A Cache-Aware Virtual Machine Placement With Network Constraints in Large-Scale Network Emulation

In large-scale network emulation, the effectiveness of network construction is vital to the availability of the platform. Meanwhile, the physical network bandwidth resource is valuable for emulating virtual network links. However, existing methods suffer from long network construction time and low utilization of network bandwidth. In this paper, we present a novel solution to solve the Virtual Machine (VM) placement problem in large-scale network emulation. First, we employ VM templates cache in physical machines (PMs) to reduce template transmission. Then, we design a Two-Level Hierarchical Decomposition Algorithm to decompose the complex problem of large-scale network construction into multiple independent sub-problems. Finally, we propose TCANC, which is a VM placement algorithm aiming at maximizing utilization of VM template cache and mitigating network bandwidth consumption. We conduct a series of experiments to compare the effectiveness of our solution with the commercial optimizer Gurobi and the aggressive heuristic algorithm. The results prove that TCANC ensures a high hit rate of VM template cache, better than Gurobi, thus significantly reducing time consumption caused by template file transmission during network construction. Moreover, the large-scale experiment with 1,000,000 VM nodes shows that our approach could be efficiently applied to large-scale virtual network environments with high resource utilization.