Load Balancing in Mobile Cloud Computing

Abstract

Now a days, Cloud computing is rapid growing and online demand services in computing environment. To control the traffic of demands or services from users on servers by allocating resources need to share the services by servers. For that Load balancing (sharing of load between servers) is needed. When multiple users send requests to a website the traffic of demand from servers increases, we need load balancing to allocate the request to servers using algorithms like round robin, IP hashing, Least connections.

We will be covering what is load balancing, features of load balancing, load balancing algorithms, and sticky session.

Introduction

With a whimsical advancement of the convenient applications and moving of dispersed registering thought, the Mobile Cloud Computing (MCC) has transformed into a potential building for the adaptable organization for customers. Mobile Cloud computing as a unity of distributing computing development with PDAs to make the PDAs resource full in conditions of computational power, memory, storing, essentialness, and association preparation.

Get quality help now

KarrieWrites

Verified writer

Proficient in: Technology

5 (339)

“ KarrieWrites did such a phenomenal job on this assignment! He completed it prior to its deadline and was thorough and informative. ”

+84 relevant experts are online

Hire writer

Portability distributed computing is the delayed consequence of interdisciplinary strategy exemplifying portable registering and distributed computing. Likewise, this trans-disciplinary field is generally called mobile cloud handling. The issue we will be discussing in this paper load balancing is about distributing the work load into entire cloud. Here load balancing could be centralized or even decentralized.

Mobile Computing

The term mobile computing is generated with the idea of cloud computing. We can say mobile cloud computing is a part or a subset of cloud computing. Today mobile or cell phone face many issues with their assets like battery life, storage, data transfer.

Mobile cloud computing helps to face these issues as these issues influence the control of the quality greatly.

MCC implies the foundation where the data storing and the data transferring takes place outside the mobile or cell phone. It very well may be thought as an issue of the distributing processing and adaptable condition. The cloud can be used for power and limit, as PDAs don't have convincing resources appeared differently in relation to standard retribution contraptions.

Load Balancing

It is the way toward redistributing the total heap of a distributing framework into individual hubs to guarantee that no node is over-burdened and no nodes were under stacked or inactive. So in a cloud situation heap balancing guarantees that no virtual machines are over-burden, where some virtual machines are under stacked or doing almost no work. Load balancing attempts to accelerate the execution time for utilizations. It moreover ensures the framework strength.

Features:

1. It can distribute the incoming traffic into the network by effectively distributing among multiple servers.
2. Reliability and high availability can be maintained with re-directing requests sent by users to only the servers which are readily available.
3. It is easy to use by adding and detaching servers in the network as per request.

Algorithms

There are many algorithms responsible for load balancing in this paper we will be discussing about 3 major algorithms namely:

1. Round Robin- Sequential demand distribution
2. Least connections- Request will be sent to the least used server in the cloud
3. IP Hash – Request will be sent to the server due to the client IP.

Round Robin

Round robin burden adjusting is a basic method to disseminate customer demands over a gathering of servers. A customer solicitation is sent to every server thusly. The calculation educates the heap balancer to return to the highest priority on the rundown and rehashes.

Simple to actualize and conceptualize, round robin is the most broadly conveyed burden adjusting calculation. Utilizing this technique, customer solicitations are steered to accessible servers on a repeating premise. Round robin server burden adjusting works best when servers have generally indistinguishable figuring abilities and capacity limit.

How Round Robin Works

More or less, round robin system burden adjusting turns association demands among web servers in the request that solicitations are gotten. For a disentangled model, expect that a venture has a group of three servers: Server 1, Server 2, and Server 3.

1. Here first request will be sent to server 1
2. Second request will be sent to server 2
3. Third request will be sent to server 3 ..the same cycle continues

The load balancer keeps passing solicitations to servers dependent on this request. This guarantees the server load is circulated equally to deal with high traffic. Advantage for the round robin is lesser memory and disadvantage is sends high traffic to hubs without respect for conveyance.

Least Connections

Here, request will be sent to the server which is processing the least number of resources or which has least requests. In order to do this the load balancer should now which server has least requests, so it might do some additional computing. It will take a while for the load balancer to know this server which has the least connections.

At the point when a virtual server is arranged to utilize the least association load balancing calculation (or technique), it chooses the administration with the least dynamic associations. This is the default technique, in light of the fact that, much of the time, it gives the best execution.

