This article reports on the design and the experimental realization of packet-switched optical network (PSON) for a data center network. In PSON, an optical interconnect consists of a combination of a fast tunable laser (FTL), an arrayed waveguide grating router, and a burst mode receiver (BMR). The conventional FTL transmitters can only accommodate short fixed-size photonic frame (PF) of tens of microseconds rather than variable-size PF due to the short-term wavelength stability. By using the FTL with a thermal compensation method, the PSON can support variable-size PFs with extended size to minimize the switching overhead in the optical switch domain. Our PSON system can achieve up to 96% throughput of 10 GbE service as the end-to-end optical switching time is reduced to 1.94  $\mu$ s that requires only 3  $\mu$ s guard-time in a 45  $\mu$ s time-slot. Total switching time includes wavelength conversion time of FTL, response time of BMR and burst mode-clock and data recovery (BM-CDR). Our PSON will achieve near non-blocking scheduling by use of a centralized scheduler running a grant-aware (GA) scheduling algorithm, which can be further improved to be non-blocking scheduling with a novel algorithm called C-RRP. The C-RRP scheduling algorithm is designed especially for the non-uniform distribution. Simulation results show that C-RRP improves the throughput and latency performance in hot-spot conditions.

Experimental Demonstration of a Photonic Frame Based Packet-Switched Optical Network for Data Centers

We introduce a scheduling algorithm for all-optical data center switch supporting high throughput and low delay for non-uniform traffic condition. Simulation result shows that our system achieves 100% throughput and shorter delay than existing methods.

Scheduling Algorithm for All-optical Switch under Non-uniform Traffic Condition

In this study, we propose a new wavelength-division-multiplexing passive optical network (WDM-PON) system to support the blended fiber-free space optics (FSO) signal access. To provide the fiber and FSO traffics simultaneously, the C- and L-band channels are applied in the presented PON, respectively. Moreover, to avoid the fiber breakpoint in the fiber access traffic, the proposed WDM access architecture also can provide the self-restored mechanism by applying simple fiber routing path. In addition, the corresponding signal performances of fiber and FSO channels are also executed experimentally for demonstration. 

/pdf/integrated-fiber-fso-wdm-access-system-with-fiber-fault-27qo1y8g.pdf

Integrated Fiber-FSO WDM Access System with Fiber Fault Protection

Crossbar switch scheduling algorithms for high performance computing: A comprehensive review

Systems and Methods for switching optical data units (ODUs) and Internet Protocol (IP) packets as Ethernet packets in an optical transport network (OTN), IP, and Ethernet switching system. The OTN, IP, and Ethernet switching system may include an Ethernet fabric having a set of M Ethernet switches each including a set of N switch ports, and a set of N input/output (IO) devices each including a set of W IO ports, a set of M Ethernet ports, an IO side packet processor (IOSP), and a fabric side packet processor (FSP). Each Ethernet switch may establish switch queues. Each IO device may establish a set of M hierarchical virtual output queues each including a set of N ingress-IOSP queues and ingress-virtual output queues, a set of W egress-IOSP queues, a set of M ingress-FSP queues, and a set of N hierarchical virtual input queues each including a set of N egress-FSP queues and egress-virtual input queues.

Disaggregated hybrid optical transport network, internet protocol, and Ethernet switching system

An optical access network plays a critical role in accommodating explosive new services such as augmented reality (AR), virtual reality (VR), machine-to-machine, and human-to-machine interaction applications. Continuous expansion of capacity as well as low latency in the optical access network are very important issues to avoid loss of connectivity for 5G and Tactile Internet. A time-division multiplexing passive optical network (TDM-PON) is an attractive solution to provide optical connectivity to end users in residential, business, and mobile applications since multiple remote nodes with simple optical power splitters give us easy-to-use optical connection anywhere in an optical distributed network. This article reviews TDM-PON-based optical access technologies for bandwidth-intensive as well as low-latency services, and introduces a recent feasibility demonstration of a Tactile Internet testbed. Technical challenges faced by TDM-PON for future networks in terms of capacity, latency, and virtual-izationlslicing are also discussed.

TDM-PON-Based Optical Access Network for Tactile Internet, 5G, and Beyond

https://onlinelibrary.wiley.com/doi/pdf/10.4218/etrij.2017-0057

Grant-Aware Scheduling Algorithm for VOQ-Based Input-Buffered Packet Switches

We propose new scheduling algorithms to support low latency packet switching targeting for all-optical data center switch. Based on simulation result, our proposed scheduling algorithms significantly improve the average delay and maximum delay compared with the existing algorithms. Especially when the traffic load is very high, the average delay of our algorithm is reduced to only 10% of the existing algorithms.

Low delay switch scheduling for data center optical switch

This paper introduces new scheduling algorithms supporting low-latency switching. The proposed grant-aware (GA) algorithm improves the average delay performance by using the grant information of previous iteration. The simulation result shows that the average delay of GA algorithm is about one-tenth of the existing algorithm in high-load condition. We also introduce two priority-based scheduling algorithms grant-aware and priority-aware (GAPA) algorithm and cyclic scheduling with the longest-queue-first (C-LQF) algorithm. In the priority-based scheduling, the scheduling priority of VoQ is determined based on its queue size. GAPA and C-LQF consider the priority only after the first iteration to prevent the starvation problem. The simulation result shows that GAPA and C-LQF scheduling achieves better performance than GA in terms of average delay, maximum delay, and hotspot throughput.

Chansung Park

Papers

TDM-PON-Based Optical Access Network for Tactile Internet, 5G, and Beyond

Grant-Aware Scheduling Algorithm for VOQ-Based Input-Buffered Packet Switches

Scheduling Algorithm for All-optical Switch under Non-uniform Traffic Condition

Low delay switch scheduling for data center optical switch

Priority-based Grant-aware Scheduling for Low-latency Switching