Article Preview
TopIntroduction
For more than 10 years since the concept was proposed, cloud computing has taken a huge leap forward and had drastic changes. Cloud computing is regarded as a revolution in the field of computer networks. Because of its emergence, social working methods and business models are also undergoing tremendous changes. Cloud computing platform has powerful computing and storage capacities, and can provide high-quality customized services. However, cloud computing is not a brand-new network technology, but a brand-new network application concept. In fact, cloud computing is the result of the mixed evolution of computer technologies such as distributed computing, utility computing, load balancing, parallel computing, network storage, and virtualization. The core concept of cloud computing is to provide fast and secure computing service and data storage on the network so that everyone can use the huge computing resources and data centers through Internet. Therefore, cloud computing is essentially a network that provides resources, and can be regarded as unlimited expansion. Users can obtain resources from the “cloud” at any time, use them as needed, as long as they pay for it according to usage. CSPs build cloud computing resource pool in the form of data center and provide it to users in the form of virtual machine (VM). After a user rents VMs in the data center, she/he can submit her/his tasks to the VMs for processing via network. After the tasks are completed, the VMs return the results to the user via network too. Cloud users only need to install a simple client in local. During the service, the CSP and the user who want to use cloud computing resources sign a service level agreement (SLA) to agree on the quality of service (QoS). Under SLA condition, the CSP tries her/his best to provide better QoS and resources management, such as faster task response, less error probability, and lower data center energy consumption, etc., to attract more users and get better economic benefits. Task response time and other indicators that affect QoS are closely related to the cloud platform’s TS strategy. So, the quality of TS strategy determines a cloud platform’s service quality and business profit (Madni, Latiff, Coulibaly & Abdulhamid, 2017). However, the problem of cloud TS optimization has not yet been well solved, and it is currently one of urgent problems of cloud computing to be solved.
Due to the advancement and convenience of the cloud computing service model, more and more users abandon the software service model in which they manage physical servers themselves, begin to use the cloud computing service model and migrate their businesses to virtual servers in remote data center. The tasks generated by different businesses usually have different characteristics. For example, some are real-time, some are with a large amount of data, some can be processed offline, and some need to be processed online. Because of the diversity of user tasks and the huge number of users, cloud computing system has to deal with a large number of tasks and data, especially when batch tasks are submitted. Due to the large number, the tasks submitted by user must first enter task queues to wait for scheduling. In this case, if all tasks are submitted into a single queue, the tasks are prone to long queuing delay when the business bursts. Furthermore, to deploy tasks on a single VM is prone to overload the VM server, leading to slow task response. The service mode of single queue and single server will severely reduce the QoS of the cloud platform. Therefore, CSPs and users tend to use a service mode with MQMC, that is, different types of tasks are loaded into different queues for submission, and multiple VMs are clustered to execute the tasks. However, The TS in this MQMC mode will be more complicated. How to perform efficient TS and reasonable resource allocation under this mode has not been studied much.