At the point when a virtual server utilizes the least connection technique, it thinks about the holding up connections as having a place with the particular administration. Along these lines, it doesn't open new connections to those administrations. Disadvantages, high usage of memory and CPU impression for association following.

IP Hash

IP hashing is useful when let’s say a client sends a request and this clients request should especially go to a set of servers so that is when IP hashing is used. Redirection can be done based on the clients IP address, only those servers which are especially close to the clients IP address will be redirected. Request should be given reference in order to reach the required server. Typical example can be given as- consider IRCTC website once logged in client should reach the same server if logged in again to retrieve data regarding his/her train information or other data. Here people will be redirected to a bunch of server farm to have more performance.

Source IP Hash load adjusting utilizes a calculation that takes the source and goal IP address of the customer and server to create a one of a kind hash key. This key is utilized to assign the customer to a specific server.

Sticky Session

It can be considered as a process in which a heap balancer makes a fondness between a customer and a particular system server for the length of a session, (i.e., the time a particular IP spends on a site). Utilizing sticky sessions can help improve client experience and advance system asset utilization.

With sticky sessions, a heap balancer allots a distinguishing credit to a client, regularly by issuing a treat or by following their IP subtleties. At that point, as indicated by the following ID, a heap balancer can begin steering the majority of the solicitations of this client to a particular server for the length of the session. Without session ingenuity, the web application would need to keep up this data over various servers, which can demonstrate wasteful—particularly for enormous systems.

Server literally keeps the session ID when the first time logged in and the same session id is given every time the request is sent. The cache should store the session information. By considering the figure above we can say that load balancers support the sticky sessions by giving a session id and making the request go to the same server for that particular id.

Optimization

Optimization in load balancing represents that performing or assigning or satisfying the user needs in less time with more accuracy (scalability and through put). In this, we have provided a new technique for optimization of load balancing based on the average of server load and users incoming for the needs. Lets have an assumption that cloud service providers are connected by ‘n’no of clients through internet , virtual machine is consisted by service provider for managing the load and there ‘m’no of servers which are considered as pool resources. Two walks in one cycle is completed by agent at the pool of hosts.

In the first walk, it gathers all from all servers by moving from first sever to the last server for settling on choice for load balancing. In the second walk, It uses standard deviation for balancing the host loads. Here we also used the Master-slave decision algorithm for transferring of jobs from high loaded server to the underloaded server and to manage the new jobs which re being arriving. At the end, Master saves all the information which is received from slave (gets information from agent).

Agent Walk

This model is explained in next 5 steps:

• Step-1: No of jobs in queue on particular server is found by agent who is activated in random server.
• Step-2: This process is repeated for all the servers which are in shared pool by the agent.
• Step-3: ‘Average’ is calculated based on the jobs on servers.
• Step-4: Based on the ‘Average ‘ status of the server is found that either overloaded or underloaded.
• Step-5: status of the server is decided as, ‘Average’ the status is overloaded.

Below ‘Average’ it is underloaded.

Mean can be calculated as M = summation of (𝑢𝑖𝑗 )

𝑀𝑖

𝑗=1

Variance is calculated as 𝑉 = 1 𝑃

Summation of ( 𝑋𝑘 − E) 2

p 𝑘=1

Where 𝑋𝑘, is sum of utilization of VM

If we calculate the Average of jobs in the servers is not sufficient to determine the status of the server then standard deviation is calculated.

PU = T+M*StdDev

If the utilization of the host is less than PU then it is overloaded otherwise, it is underloaded and the information is saved by slave. Later, information saved by slave is transferred to Master.

Agent Walk 2

Backtracking model is used by Agent. Agent backtracks from last server to first server by checking the status of each server. If any server is overloaded then jobs of particular server is send to the underloaded server, till agent reaches first server.

Here Master-slave mechanism is recommended. Information of all the slaves is containing by Master. First, Agent is at master and ready to assign new jobs and to balance the load on slaves. Agent checks the status of the slave in master by having the state and its id. It checks the condition that, if host_slave_state = overloaded then salve receives a transfer request from Master. Jobs has to transfer to underloaded host if slave sends the positive condition. If the host_slave_state = underloaded then new jobs are assigned to the host. This process is repeated to all servers until agent comes to the first server.

Conclusion

Load balancing is a significant issue in cloud computing that involves distributing workloads to prevent server overload and underutilization. This paper discussed load balancing algorithms and introduced an optimization technique for efficient load distribution. Future work includes implementing this technique in real-time cloud environments and designing more efficient algorithms to improve scalability, throughput, and resource utilization.