Abstract:When I started out at UC Santa Barbara as a PhD student in 2015, there was no funding until the AIM project proposal got accepted. Without funding no work gets done, so first I am grateful to Prof. Rod Alferness for funding me for the first year and Prof. John Bowers for funding me for the rest of my PhD. To John, thanks for giving me complete freedom at work without which I could not have explored so many different areas in research. I enjoyed our conversation in the monthly meetings with Rod, John and Adel. … Show more
“…The number of cascaded MRRs on the path of the reconfigured channels in a Flex-LIONS with the Beneš MZS network is two while that of a Flex-LIONS with multi-wavelength MRR crossbar is three, so that the bandwidth-narrowing effect is reduced. Table 1 compares Flex-LIONS with various state-of-theart wavelength-and-space selective reconfigurable switching fabrics, including indium phosphide (InP) AWGRs + semiconductor optical amplifier (SOA) gates [13], silicon (Si) echelle gratings + (micro-electro-mechanical system) MEMS arrays [14], and multi-wavelength selective crossbar [15]. In particular, the comparison study takes into account the port count, on-chip loss, and the number of switching elements.…”
Section: Working Principlementioning
confidence: 99%
“…[13] architecture to scale up to high radix. Reference [14] architecture suffers not only from the high number of switching elements (N 3 ) but also from high on-chip insertion loss since the number of waveguide crossings increases by ∼ N 2 , while the number of waveguide crossings in Flex-LIONS increases by ∼ N. Reference [15] architecture also has the issue of a high number of switching elements which makes the control plane more complex and limits the scalability. Compared with Flex-LIONS with multi-wavelength MRR crossbar, Flex-LIONS with the Beneš MZS network exhibits a lower number of switching elements and reduced bandwidth-narrowing effect at the expense of higher on-chip insertion loss and a more complex control mechanism due to the rearrangeable non-blocking nature of this solution.…”
Section: Comparison With Other Approachesmentioning
confidence: 99%
“…At the physical layer, optical interconnects are becoming the dominant communication technology in HPC and datacenters driven by the ever-increasing bandwidth scaling pushed by the widespread adoption of the cloud and emerging applications. In the past few years, we have witnessed a number of different integrated reconfigurable wavelength routing and space switching solutions that allow redefining the connectivity in both spectral and spatial domains, dynamically [6]- [15] Among these works, in [7] we proposed and demonstrated SiPh Flex-LIONS (silicon photonic flexible low-latency interconnect optical network switch), a reconfigurable photonic interconnect architecture that leverages both wavelength routing and spatial switching to spatially and temporarily steer and increase the communication bandwidth between specific node pairs (a comparison between Flex-LIONS and other existing approaches is presented in Sect. 2.2).…”
This paper summarizes our recent studies on architecture, photonic integration, system validation and networking performance analysis of a flexible low-latency interconnect optical network switch (Flex-LIONS) for datacenter and high-performance computing (HPC) applications. Flex-LIONS leverages the all-to-all wavelength routing property in arrayed waveguide grating routers (AWGRs) combined with microring resonator (MRR)-based add/drop filtering and multi-wavelength spatial switching to enable topology and bandwidth reconfigurability to adapt the interconnection to different traffic profiles. By exploiting the multiple free spectral ranges of AWGRs, it is also possible to provide reconfiguration while maintaining minimum-diameter all-to-all interconnectivity. We report experimental results on the design, fabrication, and system testing of 8 × 8 silicon photonic (SiPh) Flex-LIONS chips demonstrating error-free all-to-all communication and reconfiguration exploiting different free spectral ranges (FSR 0 and FSR 1 , respectively). After reconfiguration in FSR 1 , the bandwidth between the selected pair of nodes is increased from 50 Gb/s to 125 Gb/s while an all interconnectivity at 25 Gb/s is maintained using FSR 0. Finally, we investigate the use of Flex-LIONS in two different networking scenarios. First, networking simulations for a 256-node datacenter inter-rack communication scenario show the potential latency and energy benefits when using Flex-LIONS for optical reconfiguration based on different traffic profiles (a legacy fat-tree architecture is used for comparison). Second, we demonstrate the benefits of leveraging two FSRs in an 8-node 64-core computing system to provide reconfiguration for the hotspot nodes while maintaining minimum-diameter all-to-all interconnectivity.
