TL;DR
This paper provides a comprehensive survey of Network Function Virtualization (NFV) platforms, categorizing them by design purpose and exploring their diverse architectural choices. It aims to guide network operators and service providers in selecting or implementing the most suitable NFV solutions by detailing the design space and implementation trade-offs.
What Problem Does This Paper Address?
Traditional networks rely on expensive, proprietary hardware appliances (middleboxes) for network services. This approach is inflexible, costly, and slow to adapt to evolving demands. Network Function Virtualization (NFV) offers a solution by replacing these hardware appliances with software-based Virtual Network Functions (VNFs) running on commodity hardware. However, the NFV ecosystem has exploded with numerous platforms, each with unique design choices driven by different motivations. This makes it incredibly difficult for network operators to navigate the landscape, understand the trade-offs, and select the right platform for their needs. This paper addresses this challenge by providing a structured overview and taxonomy of existing NFV platforms.
Methods
The authors conducted a comprehensive survey of existing NFV platform implementations. Their methodology involved:
- Architectural Overview: Presenting a generic NFV platform architecture, including the MANO (Management and Orchestration) plane, service plane, and NFV Infrastructure (NFVI).
- Taxonomy: Classifying NFV platforms based on their primary design purpose into three categories: Integrated NFV Platforms, MANO Systems, and NFVI Optimization.
- Design Space Exploration: Detailing the various design choices and implementation strategies adopted by different platforms across key areas such as high-level APIs, placement, state redistribution, execution models, TCP functionality, VNF I/O, and hardware acceleration.
- Literature Review: Analyzing a broad range of academic and industry publications to gather information on existing NFV platforms.
Datasets
Not discussed in the paper.
Key Insights and Ideas
- NFV's Promise and Peril: NFV offers significant advantages in agility and cost-effectiveness over traditional networks, but the sheer diversity of platform designs creates a complex decision-making challenge for operators.
- Design Space is Vast and Interconnected: There's no single "best" NFV platform. The optimal choice depends heavily on specific use cases, application contexts, and operational requirements. Key design decisions (e.g., execution model, I/O framework, acceleration techniques) are interconnected and influence overall performance and flexibility.
- Evolution Towards Cloud-Native: The trend is shifting from VM-centric NFV to cloud-native approaches leveraging containers and Kubernetes, emphasizing microservices, declarative management, and automated lifecycle operations.
- Hardware Acceleration is Crucial: As traffic volumes grow, NFVI optimization and hardware offloading (using SmartNICs, FPGAs, GPUs) are becoming critical for achieving high performance and meeting stringent SLAs.
- State Management is a Key Challenge: Handling stateful VNFs, especially during scaling and failover events, requires sophisticated strategies like state migration or migration avoidance to maintain consistency and minimize service disruption.
- Observability and Diagnosability are Paramount: The distributed nature of NFV makes monitoring and troubleshooting complex. Platforms need robust mechanisms for performance characterization, bottleneck localization, and evidence-based auditing.
Key Results
The paper's primary contribution is a structured survey and taxonomy of NFV platforms. It doesn't present new experimental results but rather synthesizes existing knowledge to provide:
- A classification of NFV platforms based on their design purpose (Integrated NFV, MANO System, NFVI Optimization).
- A detailed exploration of the NFV design space, covering critical aspects like high-level APIs, placement strategies, state redistribution, execution models, TCP functionality, VNF I/O, and hardware offloading.
- A comprehensive table (Table 1) summarizing design choices across numerous existing NFV platforms, offering a comparative overview.
- Insights into emerging trends like cloud-native CNFs, hybrid execution models, and the increasing importance of hardware acceleration.
Limitations and Future Work
The authors acknowledge that the NFV landscape is rapidly evolving. While this survey provides a snapshot of existing platforms, it will require periodic updates to include new developments. Future work could involve:
- Deeper Dive into Specific Design Choices: While the paper explores the design space, more in-depth analysis of the performance implications and trade-offs of specific choices could be beneficial.
- Emerging Trends: The paper touches upon cloud-native CNFs and AI in NFV. Further research could focus on how these trends are shaping platform design and what new challenges they introduce.
- Benchmarking and Standardization: The authors implicitly highlight the need for better benchmarking and standardization to allow for more objective comparisons between platforms.
- Operational Aspects: While the survey covers design choices, future work could delve deeper into the operational aspects of managing and maintaining these complex NFV platforms in production environments.
Summary updated