A machine learning-based optimization approach for pre-copy live virtual machine migration

Abstract Organizations widely use cloud computing to outsource their needs. One crucial issue of is that services must be available clients at all times. However, the may temporarily unavailable due maintenance infrastructure, load balancing services, defense against cyber attacks, power management, proactive fault tolerance, or resource usage. The unavailability impacts negatively on business model providers. solution tackle service Live Virtual Machine Migration (LVM), is, moving virtual machines (VMs) from source host machine destination without disrupting running application. Pre-copy memory migration a common LVM approach used in most networked systems such as cloud. main difficulty with this high rate frequently updating pages, referred "dirty pages. Transferring these updated dirty pages during pre-copy prolongs total time. After predefined iteration, enters stop-and-copy phase and transfers remaining If are huge, downtime will very -resulting negative impact availability services. To minimize downtime, it critical find an optimal time migrate approach. address issue, paper proposes learning-based method optimize migration. It has mainly three stages (i) Feature selection (ii) Model generation (iii) Application proposed experiment results show our outperforms other learning models terms prediction accuracy significantly reduces process.