HLEM-VMP: An Effective Virtual Machine Placement Algorithm for Minimizing SLA Violations in Cloud Data Centers

Abstract

The objective of virtual machine placement in cloud data centers is to optimize resource allocation, which directly impacts the adherence to service level agreements (SLAs). Existing virtual machine placement strategies have achieved some success in reducing SLA violations through various approaches such as load balancing and minimizing host overloads. However, SLA compliance is influenced by multiple factors. Based on heuristic principles, this article fully considers various factors affecting SLA violations and proposes the HLEM-VMP virtual machine placement strategy. This strategy employs a greedy method to determine the placement of virtual machines by filtering and scoring physical hosts. We conducted comprehensive comparative experiments using the Cloudsim simulator with Planetlab and random workloads against classical algorithms such as EPSO, AFED-EF, and HS3MC-3MIN-1.5. The results show that HLEM-VMP not only significantly reduces SLA violations but also performs excellently on the combined index ESV. Building on this, we further propose the General-HLEM (G-HLEM) model, which allows for a tradeoff between energy consumption and SLA violations, thereby enhancing the model's generalizability. Experiments in three typical application scenarios demonstrate that G-HLEM can meet the needs of different scenarios and outperforms methods like EPSO, AFED-EF, and HS3MC-3MIN-1.5 on the ESV index, providing valuable insights for addressing SLA violation issues through virtual machine placement strategies.