dynamic load balancing over linux cloud
DESCRIPTION
DANAMIC LOAD BALANCING OVER LINUX PRIVATE CLOUD USING OWN ALGO, UI DEVELOP IN PYQT4, CENTOS,AND BASH IS USETRANSCRIPT
Dynamic Load Balancing
On Linux Private Cloud
Introduction .1
Load Balancing.2
Dynamic Load Balancing.3
Proposed Algorithm.4
Performance & Results Analysis.5
What done So Far.6
Overview
What’s Remaining 7
Programming Languages & References.8
Computing challenges to Businesses Modern IT providers will have to face the following challenges:
Hardware is Costly to ownCapacity Planning is complexBusiness agility is low
• Hardware is expensive
• Hardware requiresmaintenance
• Hardware needs to be replaced
• Hardware Becomes obsolete
• How many Servers required?
• How much CPU power?
• How much RAM ?
• How much Storage ?
• Loss of competitive edge
• Slow response tochanged IT
• Unchanged Traditional ITinfrastructure
• Resources cannot be modifieddynamically
What is the solution to
these challenges?
But What is Cloud
Computing?
Solution:- Cloud The only solution addressing all the business challenges
On demand, self- service, reliable, anywhere, any time and cost effective
You only pay for what you use
Respond instantly to changing IT
No need to buy, maintain & replace hardware
Eliminate all wastage by automatic capacity planning
Elastic Resources
Essential Characteristics of Cloud
Broad Networkaccess
On demand self service
Rapid elasticity
Measured service
Resource Planning
COMPUTINGCLOUD
Billing is metered and delivered as a
utility service
Capability to scale to cope with
fluctuating demands
Request driven pool of computing
resource.
Virtualized resources as a service to
businesses
Centralized Computing technology over the
Internet
Essential Characteristics
On demand, self- service, reliable, anywhere, any time and cost effective
This cloud model is composed of six essential characteristics, three service models and four deployment models
Did you know?
Cloud computing & Cloud Model Concept
This cloud model is composed of six essential characteristics, three service models and four deployment models
Source:- ESDS
“Cloud computing is a style of computing in which scalable and elastic IT-enabled capabilities are delivered as a service using Internet technologies”
- Gartner
The Cloud Model
Scaling in CloudVirtual machines within the cloud can expand and contract. This is called scaling.
UserUser User
Load Balancer
Virtual ServerVirtual Server Virtual Server
Virtual Server&
Policy Engine
Load Balancing
Important Factors consider while Developing Algo
• Collecting and managing System status information
• Estimation of load
• Comparision of load
• Performance of System
• Nature of work to be transferred
• Selection of hosts
1. Static Load Balancing.
• Priori Knowledge about the global status of distributed.
• Job resource requirements communication time are assumed
• Mapping of jobs to resources is not allowed to change after the load balancing has begun.2.Dynamic Load Balancing
• Based on the current sate of system.
• Tasks are allowed to move dynamically from an overloaded node to receive faster service
• Mapping of jobs to resources at any point of time.
Local Scheduling
1 2
3 Distributed Scheduling Policy
Load Distributing
Strategy
Determines how the CPU resources at a
single node is allocated among its resident processes
Distributes the system workload among the nodes through process
migration
4
• Process of reallocating VMs
• On another Host over the Network
• To improve both resource utilization and job response time
• Avoiding a situation where some nodes are heavily loaded while others are idle or doing little work
What is Load Balancing Type of Load Balancing
Optimize Performance
Reduce IT Capital
Expense
Reduce IT operational
expense
Increase Flexibility &
Uptime
Reduce Carbon
Footprint
Reduce Administration
overhead
Benefits from Load Balancing In Linux Cloud
This benefits pull more Business to live on cloud rather than on desktop
Five Phases of Dynamic Load Balancing
TaskMigration
Task Selection
Work Transfer Vector Calculation
Profitability Determination
Load Evaluation
VM that has CPU utilization is closer to the amount equal to Work Transfer Vector Calculation
This is done by transferring a VM from Heavily loaded to Lightly loaded band such that both system achieve a Moderately loaded band
Migration of virtual machine from one host to another if there exists one VM in Heavily loaded and one in Lightly band
All Host send load information to policy EnginePolicy Engine divide CPU utilization of host into 1.Lightly Loaded 2.Moderately Loaded3.Heavily Loaded Band
QEMU-KVM’s live migration feature is usedVirsh command is part of libvirt API
1
2
3
4
5
Policy Engine
Heart of load balancing algorithm.
Decides Every thing when to migrate
virtual
machines between hosts and runs as
normal virtual machine.
It can move itself to a different host like
any other virtual machine
Depending on load
1
2
367%
41%
Dynamic Load Balancing Algorithms & Flow chart
Bands of
CPU
Utilization
threshold = (α1+α2+α3+...+αn)/n mean = (αmin+αmax) / 2 diff = |threshold - mean|
if diff > 10 moderately loaded band = threshold ± diff
else moderately loaded band = threshold ± 10
where, αi= CPU usage of ith Host over a defined time in percentage and i =
1 to n
Calculation of moderate band:
•All VMs achieve or come into Moderately Load Band•WTVC = Threshold – CPU utilization of LLB host
Work Transfer Vector Calculation
•If there exists one VM in HLB and one LLB
Profitability Determination
1
4
2
3
KVM,QEMU Libvirt,NFS,NTP,SSH,CentOS7
Live Migration
Linux Cluster Up & Running
Modified CSLB AlgoPython,
Bash
1.Adaptive Distributed Load Balancing Algo.
