Article Preview
Top1. Introduction
Scheduling tasks over the cloud is a multi-domain problem, which includes pattern analysis, filtering, classification, clustering and prediction. Usually the following processes are followed in order to schedule cloud tasks (Asghari et al., 2020),
• Identification of undertaking boundaries from the task dataset
• Identification of machine parameters from the asset pool
• Strategizing rules and thresholds for machine and task scheduling
• Task execution on the given machine
• Evaluation of irregularities in execution, and changing methodology based to output parameters
• Post processing of tasks and machines if needed
In view of these steps, the researchers can see that at first the task parameters must be investigated. These parameters must incorporate essential assignment measurements like the undertaking execution delay, the task cutoff time, the task holding up time, while they can likewise incorporate optional parameters like undertaking mutual exclusiveness, shared reliance, and others (Nawrocki & Sniezynski, 2020). Typically, the all-out assignment execution prerequisites are administered by condition 1,
Where, is the total task execution requirement, is the total task execution delay, is the task deadline, is the task waiting time, while, are secondary application specific parameters needed to execute the task.
Once the task parameters are identified, then the resource parameters are observed, and evaluated. These parameters are again divided into primary and secondary parameters. Primary parameters include but are not limited to number of execution units available, capacity of each unit to execute the task, execution requirements for the resources, and others (Sui et al., 2019). A typical task scheduler can be observed from figure 1, wherein the tasks coming from users are given to the data center broker, the broker sends these tasks to the cloud controller for processing. The controller finally gives it to the host for further processing and scheduling on different machines.
The next section describes about such task scheduling systems in brief, and is followed by the proposed DAIRS-Q algorithms. This text further evaluates the said algorithm on different application specific datasets, and compares its efficiency with some state-of-the-art methods. Finally, the concludes with some interesting observations about the proposed protocol, and recommendations on how to further explore the field of work.