Kenji Mizutani
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Kenji Mizutani has a total of 34 co-authors. Here are the top co-authors by number of shared publications:
- Yutaka Urino
Photonics Electronics Technology Research Association (PETRA)
5 shared publications Hiroshi Yamaguchi
Photonics Electronics Technology Research Association
3 shared publications- Michihiro Koibuchi
National Institute of Informatics, Chiyoda-ku, Tokyo, Japan
3 shared publications - Tatsuya Usuki
Photonics Electronics Technology Research Association (PETRA)
3 shared publications Kenji Sato
Nanoelectron. Res. Labs., NEC Corp., Otsu
2 shared publications- Kiyo Ishii
National Institute of Advanced Industrial Science and Technology (AIST), Tsukuba, Japan
2 shared publications
Latest Publications
- Open Access
- 8 September 2025
- Published by Association for Computing Machinery (ACM)
- p. 16-18
NVMe-oF Booster: Enabling Long-Range Remote Storage Access over Optical Paths
- 8 September 2025
- Published by Association for Computing Machinery (ACM)
- p. 16-18
Open AccessNVMe-oF Booster: Enabling Long-Range Remote Storage Access over Optical Paths
- 7 November 2024
- Published by Optica Publishing Group in Journal of Optical Communications and Networking
- Vol. 17 (1) , A34-A45
In-network stable radix sorter using many FPGAs with high-bandwidth photonics [Invited]
A modern field-programmable gate array (FPGA) card can be equipped with high-bandwidth memory and high-bandwidth optical interconnection networks. This paper presents an in-network stable radix sorter on an eight-FPGA cluster. Each custom Stratix10 MX2100 FPGA card has up to 819-Gbps memory bandwidth (51.2Gbps×16channels) and up to 800-Gbps network bandwidth (25Gbps×32channels) with eight custom embedded optical modules. Our radix sorter for a 32-bit key range consists of eight 4-bit counting sorts optimized to in-network processing. Each counting sort needs only one memory read/write access for improving its throughput. We demonstrated a sorting throughput of 37.2 GB/s and an energy efficiency of 9.2 MB/J for 32-bit key range and 16-GiB data size using eight memory channels with 409.6 Gbps memory bandwidth per FPGA. It can scale up to 256 FPGAs with a sorting throughput of 983 GB/s for a 32-bit key range and 512-GiB data size.
- 7 November 2024
- Published by Optica Publishing Group in Journal of Optical Communications and Networking
- Vol. 17 (1) , A34-A45
In-network stable radix sorter using many FPGAs with high-bandwidth photonics [Invited]
A modern field-programmable gate array (FPGA) card can be equipped with high-bandwidth memory and high-bandwidth optical interconnection networks. This paper presents an in-network stable radix sorter on an eight-FPGA cluster. Each custom Stratix10 MX2100 FPGA card has up to 819-Gbps memory bandwidth (51.2Gbps×16channels) and up to 800-Gbps network bandwidth (25Gbps×32channels) with eight custom embedded optical modules. Our radix sorter for a 32-bit key range consists of eight 4-bit counting sorts optimized to in-network processing. Each counting sort needs only one memory read/write access for improving its throughput. We demonstrated a sorting throughput of 37.2 GB/s and an energy efficiency of 9.2 MB/J for 32-bit key range and 16-GiB data size using eight memory channels with 409.6 Gbps memory bandwidth per FPGA. It can scale up to 256 FPGAs with a sorting throughput of 983 GB/s for a 32-bit key range and 512-GiB data size.
- 1 January 2024
- Published by Optica Publishing Group
- p. M2G.6
Performance of Radix Sort using All-to-all Optical Interconnection Network in an Eight-FPGA Cluster
This paper presents a high-throughput dataflow processing using all-to-all communication with eight FPGAs. We demonstrated a parallel radix sorting throughput of 37.2 GB/s for 32-bit key range and 16-GiB data size.
