EASM: Efficiency-aware switch migration for balancing controller loads in software-defined networking

Tao Hu, Julong Lan, Jianhui Zhang, Wei Zhao

TL;DR

EASM (Efficiency-Aware Switch Migration) is a novel framework designed to balance loads across distributed SDN controllers without breaking the bank on migration overhead. By optimizing for both load distribution and migration costs simultaneously, EASM reduces controller response times by 21.9% and boosts throughput by over 30% compared to traditional migration strategies.

What Problem Does This Paper Address?

In Software-Defined Networking (SDN), a centralized controller manages the entire network. However, as networks scale, a single controller becomes a bottleneck. The solution is a distributed control plane—multiple controllers managing different "domains."

The challenge is that network traffic is dynamic. One controller might be overwhelmed (a "hot" controller) while another sits idle (a "cold" controller). To fix this, we use switch migration: moving a switch from the control of an overloaded controller to an underloaded one.

The authors identify a major gap in existing research: most schemes focus purely on achieving a perfect load balance but ignore the migration efficiency. Moving a switch consumes network resources, creates control overhead, and can temporarily increase latency. If the "cost" of moving a switch is higher than the benefit of the new balance, the migration is counterproductive.

Methods

The authors propose the EASM strategy, which consists of three distinct algorithmic modules:

  1. EASM-1 (Load Imbalance Detection): Instead of just looking at raw CPU or memory, this module builds a load difference matrix. It calculates a "trigger factor" based on the disparity between controllers. If this factor exceeds a dynamically calculated threshold, a migration is triggered.
  2. EASM-2 (Optimal Object Selection): This is the "brain" of the system. It selects which switch to move and which target controller should receive it. It introduces a Migration Efficiency metric that weighs the improvement in the load balancing rate against the migration cost (calculated as the product of flow rate and network hops). To find the best target controller efficiently, it employs a Simulated Annealing (SA) metaheuristic.
  3. EASM-3 (Dynamic Migration Decision): This module handles the execution. It uses Migration Triplets $[c_m, s_i, c_n]$ (source controller, migrating switch, target controller) to ensure migrations are coordinated and don't conflict with one another, maintaining a consistent network view throughout the process.

Datasets

The researchers evaluated EASM using a mix of real-world topologies and emulated traffic:

  • Topologies: The primary testing was done on the Internet2 OS3E topology (34 nodes, 42 links). To test scalability and adaptability, they also used various topologies from the Topology Zoo.
  • Traffic Generation: They used Iperf to generate TCP flows, simulating a distribution of network traffic with an average flow request rate of 200KB/s.
  • Environment: The system was implemented using the OpenDaylight controller and the Mininet network emulator.

Key Insights and Ideas

  • Migration isn't free: The core "aha" moment is treating migration as an optimization problem where "cost" is a first-class citizen. A switch with high traffic that is many hops away from a target controller is a poor candidate for migration, even if that target is idle.
  • Multi-faceted Load Calculation: The paper doesn't just look at packet counts. It formulates controller load by aggregating data interaction overhead, routing formulation overhead (Packet-in processing), and state synchronization overhead between controllers.
  • Avoiding the "Ping-Pong" Effect: By using a load difference matrix and specific trigger thresholds, EASM avoids unnecessary migrations that occur when controllers have only marginal differences in load.

Key Results

EASM was compared against three baselines: No Switch Migration (NSM), Closest Switch Migration (CSM), and Maximum Utilization Switch Migration (MUSM).

  • Response Time: EASM reduced average controller response time by at least 21.9% compared to other migration strategies.
  • Throughput: Controller throughput improved by 30.4% on average, reaching approximately 3660 packets/s.
  • Efficiency: EASM achieved significantly lower migration costs (measured in KB/s * hops) and shorter migration times than MUSM and CSM.
  • Scalability: As the network scale increased (tested via Topology Zoo), EASM maintained a higher and more stable Load Balancing Rate (LBR) than its competitors.

Limitations and Future Work

While EASM shows strong results in emulation, the authors note several areas for further exploration:

  • Real-world Deployment: The current results are based on Mininet simulations; testing in a large-scale physical testbed is a necessary next step.
  • Reliability: The paper focuses on efficiency and load, but the impact of migration on controller reliability (e.g., what happens if a controller fails mid-migration) needs deeper study.
  • Security: The process of changing switch-controller mappings introduces potential attack vectors that the current EASM framework does not explicitly address.

Summary updated

Abstract

Distributed multi-controller deployment is a promising method to achieve a scalable and reliable control plane of Software-Defined Networking (SDN). However, it brings a new challenge for balancing loads on the distributed controllers as the network traffic dynamically changes. The unbalanced load distribution on the controllers will increase response delay for processing flows and reduce the controllers'throughput. Switch migration is an effective approach to solve the problem. However, existing schemes focus only on the load balancing performance but ignore migration efficiency, which may result in high migration costs and unnecessary control overheads. This paper proposes Efficiency-Aware Switch Migration (EASM) to balance the controllers'loads and improve migration efficiency. We introduce load difference matrix and trigger factor to measure load balancing on controllers. We also introduce the migration efficiency problem, which considers load balancing rate and migration cost simultaneously to optimally migrate switches. We propose EASM to efficiently solve to the problem. The simulation results show that EASM outperforms baseline schemes by reducing the controller response time by about 21.9%, improving the controller throughput by 30.4% on average, maintaining good load balancing rate, low migration costs and migration time, when the network scale changes.
BibTeX
@inproceedings{hu2018easm,
  author = {Tao Hu and Julong Lan and Jianhui Zhang and Wei Zhao},
  title = {{EASM: Efficiency-aware switch migration for balancing controller loads in software-defined networking}},
  booktitle = {Peer-to-Peer},
  year = {2018},
  doi = {10.1007/s12083-018-0632-6},
}

Loading related content...

Figure