Medium Access Control in Wireless Network-on-Chip: A Context Analysis

Sergi Abadal, Albert Mestres, Josep Torrellas, Eduard Alarcón, Albert Cabellos-Aparicio

Abstract

Wireless on-chip communication is a promising candidate to address the performance and efficiency issues that arise when scaling current Network-on-Chip (NoC) techniques to manycore processors. A Wireless Network-on-Chip (WNoC) can serve global and broadcast traffic with ultra-low latency even in thousand-core chips, thus acting as a natural complement of conventional and throughput-oriented wireline NoCs. However, the development of Medium Access Control (MAC) strategies needed to efficiently share the wireless medium among the increasing number of cores remains as a considerable challenge given the singularities of the environment and the novelty of the research area. In this position paper, we present a context analysis describing the physical constraints, performance objectives, and traffic characteristics of the on-chip communication paradigm. We summarize the main differences with respect to traditional wireless scenarios, to then discuss their implications on the design of MAC protocols for manycore WNoCs, with the ultimate goal of kickstarting this arguably unexplored research area.

Introduction: The Manycore Bottleneck

As processor design enters the "manycore era," with chips housing hundreds or even thousands of cores, traditional wired interconnects are reaching their limits. Conventional Network-on-Chip (NoC) architectures, relying on copper wires and multi-hop routers, struggle with high latency and power consumption, particularly for global broadcast messages.

Wireless Network-on-Chip (WNoC) technology, utilizing millimeter-wave frequencies, offers a high-speed alternative. However, it introduces a critical challenge: Medium Access Control (MAC). Standard wireless protocols like Wi-Fi are unsuitable for the on-chip environment, where cores are stationary, power is constrained by thermal limits, and latency must be measured in nanoseconds rather than milliseconds.

The Hybrid WNoC Architecture

A viable WNoC does not replace wires but complements them. The proposed architecture utilizes a hybrid Network Interface (NIF) that manages both wired and wireless communication paths. This allows the system to route local, high-throughput traffic through traditional wires while offloading global, latency-sensitive, or broadcast traffic to the wireless channel.

Hybrid Network Interface (NIF) architecture integrating wired and wireless communication paths.

Context Analysis for MAC Design

Designing an effective MAC protocol for WNoC requires a deep understanding of the unique on-chip environment. The authors perform a systematic context analysis across three primary dimensions: physical constraints, workload characteristics, and architectural requirements.

The three-pronged context analysis for designing MAC protocols in WNoC environments.

1. Physical Constraints (PHY)

Unlike traditional wireless networks, a chip is a "static and controlled landscape." The locations of all antennas are fixed and known at design time. This stability allows for highly optimized channel models and the potential for collision detection mechanisms—similar to early Ethernet—which are typically impossible in long-range wireless scenarios.

2. Workload Characteristics (NET)

On-chip traffic is highly heterogeneous and bursty. By analyzing benchmarks like the SPLASH2 suite and applications like fluidanimate, researchers have identified "phase behavior," where programs alternate between intense communication and computation. Traffic also exhibits "hotspots," where specific cores dominate communication. These patterns can be quantified using parameters like the Hurst exponent ($H$) for burstiness and Gaussian standard deviation ($\sigma$) for spatial concentration.

3. Architectural Requirements (ARCH)

The performance targets for WNoC are extreme. A successful MAC protocol must achieve:

  • Latency: 1–100 nanoseconds.
  • Throughput: 10–100 Gb/s.
  • Reliability: A Bit Error Rate (BER) of $10^{-15}$, matching physical wires.
  • Energy Efficiency: 1–10 pJ/bit to stay within thermal envelopes.

Strategic Insights: The Monolithic Advantage

The paper argues that the "monolithic" nature of the chip provides a unique advantage: predictability. Because the system is closed, the compiler can be used to predict communication bursts. This enables "proactive" MAC protocols that allocate channel access before a core even requests it, rather than reactive protocols that wait for collisions to occur.

Furthermore, the existing wired network can serve as a "control plane." By handling synchronization and scheduling over wires, the wireless channel remains dedicated to high-speed data transfer, maximizing efficiency.

Conclusion and Future Challenges

While WNoC presents a transformative path for manycore scaling, several hurdles remain. Current hardware complexity makes frequency multiplexing (hundreds of separate channels) difficult to implement. Additionally, the "ACK implosion" problem—where a single broadcast triggers a flood of acknowledgments—must be addressed. Future research will likely focus on cross-layer designs and machine learning-based predictors that can reconfigure MAC protocols in real-time based on the running application's needs.

Summary updated

Abstract

Wireless on-chip communication is a promising candidate to address the performance and efficiency issues that arise when scaling current Network-on-Chip (NoC) techniques to manycore processors. A Wireless Network-on-Chip (WNoC) can serve global and broadcast traffic with ultra-low latency even in thousand-core chips, thus acting as a natural complement of conventional and throughput-oriented wireline NoCs. However, the development of Medium Access Control (MAC) strategies needed to efficiently share the wireless medium among the increasing number of cores remains as a considerable challenge given the singularities of the environment and the novelty of the research area. In this position paper, we present a context analysis describing the physical constraints, performance objectives, and traffic characteristics of the on-chip communication paradigm. We summarize the main differences with respect to traditional wireless scenarios, to then discuss their implications on the design of MAC protocols for manycore WNoCs, with the ultimate goal of kickstarting this arguably unexplored research area.
BibTeX
@article{abadal2018medium,
  author = {Sergi Abadal and Albert Mestres and Josep Torrellas and Eduard Alarcón and Albert Cabellos-Aparicio},
  title = {{Medium Access Control in Wireless Network-on-Chip: A Context Analysis}},
  journal = {IEEE Communications Magazine},
  year = {2018},
  url = {https://openreview.net/forum?id=Upt8i17A34},
}
Figure