Dynamic Load Balancing on Linux

Based Private Cloud

Md Kamran Kausar Final YearComputer Engg. Dept.Fac of Engg. & TechJAMIA MILLIA ISLAMIANew Delhi

Contents

• Cloud Computing.• Virtualization.• Live Migration.• Load Balancing.• Proposed Algorithm For Load

Balancing.• Conclusions and Future Work.• Programming Languages Use.

Cloud Computing

• Came into Existence around 1950, The underlying concept of cloud computing dates back to the 1950s, when large-scale mainframe computers became available in academia and corporations, accessible via thin clients/terminal computers, often referred to as "static terminals", because they were used for communications but had no internal processing capacities. To make more efficient use of costly mainframes, a practice evolved that allowed multiple users to share both the physical access to the computer from multiple terminals as well as the CPU time. This eliminated periods of inactivity on the mainframe and allowed for a greater return on the investment. 

• Basically virtual servers available over the Internet.• Anything we consume outside the firewall is "in the cloud," • Private cloud is the phrase used to describe a cloud computing

platform that is implemented within the corporate firewall.

Cloud Services

Cloud Rules Every Where

Virtualization

• The term "virtualization" traces its roots to 1960s mainframes, during which it was a method of logically dividing the mainframes' resources for different applications. Since then, the meaning of the term has evolved to the aforementioned.

• Virtualization is commonly defined as a technology act as a software abstraction layer between the hardware and the operating system and applications running on top of it. Main advantages include isolation,consolidation and multiplexing of resources. Other benefits of virtualization include saving on power by consolidation of different virtual machines on a single physical machine, migration of virtual machine for load balancing etc.

• Core of any virtualization is Hypervisory or VMM.It is software which allows each VMs to access and schedule the task

• Types:-1.Full Virtualization-Hypervisor controls the hardware resources and emulates it to guest OS.Guest Not require any modification. KVM is the best example of it

Virtualization Contd..

• Type 2- Paravirtualization: Here Hyperviosor controls the hardware resources and provides API to guest OS to access hardware. Guest OS requires modification to access the hardware resources. Xen is the example of paravirtualization technology.

• Top Virtualization Technologies:-

1.KVM Kernel-based Virtual Machine Implemented as loadable kernel module that converts the

Linux Kernel into a bare metal hypervisor. Implemented as regular Linux Process running by standard

Linux scheduler infact each virtual CPU appears as a regular Linux process.

This allows KVM to benifit from all features of Linux kernel2.QEMU:-process emulator and virtualizer, run many OS and programs,uses binary translation to achieve high performance

Live Migration

Moving the running VM on a physical machine (source host) to anotherphysical machine (target host) without disrupting any active network connections, while the VM is running on the source host, even after the VM is moved to the target host.

It is considered live, since the original VM is running, while the migration is in progress.

Very small downtime order of millisec, is the benefit of doing live migration.

Live Migration Steps

Pre-Migration:- Select VM to be migrated and destination host where resources required are guaranteed to be present.

Reservation:- Confirmation of necessary resources

Iterative Pre-Copy:- The guests memory is copied to the destination.

Live Migration Steps..

Stop and Copy:- VM at source is suspended and network traffic is redirected to destination host.

Commitment:- Destination host indicates source host that it has successfully received a consistent VM image.

Activation:- The migrated VM on destination host is now activated.

Load Balancing

Process of reallocating VMs on another host in the network in order to improve resource and network utilization.

Common goals are:- Maximizing throughput, minimizing response time, and/or minimizing communication time and avoiding the scenario in network that, some hosts are under-utilized and some over-utilized.

Goals of Load Balancing

To improve the performance substantially.

Fault tolerance in case of system failure.

To maintain the system stability.

To accommodate future modification in the system

Load Balancing Algorithms

Sender initiated : Algorithm initiated by Sender

Receiver initiated : Algorithm initiated by Receiver.

Symmetric : Combination of above two

Five phases of load balancing

I. Load Evaluation-It define Bands. Lightly loaded, moderately loaded and heavily loaded

II. Profitability Determination-If there exists one virtual machine in the heavily loaded band and one in lightly loaded band.

III.Work Transfer Vector CalculationIV.Task SelectionV. Task Migration

Policy Engine

Heart of load balancing algorithm.

Decides when to migrate virtual machines between hosts and runs asnormal virtual machine.

ALGORITHM for Load Evaluation

All the hosts send load information to policy engine: which

is responsible forload balancing decisions, after a predefined time interval which can be changed as per requirements.

Load Balancing Algorithms Flow chart

Conclusions and Future Work.

The work has proposed a policy engine to dynamically balance the load over the network. Originally the network was imbalanced. There were hosts in heavily as well as lightly loaded bands. After some iterations on the VM of the load balancing algorithm, all the hosts were balanced over the network i.e. all the hosts were balanced.

Cloud Computing is a vast area and load balancing plays a very important role in case of Cloud. The work has focused on CPU usage as load parameter that is applied to the Cloud Computing Technology.

There are still other parameters and approaches that can be applied to balance the load. The performance of the given algorithm can be increased by varying different parameters like memory usage, disk I/O, network load etc.

References

• KVM Kernel Based Virtual Machine Red Hat, Inc. 2014.• Ali M. Alakeel, A Guide to Dynamic Load Balancing in Distributed

Computer • Terry C. Wilcox Jr, Dynamic Load Balancing Of Virtual Machines

Hosted On Xen, Department of Computer.• 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.

• Youran Lan, Ting Yu, A Dynamic Central Scheduler Load Balancing Mechanism, Computers and Communications, pp 734-740, May 1995.

• Yi Zhao, Wenlong Huang, Adaptive Distributed Load Balancing Algorithm based on Live Migration of Virtual Machines in Cloud, Fifth International Joint Conference INC.

• WWW.PYTHON.ORG• WWW.RIVERBANKCOMPUTING.COM