*2.2.1.2 Agent selection policy*

The selection policy handles which Worker Agent is migrated whenever there is a necessity. We assign a numerical value, called credit, to every Worker Agent. The credit value designates the capacity of the agent to remain undisturbed in case of migration. For a Worker Agent, the higher its credit value, the higher its opportunity to stay at the same machine. In other words, its opportunity to be selected for migration is lower. The LBC Agent assigns credit to each Worker Agent and chooses which agent requests to be migrated using the credit. Any Worker Agent with high credit will be given more opportunity to preserve its current location (the resource the agent resides in) with less opportunity to be selected for migration.

6.The communication load is small.

*Agent Based Load Balancing in Grid Computing DOI: http://dx.doi.org/10.5772/intechopen.94219*

8.The needed Resource is available.

1.The Worker Agent's load Increases

exchanges rises the Worker Agent's load.)

it is not available. The final equation can be written as:

þ b8Pi1 þ b9Pi2 þ b10Si

excepted from the list of receiver resources.

less subject for migration.

to be migrated.

**61**

stay rather than being migrated.

9.The agent has high priority.

7.The communication path between hosts is not reliable.

In contrast, the credit value of Worker Agent reduces in the following situations:

2.The communication with Worker Agents in other resources is increased.

Using a multiple linear regression operation, we will try to gather all the mentioned factors into one equation, In linear regression, the relationship between a dependent variable, Y, and an independent variable X, is modeled by Y = a + βX. This interpretation of coefficient, it is appropriate only when the independent variable is continuous (quantitative). To incorporate qualitative independent variables into the regression model and formulate the model so the variables have interpretable coefficients, There are two commonly used methods for coding qualitative variables so they can be used in regression models, dummy coding and effect coding [4, 5]. To include the variables qualitative in the equation, we will use the simplest way which is the dummy coding, which assigns values "1" and "0" to reflect the presence and absence, For example, if we take a qualitative variable Resource availability Ri, we assigned value 1 when the resource is available and 0 if

3. Strong mobility or instant exchanges of messages (frequent message

Credit Ai ¼ b0 þ b1Wi þ b2Ui þ b3Ri þ b4Ldi1 þ b5Ldi2 þ b6Hi1 þ b7Hi2

tional load value will be excepted from the list of receiver resources.

Having a big coefficient means that this variable will make the agent tends to

b2: Communication load Coefficient: if b2 has a negative value, then we can assume that an agent has a big communication load, it is more matter to migration as its credit value will be small. Since b2 has the smallest weight among the regression coefficients then it has the weakest effect on the credit value and therefore the migrating agent selection. If b2 is a positive value, so when it is multiplied by the communication load, a resource that has a big communication load value will be

b3: Resource Availability Coefficient: when Ri = 1, b3 has relatively large values, that means that the agent finds the needed resource on the running host thus it is

b4, b5: Host Load Coefficient: b4 has a higher load than b5 because when the running host is underloaded or balanced, it is less matter to select one of its agents

b1: Computation load Coefficient: if b1 is a relatively large negative value then an agent having a big computation load is more likely to be migrated as its credit value will be reduced. If b1 is a positive value then the resource that has a big computa-

(2)

The credit value of a Worker Agent is assumed to depend on two types of parameters, namely Worker Agent dependent parameters and System dependent parameters [4]:

## *2.2.1.3 Worker agent dependent parameters*

1. Its computational load, as it represents the main source of resource loading.

