Migrating to a Public Data Center: Resource Management Case Study

Abstract

Most businesses are migrating to public data centers IT infrastructure or implementing their own data centers to provide Information technology services with cheaper operational costs. Data Centers are essential in delivering the provision of flexible resources to accommodate the growing data traffic workload demand. The paper focuses on highlighting various data center resource management of using a case study of an organization migrating to a data center public system. The paper identifies significant issues with current data centers, such as overprovisioning. The primary essential components in resource allocation when designing an optimization data center is also address as well as a presentation of various benefits expected to be achieved by accurate prediction of workload in resource management. To come up with a better data center design, a comprehensive analysis of the existing resource management categories is provided. The main types identified are physical Machine (PM), Virtual Machine (VM), and application-based resource management. The performance issues involved in the implementation of the existing resource allocation mechanisms are identified, especially in heterogeneous architecture. Other aspects, such as power management strategies within data centers, is also provided to reinforce various data center, implementation model. Finally, the paper focuses on the virtualization model to be adopted in the data center design architecture.

Is your time best spent reading someone else’s essay? Get a 100% original essay FROM A CERTIFIED WRITER!

Save 25%

Introduction

There is a massive growth in Internet-Based commerce platforms with the boom in technology advancement. As a result, data centers are being adopted to provide the growing demand for automated, reliable solutions to the existing problem domain. Cost reduction aspects related to data centers are attracting most companies globally, as well as the management overhead. With the introduction of data centers in an organization, advantages such as shared elastic computational environment to diverse businesses and corporates are provided, which allows multiple application hosting.

Literature Review

Infrastructure as a service (IaaS) is the primary mode in which data centers offer services. Other methods, such as Software as a Service (SaaS) and Platform as a Service (PaaS) are among the additional services provided to companies. For companies with leases to data center include Service Legal Agreement (SLA) which require a legal agreement between the data centers and the tenants. Tenants make regular payments to the data centers for the services provided through different charging models. Issues arise when on-demand applications are to be produced, especially in meeting efficiency demands (Bevilacqua-Linn, Comcast Cable Communications LLC 2016).

The aspect of Physical Machine (PM) and server in most studies, as used in the paper, appear to be underutilized due to overprovision of resources in peak needs. Energy inefficient problems are the significant issues of overprovisioning within the data center, which potentially results in a budget increase. Therefore, in designing a data center focusing on cost-reduction methods are critical. Resource utilization within a data center is a preexisting problem. The primary mitigation strategy used to curb this issue as proposed by most designers is server virtualization through multiplexing method in Virtual machines (VMs). Despite the various redesigning ways, the problem persists. Memory, disk, and CPU utilization in several servers randomly selected across several data centers, according to the current report, show that, amounts to 77.93%, 75.28%, and 17.76%, respectively.

According to information collected on resource utilization across the virtual machines resource utilization, in a 29 cluster within google trace records, less than 60% CPU utilization with a memory usage of less than 50% (Rodriguez et al. 2018). The primary issue within data centers, as most scholars argue, is overprovisioning of services, especially the allocation of CPU resources as it's the significant power consumption aspect as compared to other resources. It is essential to adopt other mechanisms such as an agile resource allocation technique to handle the application resource dynamics. Sharp tools ensure meeting on of on-demand application requirements hence enhancing allocation of resources in data centers without violation of SLA agreements between the data center and the tenants.

The number of VMs needed to serve the resource required by an application such as the network input/output, memory, and CPU cycle is explained as the application resource demand requirement for each Virtual machine. Resources are managed to reduce energy consumption within data centers as opposed to resource utilization optimization. Management within data centers aids in operational cost reduction experienced in running the data center ensuring resource management. To achieve the main objectives within the data center, a resource management approach is assumed to overcome the existing challenges in data center design. As an aspect of resource management, resource utilization optimization is the main focus of the paper.

There are two primary components of resource management platform: Local Resource management (LRM) and Global Resource Management (GRM). The global view of the provisioning strategy is presented using the GRM resource management methodology. The GRM determines each VM location. The location of the hosted VM by the corresponding server ensures the physical server comprises of superfluous and sufficient resources needed to host the Virtual machines (Gabriel, Sha, and International Business Machines Corp 2017). To implement global resource management, Live VM migration is applied in data center design. The GRM provision course characterizes the aspect of the timeframe in the resource allocation mechanism.

