PLAN: A Policy-Aware VM Management Scheme for Cloud Data Centres

Lin Cui, Fung Po Tso, Dimitrios P. Pezaros, Weijia Jia

Abstract

Policies play an important role in network configuration and therefore in offering secure and high performance services especially over multi-tenant Cloud Data Center (DC) environments. At the same time, elastic resource provisioning through virtualization often disregards policy requirements, assuming that the policy implementation is handled by the underlying network infrastructure. This can result in policy violations, performance degradation and security vulnerabilities. In this paper, we define PLAN, a PoLicy-Aware and Network-aware VM management scheme to jointly consider DC communication cost reduction through Virtual Machine (VM) migration while meeting network policy requirements. We show that the problem is NP-hard and derive an efficient approximate algorithm to reduce communication cost while adhering to policy constraints. Through extensive evaluation, we show that PLAN can reduce topology-wide communication cost by 38% over diverse aggregate traffic and configuration policies.

Introduction

In modern multi-tenant cloud data centers, network traffic is rarely a simple point-to-point transmission. Instead, it must often traverse a "service chain" of middleboxes—such as firewalls, Intrusion Prevention Systems (IPS), and load balancers—to satisfy security and performance policies. While virtualization allows for elastic resource provisioning, standard VM management tools often ignore these network policies during migration.

Network architecture showing middlebox service chains and VM migration paths.

When a VM is moved to consolidate server usage or save power without considering its required middleboxes, the traffic path can become long and inefficient—a phenomenon known as "tromboning." This leads to wasted bandwidth and increased latency. In the worst-case scenario, traffic might bypass a security middlebox entirely, resulting in policy violations. PLAN (PoLicy-Aware and Network-aware VM management) is designed to address this by jointly optimizing VM placement while respecting strict network security policies.

Methodology

The authors formulate the PLAN problem as an optimization task aimed at maximizing "utility" (defined as the reduction in communication cost minus the overhead of the migration itself) while respecting server capacity and policy constraints.

Key components of the approach include:

  • Mathematical Modeling: Communication cost is defined based on the entire path, including mandatory middlebox hops, weighted by the relative "expense" of different network layers (e.g., core links vs. edge links).
  • Complexity Analysis: The problem is proven to be NP-Hard via reduction to the Multiple Knapsack Problem.
  • Decentralized Heuristic Algorithm: To ensure scalability in massive data centers, a decentralized greedy algorithm is proposed. PLAN-VM allows hypervisors to periodically evaluate if moving a local VM would increase total utility, while PLAN-Server manages destination server resources.
  • Hybrid Implementation: The system utilizes a partly decentralized architecture where computationally expensive cost calculations occur at the hypervisor level, while a centralized SDN controller provides global network topology and flow statistics.

Experimental Evaluation

The performance of PLAN was evaluated using the ns-3 network simulator on a large-scale Fat-tree topology (250 servers, 2,320 VMs) and a physical testbed using Xen hypervisors and Open vSwitch. The simulation environment used a traffic profile where each VM averaged 10 random outgoing connections, with 90% of flows constrained by policies requiring traversal of 1–3 middleboxes.

Performance evaluation of PLAN across total cost, route length, and link utilization over time.

Key Results

  • Cost and Path Efficiency: PLAN reduced topology-wide communication costs by 38% compared to non-policy-aware schemes like S-CORE. The average end-to-end route length was reduced by approximately 20%.
  • Congestion Mitigation: Link utilization at the core and aggregation layers was reduced by 30.5% and 7%, respectively, creating significant headroom in the network.
  • Convergence and Stability: The decentralized algorithm demonstrated stability; as shown in the time-series data, most VMs required only a single migration to reach an optimized state.
  • Low Overhead: Testbed results confirmed that the flow-tracking module consumed less than 5% CPU utilization, even when managing 10,000 concurrent flows.

Conclusion and Future Work

PLAN demonstrates that policy-awareness is critical for efficient VM management in modern data centers. By accounting for middleboxes, PLAN significantly reduces network overhead without compromising security. Future work includes extending the model to environments where middleboxes themselves are virtualized (NFV) and can be migrated, as well as integrating traffic prediction models to enable proactive migration decisions.

Summary updated

Abstract

Policies play an important role in network configuration and therefore in offering secure and high performance services especially over multi-tenant Cloud Data Center (DC) environments. At the same time, elastic resource provisioning through virtualization often disregards policy requirements, assuming that the policy implementation is handled by the underlying network infrastructure. This can result in policy violations, performance degradation and security vulnerabilities. In this paper, we define PLAN, a PoLicy-Aware and Network-aware VM management scheme to jointly consider DC communication cost reduction through Virtual Machine (VM) migration while meeting network policy requirements. We show that the problem is NP-hard and derive an efficient approximate algorithm to reduce communication cost while adhering to policy constraints. Through extensive evaluation, we show that PLAN can reduce topology-wide communication cost by 38% over diverse aggregate traffic and configuration policies.
BibTeX
@inproceedings{cui2015plan,
  author = {Lin Cui and Fung Po Tso and Dimitrios P. Pezaros and Weijia Jia},
  title = {{PLAN: A Policy-Aware VM Management Scheme for Cloud Data Centres}},
  booktitle = {Ucc},
  year = {2015},
}
Figure