3. PROPOSED TECHNIQUE
3.1 WHALE OPTIMIZED MULTICRITERION CORRELATION TECHNIQUE FOR RESOURCE AND COST OPTIMIZATION IN CLOUD
Service provisioning is a significant and demanding issue in the widely distributed systems such as cloud computing environments. A cloud service offers a large flexibility in provisioning to balance the user’s required services. A service provider delivers the users required services to users through the internet. During the service provisioning, the resource utilization is a main demanding issue in the cloud. In general, the cloud includes a number of resources like energy, memory, bandwidth, cost and so on. The cloud server renders the different services to the user with a minimum level of resource utilization. Based on this motivation, an effective optimization technique called Whale Optimized Multicriterion Correlation (WOMC) is developed for resource and cost optimization in the cloud. Architecture diagram of the WOMC technique is shown in figure 1.
Figure 1 describes the architecture diagram of WOMC technique. The number of cloud users sends their requests to cloud server (CS). The cloud service provider (CSP) provides different services to end users with less resource utilization.
The WOMC technique initially measures the correlation between the number of users requested services and already available services in a cloud server. This correlation measurement is based on the multiple criterions in which the user requested services has certain resource criterion. Based on the correlation, the user required services are identified. After that, optimal resources are identified for rendering the user requested services in an effective manner. A brief explanation of the WOMC technique is described in following sections.
3.1.1 Multicriterion Correlation Measure
The first step in the design of WOMC technique is to perform correlation between the cloud users requested services and availability of services in the cloud server. WOMC technique uses multicriterion correlation that which is used for finding the mutual relationship between the users requested service and number of available resources in the cloud. A correlation is a statistical method that provides strong pairs of variables (i.e. services and the availability of services) which are related to each other. Multicriterion correlation analysis is carried out where the relationship between the variables are measured based on more than one criterion. The multiple criterions such as energy, bandwidth, cost, memory are used for correlation measures. Initially, the user sends a number of requests to cloud server. The cloud server measures correlation with user required resource criterion. The resources utilized in the cloud service provisioning are energy, bandwidth, memory and cost. Energy is an amount of power utilized by virtual machine for providing the services to user with the specific time. The energy is computed using following mathematical equations,
From (1), ‘E’ denotes energy for resource provisioning which is measured in terms of a joule (J). represents total power consumption for providing the services and denotes a service provisioning time. The other criterion is the bandwidth which is measured as the average rate of successful request transfer through a communication path. It is measured in the unit of bit per second (Bps).
From (2), denotes a bandwidth and ‘T’ represents time. The memory consumption is a total amount of storage space consumed by virtual machine for rendering the multiple user requested services. It is measured in terms of mega bytes (MB). Cost is the other major criterion which is the amount of time taken by the cloud server for responding the user requested services. Based on the above said resources, the user requests the services with less resource utilization. With the multiple criterions such as minimum energy utilization, memory consumption, cost and bandwidth utilization, the WOMC technique measures the correlations as follows,
From (3), denotes a correlation coefficient, represents a number of services in cloud. consitutes a user requested services with multiple criterions, is the available services. The correlation coefficient provides the values from -1 to +1. As a result, the relationship between the numbers of user requested services and already available resources are identified effectively in the cloud of a server for minimizing the response time. The algorithm of multicriterion correlation is illustrated as follows.
Algorithm 1 Multicriterion correlation algorithm
Input: Number of user request available services
Output : Identity the user requested services
1.For each user request
2. Measure the multicriterion correlation using (3)
3. if correlation coefficient then
4. positive correlation
5. else if correlation coefficient -1 then
6. Negative correlation
7. else if correlation coefficient 0 then
8. No correlation
9. end if
10. end for
Algorithm 1 provides the correlation measure of user requested services and available services based on the multiple criterions. By applying the correlation, the coefficient offers three dissimilar results. The =’+1′ means that the perfect relationship between user required services and available services. ‘0’ indicates the no correlation and ‘-1’ represents a negative correlation between requested services and available services. As a result, the user required services are identified. This in turn minimizes the response time of user requested services.
