Article Preview
TopIntroduction
Dissipating resource benefits and the ability to share resources and time-limited usage are two key advantages that cloud computing provide to the Information technology (IT) business. Models and tools that produce new system requirements are often used to apply effective analysis to predict the most sufficient quantity of resources for every request. Live virtual machine migration allows users to reorganise resources in the data centers (DC) without affecting the company's services. Using the term “live VM migration,” one can move an entire system and its related software through one host to the other without disturbing the customer's experience. As part of resource management, it provides web service, bandwidth allocation, and power usage. It has become increasingly challenging to manage and distribute resources during VM migration because of the high dynamism of cloud services, resources flexibility, and guaranteed uptime and consistency. Large, virtualized data centers are fully dependent on the design and network interactions between VMs. Ethernet connections can be reduced by reducing number of migration. Operating costs and network demand must be reduced in a cloud computing environment. Changes in Cloud Data Centers (CDC) resource consumption might occur when new VMs are added or hosts fail, or when existing VMs are removed from the CDC. Load balancing or server consolidation may be necessary based on the EC and other Quality of Service (QoS) factors.
The energy efficiency in CDC is broken down into four categories:
- 1.
Multiple VMs must be switched to another host to meet SLA if a host is overloaded.
- 2.
For underloaded hosts, migrate all VMs and sleep the inactive hosts.
- 3.
Choose the VMs to be relocated from the overloaded host.
- 4.
VM placement and migrations addresses the relocation of chosen VMs.
This paper offers an adaptive upper threshold method for predicting overloaded hosts. The adaptive threshold is based on statistical data acquired over the lifespan of VMs. Because workloads are dynamic, the data may contain non-normal distribution outliers. The proposed technique detects an overloaded host based on CPU utilization using an adaptive upper threshold (UTh) approach. When the overloaded host detection process is started, it compares the actual CPU utilization to an adaptive upper threshold (UTh). If CPU utilization is greater than the Upper threshold (UTh) then the host is detected as overloaded. After determining an overloaded host, VM selection and VM placement continue as normal.
The following is a brief description of the paper contributions:
- •
A VMC is considered as a MOOP aiming to reduce energy consumption, SLA and number of VMs migration.
- •
This study focuses on determining the upper threshold based on historical Central Processing Unit (CPU) consumption using the adaptive threshold approach.
- •
The modified Iqr host overload detection algorithm outperform than existing overloaded host detection algorithm (i.e. Iqr, Lr and DVFS) algorithm, in terms of EC, SLA and the number of VMs migrations.
TopLiner prediction model and heuristics approaches for regulating number of migration are presented by Ferreto et al. (2011). The strategy was compared to previous eager-migration-based methods using workloads from TU Berlin and Google CDC. Migrating VMs with consistent capacity decreases migrations without increasing physical server count.
Bing Wei et al. (2011), proposed a method for performing live migration in a CDC. The energy-aware migration model looks at the actual migration procedure and the linear relationship between energy use and shifting VM resources. The proposed approach has been demonstrated to reduce EC while maintaining performance. Migration times and EC are reduced by increasing the number of co-migrations VMs.