A Scalable, Partially Configurable Optical Switch for Data Center Networks

Modeling and performance analysis of a data center network (DCN) combining optical packet switching (OPS) and agile optical circuit switching (OCS) in a unified platform are presented. Hybrid optoelectronic packet routers which combine optical switching fabric and electronic buffers are used, supporting 100 Gbit/s (25 Gbit/s × 4 wavelengths) optical packets. OCS transmission is implemented by a novel “express path,” (ExP) an instant wavelength path on demand, which enables certain flows following the ExP with no packet contention, whereas ordinary OPS packets are switched with a deflection routing algorithm and buffering for contention resolution. This architecture can support OCS for large and reliability-sensitive data transferring and OPS for latency-sensitive applications. The performance of the network is investigated via numerical simulations of top-of-rack switch (ToR)-to-ToR traffic in an N-dimensional (up to 4096 node) torus topology. We have made the following observations: 1) torus topology results in high throughput and can be scaled out by increasing the dimension; 2) contention resolution between OPS and agile OCS is needed and our proposed strategies enhance the overall performance; and 3) in a lightly to moderately loaded case of OPS traffic, more than 100 ExPs can be supported simultaneously with acceptable affect on OPS.

Multi-tenancy is a key feature of modern data centers. It allows for the existence of multiple independent virtual infrastructures, called virtual slices, on top of the same physical infrastructure, each of them specially tailored to the tenants' needs. In such a scenario, an optimal mapping of the virtual slices plays a capital role toward an efficient utilization of the data center network resources, potentially saving costs for the data center owner. However, due to the increasing trend of bringing optics to data center networks, specific virtual slice mapping mechanisms accounting for the particularities of the optical medium (e.g., wavelength continuity constraint) have to be investigated. For this, we present an integer linear programming (ILP) model for optimally mapping a set of virtual slices from different tenants in a hybrid optical circuit switching (OCS)/optical packet switching (OPS) data center network with the aim to minimize the necessary optical transponders to be equipped in the network. Additionally, we also present a lightweight heuristic for the cases in which the ILP model scalability is compromised. The benefits of the proposals are highlighted by benchmarking them against a pure OCS solution through extensive simulations.

Scheduling traffic for maximum switch lifetime in optical data center fabrics

Data center networks (DCNs) are at the forefront of a radical transformation, fueled by the rapid rise of artificial intelligence (AI), cloud-native adoption, and enterprise digitalization, which together place unprecedented demands on their design and performance. From healthcare AI and intelligent transportation to business analytics, modern DCNs are expected to support a diverse and evolving mix of bandwidth-demanding applications, each with distinct quality of service (QoS) requirements. Yet, most state-of-the-art DCN solutions rely on oversimplified DCN data traffic models, failing to address the required application-level differentiation in terms of end-to-end (E2E) latency demands. In response, we propose ALAS-DCN, a novel, to our knowledge, application-aware, low-latency, scalable DCN architecture, based on optical packet switching in the intra-rack and inter-rack domains that is both designed and rigorously evaluated using commercial QoS specifications and realistic traffic patterns of different applications coexisting within the same DCN. ALAS-DCN integrates three hierarchical optical sub-networks: intra-rack, inter-rack, and bridge, each governed by a custom wavelength division multiplexing (WDM) medium access control (MAC) protocol that dynamically prioritizes critical traffic packets, according to the specific application E2E latency demands. Unlike prior approaches, our evaluation leverages realistic workload traces generated by established traffic generators, capturing the application diversity within modern data center (DC) environments. Extensive experimental simulations across five representative DC categories, each hosting a different mix of modern applications, demonstrate that ALAS-DCN consistently achieves up to 95% throughput efficiency and sub-µs latency for high priority traffic across a variety of applications, consistently meeting their differentiated latency constraints. Scalable up to 20,000 servers, ALAS-DCN redefines the DCN as not just a transport layer but an intelligent enabler of next-generation application performance.

In optical packet switching (OPS)-based data center networks (DCNs), it is preferred to set up temporal optical circuit switching (OCS) paths for large-data transmission that requires high reliability. However, inappropriate OCS path selection could severely affect the original OPS traffic. This is because an OCS path may break the network topology and cause OPS packet circulations. We propose to maintain bypassing routes for every link of the OCS path when it is established. Then, the OPS packets can bypass the OCS links by being deflected to the bypassing routes. Therefore, this method facilitates the coexistence of OCS in OPS-based DCNs. Numerical simulations demonstrate the benefit achieved by our proposal compared with non-bypassing-route scenarios.

