NetCompute 2018 - 2018 Workshop on In-Network Computing (NetCompute 2018)

In-network computing is an emerging topic that draws a lot of attention from both academia and industry. In-network computing exploits the capabilities of new programmable network devices such as programmable switch ASICs, network processors, FPGAs, and programmable NICs to offload computing from CPUs to the network. While in-network computing can be dated back to early efforts such as active networking two decades ago, many believe that the time has finally come due to a combination of hardware and software innovations. On the hardware side, many hardware vendors have released products that provide programmability without sacrificing performance, such as Barefoot Tofino, Intel FlexPipe, Cavium XPliant, and Netronome Agilio. On the software side, besides new network functionalities such as in-network telemetry and layer-4 load balancing, many new application-level functionalities beyond traditional packet processing have been proposed, such as key-value caching, consensus, and even machine learning.
We believe that our community must expedite the research on in-network computing to make a profound influence on the theory and practice of future computer systems. There are significant open challenges: what are the killer apps of in-network computing; what are other application-level functionalities that can be offloaded to the network; fundamentally, what should be offloaded to the network and what should be kept on hosts; how to design better hardware primitives and software development platforms (e.g., domain-specific programming languages and compilers) to support the development of new applications; how to manage, monitor and debug heterogeneous systems that span servers and switches; how to incrementally deploy new applications; and many others.
The ACM SIGCOMM Workshop on In-Network Computing (NetCompute 2018) is organized with the goal of bringing researchers and engineers working on in-network computing together to present and discuss their latest ideas, research results and system experiences, thereby promoting the development and evolution of this area. All submissions on in-network computing related to architecture, design, implementation, simulation, modeling, analysis, and measurement are welcomed. We highly encourage novel and innovative early stage work that will encourage discussion and future research on in-network computing.
Topics of Interest
Applications of programmable network devices, e.g., machine learning, deep learning and key-value storage
Architectures for in-network computing
Control plane for in-network computing
Switch and NIC hardware design for in-network computing
Virtualization for in-network computing systems
Programming languages and compilers for in-network computing
Measurements and performance studies of programmable network devices
Diagnosing and troubleshooting in-network computing systems
In-network computing in clouds and edge clouds
Security and privacy of in-network computing
Deployment strategies and backward compatibility with traditional network functionalities
Experiences and best-practices of in-network computing systems in production