2.Central Scheduler Load Balancing Algo (CSLB)
3.Modified CSLB
Implemetation Of Algo
Dynamic Load Balancing Algo
What’s Done So Far Project is decompose into following Phases
What’s Remaining User Interface to intract with the Software
4
Graphical User Interface
PERFORMANCE AND RESULTS ANALYSIS
OS:- Centos7,
CPU:- Intel i5-2400 3.1 GHz * 3,
Memory:- 3 GB
Dynamic Load Balancing Algorithm
Host 1 CPU Usage 42.4 %
Host 2 CPU Usage 19.07 %
Host 3 CPU Usage 62.18 %
S146.50
S29.60
S33.00
K148.50
K28.70
K323.80
A193.50
A224.80
A32.90
Dynamic Load Balancing Algorithm
Host 1 CPU Usage 42.4 %
Host 2 CPU Usage 19.07 %
Host 3 CPU Usage 62.18 %
S146.50
S29.60
S33.00
K148.50
K28.70
K323.80
A193.50
A224.80
A32.90
MLB LLB HLB
MLB=Thres±10MLB
31.21 51.21
Threshold=41.21Mean=40.62Difference=0.58
Dynamic Load Balancing Algorithm
Host 1 CPU Usage 42.4 %
Host 2 CPU Usage 19.07 %
Host 3 CPU Usage 62.18 %
S146.50
S29.60
S33.00
K148.50
K28.70
K323.80
A193.50
A224.80
A32.90
MLB LLB HLB
Threshold=41.21Mean=40.62Difference=0.58
MLB=Thres±10MLB
31.21 51.21W.T.V.C = 22.14
Dynamic Load Balancing Algorithm
Host 1 CPU Usage 40.99 %
Host 2 CPU Usage 53.33 %
Host 3 CPU Usage 29.00 %
S149.40
S210.10
S31.00
K149.00
K27.00
K324.50
A194.50
A223.40
A31.90
Dynamic Load Balancing Algorithm
Host 1 CPU Usage 40.99 %
Host 2 CPU Usage 53.33 %
Host 3 CPU Usage 29.00 %
S149.40
S210.10
S31.00
K149.00
K27.00
K324.50
A194.50
A223.40
A31.90
MLB HLB LLB
Threshold=41.10Mean=41.16Difference=0.06
MLB=Thres±10MLB
31.01 51.10
Dynamic Load Balancing Algorithm
Host 1 CPU Usage 40.99 %
Host 2 CPU Usage 53.33 %
Host 3 CPU Usage 29.00 %
S149.40
S210.10
S31.00
K149.00
K27.00
K324.50
A194.50
A223.40
A31.90
MLBHLB
LLB
Threshold=41.10Mean=41.16Difference=0.06
MLB=Thres±10MLB
31.01 51.10W.T.V.C = 12.10
Dynamic Load Balancing Algorithm
Host 1 CPU Usage 42.58 %
Host 2 CPU Usage 50.59 %
Host 3 CPU Usage 27.31 %
S151.40
S210.40
S31.00
K146.10
K21.80
K325.10
A195.10
A251.90
A32.10
Dynamic Load Balancing Algorithm
Host 1 CPU Usage 42.58 %
Host 2 CPU Usage 50.59 %
Host 3 CPU Usage 27.31 %
S151.40
S210.40
S31.00
K146.10
K21.80
K325.10
A195.10
A251.90
A32.10
MLBHLB
LLB
Threshold=40.16Mean=38.95Difference=1.21
MLB=Thres±10MLB
31.16 50.16
Dynamic Load Balancing Algorithm
Host 1 CPU Usage 42.4 %
Host 2 CPU Usage 19.07 %
Host 3 CPU Usage 62.18 %
S151.40
S210.40
S31.00
K146.10
K22.10
K325.10
A194.10
A251.90
A31.80
MLBHLB LLB
Threshold=40.16Mean=38.95Difference=1.21
MLB=Thres±10MLB
31.16 50.16W.T.V.C = 12.85
Dynamic Load Balancing Algorithm
Host 1 CPU Usage 41.17 %
Host 2 CPU Usage 35.77 %
Host 3 CPU Usage 38.78 %
S152.20
S210.60
S31.00
K141.80
K21.00
K324.70
A195.60
A261.90
A31.70
Dynamic Load Balancing Algorithm
Host 1 CPU Usage 41.17 %
Host 2 CPU Usage 35.77 %
Host 3 CPU Usage 38.78 %
S152.20
S210.60
S31.00
K141.80
K21.00
K324.70
A195.60
A261.90
A31.70
MLB MLB MLB
Threshold=38.57Mean=38.47Difference=0.1
MLB=Thres±10MLB
28.57 48.57
KVM Kernel Based Virtual Machine Red Hat, Inc. 2014.1
Ali M. Alakeel, A Guide to Dynamic Load Balancing in Distributed Computer 2
Terry C. Wilcox Jr, Dynamic Load Balancing Of Virtual Machines Hosted On Xen, Department
of Computer.3
Jyotiprakash Sahoo, Subasish Mohapatra, Radha Lath,Virtualization: A Survey On Concepts,
Taxonomy And Associated Security Issues, Second International Conference on Computer
and Network Technology, 2010.
4
Youran Lan, Ting Yu, A Dynamic Central Scheduler Load Balancing Mechanism, Computers
and Communications, pp 734-740, May 1995.5
Yi Zhao, Wenlong Huang, Adaptive Distributed Load Balancing Algorithm based on Live
Migration of Virtual Machines in Cloud, Fifth International Joint Conference INC.6
References :-
7
8