Masters Thesis

Joint virtual machine placement and migration in dynamic policy-driven data centers

Policy-Driven Data Centers (PDDCs) are data centers in which all Virtual Machine (VM) traffic must traverse a sequence of Middleboxes (MBs). These MB policies help guarantee a given level of security and performance for all applications hosted in the PDDC at the cost of increased network traffic and energy consumption. This thesis proposes a new VM optimization framework called VM2P (Virtual-Machine Migration in PDDCs), which, when given an existing PDDC with existing VM placements and policies, will migrate VMs within a PDDC in order to minimize energy consumption while still maintaining all policy constraints. This thesis will also show how the VM placement problem VMP2 (Virtual-Machine Placement in PDDCs), the problem of distributing VMs into an empty PDDC such that energy consumption is minimized, is actually a special case of VM2P. The performance of VM2P and VMP2 are evaluated via simulations wherein the algorithms will compete with state-of-the-art policy-agnostic algorithms.