As the scale of data centers and cloud computing applications increases, data center networks play a critical role in meeting the huge communication bandwidth requirement of such applications. The scalability of conventional electronic data center networks is limited by wiring complexity and reaching distance of links under fixed power budget. To overcome this problem, in this paper we propose a nonblocking hybrid switch architecture, called NEO (Nonblocking Electronic and Optical), which is able to provide nonblocking interconnections for as many as 1,000,000 servers in a data center. NEO maintains electronic interconnections for intra-pod networks, while providing interpod interconnections by optical core switches, which not only increases the scalability of the switch architecture, but also lowers the switch cost and power consumption compared to other existing optical switch architectures. We also design a packet scheduler for NEO, which adopts a credit flow control mechanism and a parallel scheduling algorithm to avoid packet loss, and provide low communication latency. Our simulation results demonstrate that NEO achieves very low average packet delay compared to other existing optical switching architectures under various traffic patterns.

All optical or hybrid electrical/optical switching data center network (DCN) architectures are capable of adapting the optical-layer topology to various network traffic demands based on the flexibility of optical switching technology. The topology reconfiguration strategies in previous research simply attempt to reconfigure optical circuit switching (OCS) links to offload heavier rack-to-rack traffic. However, due to the complex multi-layer interconnections and the various network traffic patterns in DCNs, the OCS links will not only offload the network traffic between its source and destination racks, but also forward the network traffic that belongs to the adjacent racks of the source and destinations racks. Previous strategies ignore the forwarding traffic and result in sub-optimal solutions. In order to overcome the shortcomings of previous methods and make full use of the OCS link resources in DCNs, an effective topology reconfiguration strategy named optical circuit bridge (OCBridge) has been proposed in our recent work. The main idea of our OCBridge is to take into account the traffic demands between the source and destination regions of each OCS link. In this paper, we first review our OCBridge, and then in simulation, we evaluate the performance of our OCBridge by testing it in two typical OCS-based DCNs, including OpenScale and Helios. Under our OCBridge, we analyze the outstanding network throughput performance of the DCNs and investigate the communication distance of traffic demands. Finally, in experiment, we deploy our OCBridge to a small prototype of OpenScale and verify that our OCBridge is capable of accelerating the distributed computing applications compared with previous topology reconfiguration strategies.

This paper presents performance evaluations and comparisons of different sharing approaches applied to the available optical bandwidth in data center networks. The recently proposed hybrid optical network concept, which combines multi-wavelength optical packet switching (OPS) and optical circuit switching (OCS) in WDM networks, is compared with a service-specific sharing model. Simulation results show that a careful design of dedicated channel allocation gives similar performance to a complete channel sharing with prioritized scheduling, which is however more flexible.

We present a control plane architecture to enable software-addressable optical acceleration from the application layer. The architecture is experimentally examined on a cluster-computing test-bed by enabling physical layer optical multicasting on-demand for the application layer to achieve non-blocking performance.

Dynamic optical interconnects using fast Optical Circuit Switches (OCS) are emerging as a scalable and energy efficient alternative to increasing network demands. Initial concerns regarding slow switching speeds of OCSs were recently overcome, with prototypes enabling circuit setup in a few microseconds. This can potentially broaden the classes of traffic patterns that can be carried efficiently by an alloptical interconnect. However, application performance on such newer interconnects has not been fully understood yet. In this paper, we explore the the gap between advances in faster OCS hardware and the potential success of such newer technologies in terms of application performance and cluster energy efficiency. We evaluate the performance of latencysensitive distributed applications running on a fast OCS environment, analyzing its impact to overall server and network energy efficiency. We also discuss scheduling inefficiencies of current fast OCSs and evaluate ideas to solve them. We find that while some distributed applications suffer minimal performance penalty when running on fast OCSs, more flexible schedulers, like the ones outlined in this paper, improve application performance and OCS efficiency by up to 2.44 times compared to the strategies in the literature. Dynamic optical interconnects using fast Optical Circuit Switches (OCS) are emerging as a scalable and energy efficient alternative to increasing network demands. Initial concerns regarding slow switching speeds of OCSs were recently overcome, with prototypes enabling circuit setup in a few microseconds. This can potentially broaden the classes of traffic patterns that can be carried efficiently by an all-optical interconnect. However, application performance on such newer interconnects has not been fully understood yet. In this paper, we explore the the gap between advances in faster OCS hardware and the potential success of such newer technologies in terms of application performance and cluster energy efficiency. We evaluate the performance of latency-sensitive distributed applications running on a fast OCS environment, analyzing its impact to overall server and network energy efficiency. We also discuss scheduling inefficiencies of current fast OCSs and evaluate ideas to solve them. We find that while some distributed applications suffer minimal performance penalty when running on fast OCSs, more exible schedulers, like the ones outlined in this paper, improve application performance and OCS efficiency by up to 2.44 times compared to the strategies in the literature.