3.1.2 Nature-inspired Multicriterion whale optimization for optimal resource utilization
After identifying the user required services through the correlation measure, the optimization is carried out for selecting a virtual machine which utilizes optimal resources for providing the user services. This process is done by using nature-inspired multicriterion whale optimization. By applying this optimization technique, the cloud server offers efficient services with optimal resource utilization based on multiple criterions such as minimum energy consumption, less bandwidth utilization and memory consumption and less computational cost. In whale optimization, the whale is a very intelligent animal with emotion. The whale optimization is inspired by the distinctive hunting nature of humpback whales. The whale searches the location of the prey and spins around the prey to create different bubbles along a path. In nature-inspired multi criterion whale optimization, the whales are considered a number of virtual machines ‘ ‘. A CSP offers computing services to cloud users through a virtual Machine ( ). User requested services are considered as prey. As a result of optimization, the best solution (i.e. virtual machine) is selected for providing user requested services with less resource utilization. The flow process of Nature-inspired Multi criterion whale optimization is described as follows,
Figure 2 shows the flow process of optimization algorithm to identify resource optimized virtual machine for providing the user requested services from the number of available resources in the cloud.
Figure 2 Flow process of Nature-inspired Multi criterion whale optimization
The numbers of whales (i.e. VM) are initialized. The fitness of VM is calculated and it satisfies the fitness condition for selecting the best solution. The fitness condition is expressed with multiple criterions as follows,
From (4), denotes an energy consumption, represents a bandwidth utilization, means memory consumption, CC refers to a computational complexity. arg min function attains the smallest value of all the resources of virtual machine. From (4), FC denotes fitness condition of the virtual machine to provide the services with minimum resource utilization. Based on the fitness condition measure, the current best VM is selected. After that, three processes are carried out such as encircling prey, bubble-net feeding method and searching the prey. Followed by, an optimal resource optimized virtual machine is selected for providing the user requested cloud services. In encircling prey phase, the whale identifies the location of prey and surrounds them. Because the location of an optimum designs in the search space is not identified previously. For that reason, the optimization algorithm considers that the present solution is an optimal. After identifying the current best solution, the position updates are carried out for comparing the current best solution with other solution (i.e. whale) to find the resource optimized virtual machine. This updating behavior is represented as follows,
From the equation (6), d denotes a distance between the position vector of the prey and whale ( . represents a coefficient vector, denotes a current iteration. From (5), denotes an updated position and represents a coefficient vector and it is expressed as follows,
From (7), ‘b’ is linearly reduced from 2 to 0 over the course of iterations and ‘R’ indicates a random vector in the range values 0, 1. After that, the bubble-net behavior of whales are designed based on shrinking encircling approach and spiral updating position. Bubble-net behavior refers the foraging behaviors of the whales. In the first approach, the variation range of ‘a’ is also decreased by b. The random values for ‘a’ are assigned as ?1, 1, the new location of the whale is identified in any place in between the initial position of the whale and the position of the present solution. In the second approach, the distance between the position of the whale and prey is calculated. A spiral equation is then generated among the position of whale and prey to mimic the helix-shaped motion of humpback whales. Then the current best solution is updated and compared with other whales to find the global optimum. The updating results are expressed as,
From (10), denotes an updated distance among whale and ‘p’ is a constant for describing the structure of the logarithmic curve. Exponential function ‘e’ is the base of natural logarithms. From (9), ‘q’ is the random number and their ranges are -1, 1. By performing the optimization, positions of the whale is updated as follows,
From (11), ‘P’ is a random number in 0, 1, finally, global optimum values are obtained by updating the whale’s position with randomly chosen whale (i.e. VM) rather than the current best whale. The updating behavior is expressed as follows,
From (13), denotes a random position vector of a random whale. Finally, updated VM is a best solution for providing the user required services. The algorithmic explanation is presented as follows,
Input: Number of resources, number of user requests (i.e. prey), Virtual machines (i.e. whales)
Output: Select resource optimized virtual machines for providing user requested services
1. Initialize the whale’s populations
2. Initialize the value of a, b, R, p, q
3. for each whale
4. Calculate the fitness condition for selecting the current best solution using (4)
5. if (P