Task Load and Work Property-Based Virtual Machine Dispatching Algorithm

Abstract

This research proposes a virtual machine dispatching algorithm that takes into account the task load and work property of a virtual machine. The algorithm utilizes an interruption monitoring module to assess the interruption frequency of the virtual machine and determine whether it belongs to a CPU-dense or I/O-dense type. Additionally, a CPU monitoring module tracks the current CPU utilization rate to monitor the task load. By considering the historical information within a specified time range, the algorithm calculates a dispatching time segment for the virtual machine and notifies the dispatcher to update the time segment. The dispatcher module adjusts the credit value dispatcher based on inputs from the monitoring and CPU modules. It applies different dispatching time segments to virtual machines with varying work properties, aiming to reduce I/O request delays, provide sufficient time for I/O request handling, and minimize additional overhead. Notably, the modifications occur within the virtual machine monitor, ensuring practicality and adaptability.