way, Non-distributed system is executed by a central node in the whole
system. The other nodes interact only with the central node.
• In semi-distributed system, nodes are partitioned into clusters, where each cluster is
of centralized form.
• A central node is elected in each cluster by appropriate election technique.
Monday, May 11, 2020 Soumen Santra 11
12. Cloud Simulator : CloudSim3.0
A new, generalized, and extensible simulation framework that allows seamless modeling,
simulation, and experimentation of emerging Cloud computing infrastructures and
application services.
Support for modeling and simulation of large scale federated Cloud environment .
Virtualized and customizable scalable, dynamic scheduled ,low energy consumed
computational resources.
Monday, May 11, 2020 Soumen Santra 12
13. Cont..
Here using Cloudsim3.0 we implemented RoundRobin scheduling policy for VM. In CloudSim3.0
normally overriding two class VmSchedulerSpaceShared and VmSchedulerTimeshared.
We can implemented FCFS and RoundRobin respectively But here in this project we may do same
thing using overriding few classes like Datacenter, DatacenterBroker, Host,Cloudlet, CircularHost,
RoundRobin , RoundRobinVmAllocationPolicy etc.
Monday, May 11, 2020 Soumen Santra 13
14. Cont..
• Here VmSchedulerTimeShared and VmSchedulerSpace Shared are represented as
round robin and FIFO.
• These two classes extends another class VmScheduler to allocate the vm for a
datacenter to allocate for customer.
Monday, May 11, 2020 Soumen Santra 14
19. Result
• Analysis of simulation :
• Number of VM : 5
• Number of CLOUDLET : 20
Monday, May 11, 2020 Soumen Santra 19
20. Cont..
• Number of VM : 5
• Number of CLOUDLET : 20
Monday, May 11, 2020 Soumen Santra 20
21. Result
• Number of VM : 5
• Number of CLOUDLET : 20
Monday, May 11, 2020 Soumen Santra 21
22. Result
• Number of VM : 5
• Number of CLOUDLET : 20
Monday, May 11, 2020 Soumen Santra 22
23. Result
• Number of VM : 5
• Number of CLOUDLET : 20
Monday, May 11, 2020 Soumen Santra 23
24. Result
• Number of VM : 5
• Number of CLOUDLET : 20
Monday, May 11, 2020 Soumen Santra 24
25. Result
• Analysis of simulation :
• Number of VM : 5
• Number of CLOUDLET : 20
Monday, May 11, 2020 Soumen Santra 25
26. Cloud Simulator : CloudReport
II. CloudReports:
• CloudReports is a graphic tool that simulates Cloud computing environments.
• It uses CloudSim as its simulation engine even we can add Netbeans/Eclipse also.
• Easy-to-use user interface, report generation features and creation of extensions.
• The application simulates an Infrastructure as a Service (IaaS) provider with an
arbitrary number of customized datacenters.
• Includes processing capacity, amount of RAM, available bandwidth, power
consumption and scheduling algorithms (Round Robin).
• HTML reports of each simulation to explain log report.
Maven : An accumulator tool which manages the code
Allows proper migration from one platform to another platform
Monday, May 11, 2020 Soumen Santra 26
31. Conclusion
We have seen the VM load balancing algorithms work in cloud environment
those are implemented in CloudSim or CloudReport type simulators .
We conclude that if we select an efficient virtual machine then it effect the overall
performance of the cloud Environment and also decrease the average response time.
Monday, May 11, 2020 Soumen Santra 31
32. FUTURE SCENARIO
• NEW DELHI: The cloud computing market in India is expected
to grow at 40 per cent by 2014, from an estimated $66.7
million in 2009, driven by cost and performance efficiencies,
IT research firm IDC said.
• According to IDC's 'India Cloud Computing Market: Current
State and Future Roadmap Study 2010', the Indian public
cloud computing market was estimated to be $66.7 million in
2009.
• "A significant segment of the market is aware of the concept,
but do not really understand what it actually means or how it
can benefit them," IDC India Lead Analyst (User Research)
Indranil Dutta said.
Monday, May 11, 2020 Soumen Santra 32
34. Cont..
•Optimization of collective communication operations in MPICH, R. Thakur, R.
Rabenseifner, and W. Gropp, in International Journal of High Performance
Computer Applications, vol. 19, no. 1, 2005, pp. 49–66.
• A Comparative Study into Distributed Load Balancing Algorithms for Cloud
Computing, Martin Randles, David Lamb, A. Taleb-Bendiab, 2010 IEEE 24th
International Conference on Advanced Information Networking and Applications
Workshops.
• A Guide to Dynamic Load Balancing in Distributed Computer
Systems, Ali M. Alakeel, IJCSNS International Journal of Computer Science and
Network Security,VOL.10 No.6, June 2010.
• CONTINUOUS DYNAMIC OPTIMISATION USING EVOLUTIONARY
ALGORITHMS by TRUNG THANH NGUYEN The University of Birmingham.
• A Survey With Proposed Approach On Dynamic Load Balancing: via VM Migration
Strategy In Cloud Computing in Amity University, CONFLUENCE 2013
Monday, May 11, 2020 Soumen Santra 34