NOVIA: A Framework for Discovering Non-Conventional Inline Accelerators

Trilla, David; Wellman, John-David; Buyuktosunoglu, Alper; Bose, Pradip

doi:10.1145/3466752.3480094

Cited by 4 publications

(1 citation statement)

References 51 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…As seen in the figure, the main processing device is still the CPU, however, it is now supported by specific HW accelerators which can be related to networking or other functionalities. There are several studies on the benefits of such an architecture and how to identify good candidates for HW acceleration, such as the Accelerometer study from Facebook [99], a framework for discovering Non-Conventional Inline Accelerators from IBM (NOVIA) [100] or EVITA project for embedded security functions with the HSM [101]. Networking in particular, is an area with great potential for HW acceleration because of its intrinsic characteristics.…”

Section: B Network Processors With Hw-based Network Offloading Capabi...mentioning

confidence: 99%

The Future Roadmap of In-Vehicle Network Processing: A HW-Centric (R-)evolution

2022

View full text Add to dashboard Cite

The automotive industry is undergoing a deep revolution. With the race towards autonomous driving, the amount of technologies, sensors and actuators that need to be integrated in the vehicle increases exponentially. This imposes new great challenges in the vehicle electric/electronic (E/E) architecture and, especially, in the In-Vehicle Network (IVN). In this work, we analyze the evolution of IVNs, and focus on the main network processing platform integrated in them: the Gateway (GW). We derive the requirements of Network Processing Platforms that need to be fulfilled by future GW controllers focusing on two perspectives: functional requirements and structural requirements. Functional requirements refer to the functionalities that need to be delivered by these network processing platforms. Structural requirements refer to design aspects which ensure the feasibility, usability and future evolution of the design. By focusing on the Network Processing architecture, we review the available options in the state of the art, both in industry and academia. We evaluate the strengths and weaknesses of each architecture in terms of the coverage provided for the functional and structural requirements. In our analysis, we detect a gap in this area: there is currently no architecture fulfilling all the requirements of future automotive GW controllers. In light of the available network processing architectures and the current technology landscape, we identify Hardware (HW) accelerators and custom processor design as a key differentiation factor which boosts the devices performance. From our perspective, this points to a need -and a research opportunity -to explore network processing architectures with a strong HW focus, unleashing the potential of next-generation network processors and supporting the demanding requirements of future autonomous and connected vehicles.

show abstract

Section: B Network Processors With Hw-based Network Offloading Capabi...mentioning

confidence: 99%