“…The number of cascaded MRRs on the path of the reconfigured channels in a Flex-LIONS with the Beneš MZS network is two while that of a Flex-LIONS with multi-wavelength MRR crossbar is three, so that the bandwidth-narrowing effect is reduced. Table 1 compares Flex-LIONS with various state-of-theart wavelength-and-space selective reconfigurable switching fabrics, including indium phosphide (InP) AWGRs + semiconductor optical amplifier (SOA) gates [13], silicon (Si) echelle gratings + (micro-electro-mechanical system) MEMS arrays [14], and multi-wavelength selective crossbar [15]. In particular, the comparison study takes into account the port count, on-chip loss, and the number of switching elements.…”
Section: Working Principlementioning
confidence: 99%
“…[13] architecture to scale up to high radix. Reference [14] architecture suffers not only from the high number of switching elements (N 3 ) but also from high on-chip insertion loss since the number of waveguide crossings increases by ∼ N 2 , while the number of waveguide crossings in Flex-LIONS increases by ∼ N. Reference [15] architecture also has the issue of a high number of switching elements which makes the control plane more complex and limits the scalability. Compared with Flex-LIONS with multi-wavelength MRR crossbar, Flex-LIONS with the Beneš MZS network exhibits a lower number of switching elements and reduced bandwidth-narrowing effect at the expense of higher on-chip insertion loss and a more complex control mechanism due to the rearrangeable non-blocking nature of this solution.…”
Section: Comparison With Other Approachesmentioning
confidence: 99%
“…At the physical layer, optical interconnects are becoming the dominant communication technology in HPC and datacenters driven by the ever-increasing bandwidth scaling pushed by the widespread adoption of the cloud and emerging applications. In the past few years, we have witnessed a number of different integrated reconfigurable wavelength routing and space switching solutions that allow redefining the connectivity in both spectral and spatial domains, dynamically [6]- [15] Among these works, in [7] we proposed and demonstrated SiPh Flex-LIONS (silicon photonic flexible low-latency interconnect optical network switch), a reconfigurable photonic interconnect architecture that leverages both wavelength routing and spatial switching to spatially and temporarily steer and increase the communication bandwidth between specific node pairs (a comparison between Flex-LIONS and other existing approaches is presented in Sect. 2.2).…”
This paper summarizes our recent studies on architecture, photonic integration, system validation and networking performance analysis of a flexible low-latency interconnect optical network switch (Flex-LIONS) for datacenter and high-performance computing (HPC) applications. Flex-LIONS leverages the all-to-all wavelength routing property in arrayed waveguide grating routers (AWGRs) combined with microring resonator (MRR)-based add/drop filtering and multi-wavelength spatial switching to enable topology and bandwidth reconfigurability to adapt the interconnection to different traffic profiles. By exploiting the multiple free spectral ranges of AWGRs, it is also possible to provide reconfiguration while maintaining minimum-diameter all-to-all interconnectivity. We report experimental results on the design, fabrication, and system testing of 8 × 8 silicon photonic (SiPh) Flex-LIONS chips demonstrating error-free all-to-all communication and reconfiguration exploiting different free spectral ranges (FSR 0 and FSR 1 , respectively). After reconfiguration in FSR 1 , the bandwidth between the selected pair of nodes is increased from 50 Gb/s to 125 Gb/s while an all interconnectivity at 25 Gb/s is maintained using FSR 0. Finally, we investigate the use of Flex-LIONS in two different networking scenarios. First, networking simulations for a 256-node datacenter inter-rack communication scenario show the potential latency and energy benefits when using Flex-LIONS for optical reconfiguration based on different traffic profiles (a legacy fat-tree architecture is used for comparison). Second, we demonstrate the benefits of leveraging two FSRs in an 8-node 64-core computing system to provide reconfiguration for the hotspot nodes while maintaining minimum-diameter all-to-all interconnectivity.
“…The first demonstration of a μm-scale silicon ring resonator by Xu et al stimulated the research of MRR-based photonic integrated circuits [20]. There have been a number of demonstrations of microringbased optical switch fabrics, scaling from 2 to 8 ports [9], [17], [18], [21]- [26]. Table I summarises notable demonstrations of microring-based optical switching circuits chronologically.…”
Section: State-of-the-art Of Microring-based Silicon Optical Switmentioning
confidence: 99%
“…Key metrics, such as port count, insertion loss, crosstalk, and optical power penalty are highlighted and compared. This type of switch tends to utilize the MRR element as a 1 × 2 spatial adddrop unit assembled in certain topologies, such as the hitless [17] and five-port [21] routers, the crossbar switches [9], [23], [26], and the Omega switch with dual-ring intersections [18]. In these configurations, while each 1 × 2 elementary cell is only traversed by one signal, the crosstalk leakage gets aggregated with adjacent channels.…”
Section: State-of-the-art Of Microring-based Silicon Optical Switmentioning
We propose and analyze a scalable microring-based Clos switch fabric architecture constructed with switch-and-select switching stages. A silicon 4 × 4 building block that was designed and fabricated through American Institute for Manufacturing Integrated Photonics is used for the proof-of-principle demonstration of a 16 × 16 Clos switch fabric. By fully blocking the first-order crosstalk, the 4 × 4 device is measured to show a crosstalk ratio in the range of-57 to-48.5 dB, enabling better than-39 dB crosstalk for the 16 × 16 switch. Our study shows that the three-stage Clos design enables up to a factor of 4 in the reduction of the number of switching cells compared to single-stage switch-and-select fabrics. We further explore the design space for both first-order and secondorder switching elements using the foundry-validated parameters and how these factors impact the performance and scalability of the three-stage Clos switch. A detailed power penalty map is drawn for Clos switch fabrics with various scales, which reveals that the ultimate key limiting factor is the shuffle insertion loss. An optimized 32-port Clos switch fabric using foundry-enabled parameters is shown to have a less than 10-dB power penalty.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.