Related works
According to our knowledge, we have not noticed any comprehensive taxonomy journal article on IaaS cloud resource allocation approaches. However, a number of related surveys and review book chapters that referred to IaaS cloud resource allocation have been published. In this section, we describe only those surveys and reviews. The detail description of referred algorithms and their original reference are in section IV.
In [10], the authors studied open source cloud platforms. This work compared solutions and their business model (hardware, middleware and user level) according to configuration flexibility. It also compared the service, infrastructure and users of those systems. The important of cloud resource allocation was stated, but none of detail issues were discussed.
The work in [1] extended a taxonomy and survey of cloud computing system to both open source and commercial cloud platforms. The cloud systems are mainly characterized with architecture, virtualization management, service, fault tolerance and security. Related to resource allocation, the authors referred only the load balance feature. In which, most studied systems use simple algorithms such as Round Robin, Greedy or server load equalization at IaaS level.
In [11], the author presented the taxonomy and survey of energy-efficient data centres and cloud computing systems. This work discussed many energy-saving techniques ranging from hardware level, OS level, Virtualization level to data centre level. At data centre level, the authors described several research works about saving energy techniques. Those techniques mostly are based on DVFS (Dynamic Voltage and Frequency Scaling), VM consolidation and power switching.
In [12], the authors discussed various scheduling techniques for traditional distributed systems, grid computing systems and cloud computing systems. However, the author only touched the surface of real works for cloud computing. The IaaS cloud employs the VM concept. Each VM could have multiple CPUs and must be allocated completely within a physical machine. From the point of resource allocation, this is the distinguished character of Cloud IaaS compared with traditional distributed system and Grid computing. In other types of distributed system, one job including many processes can be spread out to multiple physical machines.
The work in [63] took some scheduling algorithms for cloud computing and performed experiment to do comparative analysis. The main comparative criteria include execution time, resource use rate and cost of algorithm. Also following this way, the work in [64] focused on scheduling schemes for on-demand IaaS requests. However, the authors of [64] used the analytical model and studied the ability of reducing energy consumption.
Resource allocation algorithms taxonomy
The resource allocation algorithm responds for finding the resource allocation solution that satisfies a specific goal of the cloud provider. This goal could be optimizing power consumption, optimizing cost, ensuring SLA, etc. The typical resource allocation architecture for IaaS cloud is presented in Figure 4.
In general, the input for resource allocation includes resource information and workload information. Based on this input information, the resource allocation algorithm finds out resource allocation solution.
Figure 4. Typical resource allocation architecture
Chia sẻ với bạn bè của bạn: |