VM based activities such as creating, migrating VM, starting, and termination operations are achieved by installing LRM in every VM machine. The GRM controls the various operation on a time-based scale. At the moment, server-based resources are assigned by LRM to the VMs hosting, depending on the information from the GRM (Rossi et al. 2017). Due to real-time resource demand, resources are allocated dynamically to each VM. The technique abstracts the LRM from getting server-related information such as resource utilization and application performance of all applications currently running on the server.

To determine the workload demand of the current object with a specific period under predication, workload demand predictor is used, which gives possible estimates depending on the data traces as recorded within the database. Application, the server, or the VM are the main objects used in the data center workload. Different resources are needed for different objects when determining the workload (Palmer, 2018). For specific purposes under selection, such as the virtual machine, the primary resources likely to be identified are the network I/O, memory, and CPU cycles resource utilization. The object selected in the earlier case (VM) would represent a dedicated server. When the object under selection becomes the server, the public resources being referred to will be the summation of all possible resources being utilized by the virtual machines supported.

The workload for the application in the case of object selection, the resources needed would be represented by the number of requests an application has to send over a specific time duration.

Since different objects have different characteristics reflecting varied workload demands, different prediction models can, therefore, be adopted in determines the future needs of an object. In such cases, GRA plays a more significant role in ensuring resources are assigned to each object depending on the workload estimated market and then mapping the resource estimate to several servers to ensure the SLAs are met as well as maximizing resource utilization within the data centers.

When GRA is integrated with the GRM, different workload estimation strategies will be adopted when choosing the available objects. The workload demand predictor is, however, not always implemented by the LRM on every individual server due to the resource-intensive nature. The main fear is that the algorithms are likely to deplete server resources. The object workload, instead, is determined by its performance hence allowing the LRM to trace its data. The information obtained from the data tracing enables the LRM to satisfy the resource provided by the local resource allocator (LRA).

There are two main reasons why workload demand prediction in resource management is essential. An object's workload keeps changing over time; the technique used to allocate resources at a specific time is also likely to change for the next resource allocation. During resource allocation, there is a high likelihood of incurring VM migration. Workload prediction is essential in managing the resource requirements of different objects. Managing resources using workload prediction helps in achieving better resource utilization as opposed to managing resources in workload prediction. The algorithm used in designing workload predictors should have accuracy since it determines the overall performance of the system.

Proposed Data Center Architecture

The proposed data center design will comprise of hardware components such as power distribution units and server racks. Other requirements will be network monitoring tools and server operating systems. The primary function of the data center to be developed is to host various components needed according to the company requirements addressed earlier. To assist in business running and maintenance, the data center will support multiple applications, cooling systems, storage devices, network computers, and the provision of computational power. The business operations, such as data processing requests, will be handled within the data center. The various data center requirements will include an electronic health record (HER) and customer relations management (CRM). To meet these requirements, there are several conditions. To address security requirements, the HER will be contained within the warehouse within the customer relations management. The proposed data center requires a high level of data security. Therefore, designing the data center will incorporate aspects such as physical security against external and internal threats. Other security measures, such as cybersecurity concerns, will also be considered vital. By having high-security levels, issues such as data integrity and trust issues will be maintained across the business operation. Other essential factors to be found in data center design include scalability aspects.

The case company is likely to expand shortly. Therefore, the design must meet collocation services while supporting the current operations. To handle the growing number of users accessing different services, scalability should be considered a vital aspect since growth within the enterprise is needed. The design incorporates both the long run aspects, such as creating room for more considerable data traffic processing. To address the changing technology needs, which is likely to call for other essential changes within the data center, the design assumes that the future changes will require hardware components and other critical components.

For manageability, it is recommended that the data centers must meet ease managing services. Where the data center is to be located will be determined by thoroughly evaluating security needs. Other essential issues addressed include data manageably. The data stored within the servers will allow easy evaluation by the company analysis. He cost problems to be treated the Operational Expenditure (OPEX) and the Total owners...