This article proposes a secure in-network computing system based on a simple reduced instruction set architecture, which can be used for processing artificial intelligence and machine learning models in network devices, in an AI-based Industrial Internet of Things (IoT) system. The architecture exploits the capabilities of upcoming generations of packet processing pipelines in programmable network switches. This instruction set enables processing of data at multiple terabits-per-second, which is beyond the processing power of current servers. Instructions for regular expressions, basic arithmetic, and logical operations are defined as a proof of concept. A packet containing both instruction and data blocks is presented as an input to the pipeline by bundling both the function and its arguments into the packet. The primary challenge in opening up network switches for executing a user-defined code is security. In this context, this article presents a secure execution model (SEM), which provides additional levels of security by deliberately disallowing memory allocation and modifications to persistent state of the network switch. Furthermore, real-life use cases are presented in this article to demonstrate the utility of the proposed instruction set architecture, as also applicable to IoT data processing. This instruction set is implemented in the programming protocol-independent packet processors language, verified on a mininet-based software switch and demonstrated on Xilinx NetFPGA SUME boards. The performance results show line rate packet processing with zero packet loss, at 10 Gb/s, and average packet latency of 3.66${\mu }\text{s}$.

With the emergence of programmable network devices that match the performance of fixed function devices, several recent projects have explored in-network computing, where the processing that is traditionally done outside the network is offloaded to the network devices. In-network computing has typically been applied to network functions (e.g., load balancing, NAT, and DNS), caching, data reduction/aggregation, and coordination/consensus functions. In some cases it has been used to accelerate stream-processing tasks that involve small payloads and simple operations. In this work we focus on leveraging in-network computing for stream processing of scientific datasets with large payloads that require complex operations such as floating-point computations and logarithmic functions. We demonstrate in-network computing for a real-world scientific application performing streaming normalization of a 2-D image from a light source experiment. We discuss the challenges we encountered and potential approaches to address them.

This paper deals with the efficient design of network switch/routers for an optical data center network. Each switch has multiple components such as ingress/egress interfaces, optical and/or electronic buffers, interconnection switching fabric and so on. There are several possible choices available for each of these components. This paper presents a systematic approach to designing the switch architecture using a combination of these component choices, while meeting specified design criteria. It requires formally defining the structure of a switch and enforcing semantics across components. This is formulated as a constraint optimization problem with formal language grammar guiding its search process. This problem formulation is used to identify the best-possible architecture for a hierarchical DCN. Two of the three solutions identified were new and were not reported in literature. These solutions were also validated experimentally.

Data center networks generate high volumes of traffic. In order to reduce packet latency, packet transmissions are often centrally scheduled. Such approaches have been proposed for both packet-switched and hybrid optical-packet switched networks. This paper investigates algorithm design choices for transmission scheduling in a tightly synchronized hybrid optical packet data center network. This problem is studied in two cases: with precedence where the requests are scheduled in the order of arrival, and without precedence, where the requests can be reordered in time. It is shown that the problem without any precedence constraints is NP-complete. For scheduling with precedence constraints, a greedy algorithm is proposed and shown to be optimal. Theoretical approximation for the performance of scheduling with the greedy algorithm is presented. Simulation experiments were performed on a two-tier network with 1024 servers and 64 wavelengths. Parallel implementation aspects of the scheduling algorithm are also discussed.

This paper investigates the time synchronization aspect of transmission scheduling in an optically groomed data center network (OGDCN). The architecture is based on a hybrid optical-packet approach and uses broadcast domains and wavelength division multiplexing for communication. The salient feature of this architecture is that all network paths are readily available in the optical domain and there is no need for optical path establishment. A source-destination pair must tune to a predefined wavelength at a scheduled time during data transfer. Every compute and storage node (CSN) is equipped with one or more tunable optical transceivers. As with any broadcast network, multiple transmitters that share a link segment cannot use the same wavelength on a link at the same time since this would result in collisions. Hence, transmission scheduling is required to prevent collisions and to allot an exclusive time duration for every transmission request. This paper's focus is on the time synchronization aspect that is critical for scheduling. Two schemes — continuous and discrete (slotted) time — are defined and evaluated. The objective of this paper is to understand the various factors affecting performance of these synchronization mechanisms. With the former scheme, clock accuracy has a significant impact on performance. With the latter scheme, propagation delay variance and packet length distribution impact performance in terms of utilization. The paper presents an evaluation of the performance of these mechanisms in the context of the OGDCN architecture. The results show that continuous time is able to better efficiently utilize network resources.

With data center network traffic growing significantly, power-efficient optical and hybrid optical architectures are considered as an alternate to packet switching. Typically, hybrid optical architectures use fast optical switching elements to ensure any-to-any route reachability. In this paper, an optically groomed data center network (OGDCN) framework is proposed. The proposed framework supports any-to-any route reachability without using fast optical switching elements in the network. Different components can be combined to realize an OGDCN, as described in this paper. The framework is evaluated and compared to other architectures in terms of scalability and power consumption. A particular OGDCN realization is presented and is shown to be better than other architectures in terms of power consumption. The power consumption is lower by at least 47% than the next best existing architecture proposed in the literature.

In this paper, we examine the effects of a power saving scheme in Passive Optical Networks (PON). Buffers in network equipment have been a cause of concern since they consume significant power. In a PON, the Optical Network Unit (ONU) buffers client node packets before forwarding to the Optical Line Terminal (OLT). We propose a scheme that reduces or eliminates buffers in the ONU thereby providing reduction in cost and power. However, this can potentially increase the packet delay due to buffering at the client node. This paper analyzes the delay performance using simulated and theoretical models. The increase in delay due to buffering at the EN instead of the ONU is found to be less than 300 microseconds, for the studied workloads; the total packet delay is also less than 800 microseconds. This is within the access network delay budget (typically 2 milliseconds) for different traffic types and loads.