Optical Circuit Switching (OCS) provides a good support for the services requiring high bandwidth and Quality of Service (QoS). When the character of network traffic changes from voice-oriented to data-oriented, OCS must cooperate with new switching mechanisms such as Optical Burst Switching (OBS) to improve performance. So a Hybrid Optical Switching (HOS) system is proposed. In HOS, different traffics are classified and both OBS and OCS can be processed. At the same time, delay buffer mechanisms, which are Fiber Delay Line (FDL) and electronic buffering, are introduced in HOS to solve the contention of OBS bursts. And a wavelengths classification mechanism, which assigns some wavelengths dedicated to OCS traffics, is provided to guarantee the QoS of OCS. To analyze the performance in theory, models based on Markovian theory and iterative method are provided. The results got by theory are accurate compared with simulations. The use of delay buffer can decrease the blocking probability of OBS and increase resource utilization. Wavelength classification mechanism benefits to improve OCS's performance.

Optical switching based small-world data center network

We study the fundamental properties of the hybrid combination of optical circuit switching (OCS) and electrical packet switching (EPS) used in an ultra-high-speed local area network (LAN) of the future, where chunks of data such as those containing non-compressed super-high-definition images are frequently transmitted as well as text-based data. We assume the logical topology of the OCS network in such a LAN to be single star in which the end-to-end connections are provided through a non-blocking optical matrix switch located at the center, while a conventional modest-speed EPS-based network such as Ethernet is co-installed to function as another data plane as well as the control plane of the OCS network. We refer to queueing theory in order to understand the fundamental performances issues of OCS and packet switching, and show that the hybrid use of OCS and EPS is very preferable especially when the variance of file size is very large. Simulation results show that the hybrid configuration can significantly outperform a network based on OCS only.

DWDM multi-mode switching network provides a unified approach to DWDM-based communications by allowing electronic packet switching (EPS), optical burst switching (OBS) and optical circuit switching (OCS) to be carried in the same network, and on the same router platform. In this paper, we present a sub-millisecond dynamic optical path setup scheme for the multi-mode switching router. More specifically, we devised a reconfigurable asymmetric optical burst switching (RA-OBS) scheduling scheme which integrates burst scheduling in the OBS mode, connection setup/teardown in the OCS mode, and channel reconfiguration in the EPS mode in a seamless fashion. The proposed scheme allows wavelengths to be dynamically shared among three different switching modes in real-time. The performance of the proposed integrated scheduler has been verified using OBS-ns2 simulator as well as in a hardware testbed. In particular, we have implemented the proposed integrated multi-mode scheduler in FPGA hardware in an optical switching testbed, and have demonstrated application triggered real-time sub-millisecond optical path setup. Another key contribution is that the proposed approach brings out the practical value of optical burst switching, which has stayed mostly in the theoretical regime in the past decade. We show that a truly dynamic optical network can be built with off-the-shelf optical components.

The dynamic traffic changes in an optical circuit switching based data-center network (ODCN) can make the utilization of its optical connections fragmented, degrading the efficiency of service provisioning in the ODCN. Consequently, connection defragmentation becomes imperative. However, consolidating traffic onto fewer connections may unilaterally add the risk of bandwidth contention and undermine ODCNs’ robustness against unexpected traffic bursts. To address this issue, we propose risk-aware connection defragmentation (RA-cDF), which explores the topology flexibility of ODCN to consolidate traffic such that the active optical connections through optical circuit switching (OCS) switches can be minimized together with the risk of future bandwidth contention on remaining connections. We formulate a mixed integer linear programming (MILP) model to address the RA-cDF problem exactly, followed by a heuristic to solve it time-efficiently. Extensive simulations confirm the effectiveness of our proposals.