NFVactor: A Resilient NFV System using the Distributed Actor Model

Jingpu Duan, Xiaodong Yi, Shixiong Zhao, Chuan Wu, Heming Cui, Franck Le

Abstract

Resilience functionality, including failure resilience and flow migration, is of pivotal importance in practical network function virtualization (NFV) systems. However, existing failure recovery procedures incur high packet processing delay due to heavyweight process-checkpointing, while flow migration has poor performance due to centralized control. This paper proposes NFVactor, a novel NFV system that aims to provide lightweight failure resilience and high-performance flow migration. NFVactor enables these by using the actor model to provide a per-flow execution environment, so that each flow can replicate and migrate itself with improved parallelism, while the efficiency of the actor model is guaranteed by a carefully designed runtime system. Moreover, NFVactor achieves transparent resilience: once a new NF is implemented for NFVactor, the NF automatically acquires resilience support. Our evaluation result shows that NFVactor achieves 10Gbps packet processing, flow migration completion time that is 197 times faster than existing systems, and packet processing delay stabilized at around 20 microseconds during replication.

Introduction

In modern cloud networking, traditional hardware middleboxes like firewalls and load balancers are increasingly replaced by software-based Network Functions (NFs). A critical requirement for these systems is resilience: the ability to survive hardware failures and migrate traffic flows seamlessly during server overloads. Current NFV solutions typically rely on heavyweight process-level checkpointing, which causes significant latency spikes (sometimes thousands of microseconds), or centralized controllers that become bottlenecks during large-scale flow migrations involving millions of concurrent flows.

System Architecture

NFVactor addresses these challenges by reimagining the NFV framework through the lens of the Distributed Actor Model. Instead of a monolithic process, NFVactor treats every individual network flow as an independent "actor"—a lightweight, isolated execution unit.

NFVactor System Architecture Overview

Key Innovations

  • One-Actor-One-Flow: By assigning each flow its own actor, NFVactor enables flows to replicate or migrate themselves in parallel. This decentralization removes the global controller from the critical path of state management, allowing for massive scalability.
  • Transparent Resilience: The framework provides high-level APIs (such as process_pkt and allocate_new_fs) that decouple NF logic from state management. NFs built on these APIs automatically inherit resilience features without additional developer effort.
  • High-Performance Runtime: To meet 10Gbps line-rate requirements, NFVactor utilizes a custom runtime built on DPDK (Data Plane Development Kit). It employs a "run-to-completion" model and a Module Graph Scheduler to minimize context switching and kernel overhead.
Internal Runtime Design and Actor Communication
  • Liaison Actor: To bridge external control signals with millions of flow actors, a specialized "Liaison Actor" coordinates management tasks on each runtime. This ensures that the system can handle control messages without becoming a performance bottleneck.

Evaluation and Results

The system was evaluated using synthetic traffic (64-byte TCP packets) generated by a customized BESS switch, simulating worst-case scenarios for packet processing at rates up to 10Gbps.

Performance Metrics

  • Throughput: NFVactor maintains 10Gbps line-rate processing for common NFs, including firewalls, NATs, and Load Balancers.
  • Migration Efficiency: The decentralized protocol allows NFVactor to migrate flows up to 197 times faster than traditional centralized systems. In tests, it successfully migrated 600,000 concurrent flows in approximately 700 milliseconds.
  • Latency Stability: While checkpoint-based systems often see latency spikes exceeding 3,000µs, NFVactor stabilizes packet processing delay at approximately 20 microseconds even during active state replication.
  • Recovery: Full runtime recovery after a crash is achieved in under 70 milliseconds, ensuring high availability for critical network services.

Conclusion

NFVactor demonstrates that the distributed actor model is a powerful paradigm for building resilient, high-performance NFV systems. By shifting from centralized management to per-flow autonomy, it achieves unprecedented migration speeds and latency stability. While it requires NFs to be ported to its specific API, the performance gains in resilience and scalability offer a compelling path forward for cloud-native networking.

Summary updated

Abstract

Resilience functionality, including failure resilience and flow migration, is of pivotal importance in practical network function virtualization (NFV) systems. However, existing failure recovery procedures incur high packet processing delay due to heavyweight process-checkpointing, while flow migration has poor performance due to centralized control. This paper proposes NFVactor, a novel NFV system that aims to provide lightweight failure resilience and high-performance flow migration. NFVactor enables these by using actor model to provide a per-flow execution environment, so that each flow can replicate and migrate itself with improved parallelism, while the efficiency of the actor model is guaranteed by a carefully designed runtime system. Moreover, NFVactor achieves transparent resilience: once a new NF is implemented for NFVactor, the NF automatically acquires resilience support. Our evaluation result shows that NFVactor achieves 10Gbps packet processing, flow migration completion time that is 197 times faster than existing system, and packet processing delay stablized at around 20 microseconds during replication.
BibTeX
@article{duan2019nfvactor,
  author = {Jingpu Duan and Xiaodong Yi and Shixiong Zhao and Chuan Wu and Heming Cui and Franck Le},
  title = {{NFVactor: A Resilient NFV System using the Distributed Actor Model}},
  journal = {IEEE J. Sel. Areas Commun.},
  year = {2019},
  doi = {10.1109/jsac.2019.2894287},
}
Figure