Tra!c-Aware Live Virtual Machine Migration: A Priority and SLA-Sensitive Approach

Abstract

Abstract Live migration of virtual machines (VMs) is critical in cloud environments for maintaining service continuity during load balancing, fault recovery, and infrastructure maintenance. However, traditional techniques such as pre-copy, postcopy, and hybrid migration often overlook the interplay between migration tra!c and application workloads, leading to increased downtime and degraded performance due to network contention. This work proposes a Tra!c-Aware Live VM Migration algorithm that dynamically selects the optimal migration strategy and performs adaptive bandwidth reservation based on real-time tra!c, migration urgency, and Service Level Agreement (SLA) constraints. The algorithm classifies network and workload tra!c patterns to choose between pre-copy and post-copy techniques and prioritizes migration tasks according to SLA sensitivity and urgency level, adapting between parallel and serial execution as needed. Empirical evaluation shows that high-priority migrations experience a 60–65% reduction in total migration time, medium-priority migrations improve by around 45%, and low-priority ones show minimal to no improvement, occasionally performing worse due to conservative resource use. In terms of downtime, low-priority migrations often perform best, medium-priority migrations o”er a balanced tradeo”, and high-priority migrations do not always outperform the best traditional methods. The approach also reduces performance degradation for co-located applications through adaptive tra!c shaping. Overall, the results highlight the benefits of integrating tra!c awareness, SLA compliance, and urgency-based prioritization to enable e!cient, non-disruptive VM migration in modern data centers.