- 1 January 2024
- Published by Optica Publishing Group
- p. M2G.6
Performance of Radix Sort using All-to-all Optical Interconnection Network in an Eight-FPGA Cluster
This paper presents a high-throughput dataflow processing using all-to-all communication with eight FPGAs. We demonstrated a parallel radix sorting throughput of 37.2 GB/s for 32-bit key range and 16-GiB data size.
- 31 January 2022
- Published by Optica Publishing Group in Journal of Optical Communications and Networking
- Vol. 14 (2) , A166-A179
Accelerating parallel data processing using optically tightly coupled FPGAs
A cutting-edge field programmable gate array (FPGA) card can be equipped with high-bandwidth inputs and outputs by high-density optical integration, e.g., onboard Si-photonics transceivers or co-packaged optics. It is possible to implement highly parallel data processing by using tightly coupled FPGAs. We previously proposed a lightweight, fully connected inter-FPGA network called OPTWEB for efficient collective communications. OPTWEB discards the traditional packet communication mechanism and introduces dedicated memory processing to minimize communication overhead. In this paper, we provide the design guidelines for a parallel data application on FPGAs connected by OPTWEB. We then present a generalized diameter-two hypercube network topology for connecting multiple FPGAs, which enables larger applications. In the guidelines, data to be exchanged should be directly processed by the dedicated memory on each FPGA, and parallelism should be implemented to match the collective communications provided by OPTWEB. We illustrate two case studies, one on parallel counting sort and the other on sharing averaged parameters, on eight custom Stratix10 MX2100 FPGA cards with 800 Gbps network bandwidth. The parallel counting sort operates up to 5.3 times faster than a comparable CPU cluster with the conventional InfiniBand network, and with sharing averaged parameters, up to 250 times faster operation is achieved. The two applications utilize a total of 80.1% and 60.6% of the adaptive logic modules of the FPGA, respectively.
- 31 January 2022
- Published by Optica Publishing Group in Journal of Optical Communications and Networking
- Vol. 14 (2) , A166-A179
Accelerating parallel data processing using optically tightly coupled FPGAs
A cutting-edge field programmable gate array (FPGA) card can be equipped with high-bandwidth inputs and outputs by high-density optical integration, e.g., onboard Si-photonics transceivers or co-packaged optics. It is possible to implement highly parallel data processing by using tightly coupled FPGAs. We previously proposed a lightweight, fully connected inter-FPGA network called OPTWEB for efficient collective communications. OPTWEB discards the traditional packet communication mechanism and introduces dedicated memory processing to minimize communication overhead. In this paper, we provide the design guidelines for a parallel data application on FPGAs connected by OPTWEB. We then present a generalized diameter-two hypercube network topology for connecting multiple FPGAs, which enables larger applications. In the guidelines, data to be exchanged should be directly processed by the dedicated memory on each FPGA, and parallelism should be implemented to match the collective communications provided by OPTWEB. We illustrate two case studies, one on parallel counting sort and the other on sharing averaged parameters, on eight custom Stratix10 MX2100 FPGA cards with 800 Gbps network bandwidth. The parallel counting sort operates up to 5.3 times faster than a comparable CPU cluster with the conventional InfiniBand network, and with sharing averaged parameters, up to 250 times faster operation is achieved. The two applications utilize a total of 80.1% and 60.6% of the adaptive logic modules of the FPGA, respectively.
- Open Access
- 24 March 2021
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Computers
- Vol. 70 (6) , 849-862
OPTWEB: A Lightweight Fully Connected Inter-FPGA Network for Efficient Collectives
Modern FPGA can be equipped with many high-bandwidth network I/Os enabled by onboard optics. Some dozens of tightly-coupled FPGAs form an emerging computing platform for distributed processing. However, a conventional indirect packet network imposes an unacceptably large amount of the logic for handling such high-bandwidth interconnects on an FPGA. Besides an indirect network, another approach builds a direct packet network. Existing direct inter-FPGA networks have a low-radix network topology. However, the low-radix network has the disadvantage of a large diameter and large average shortest path length that increases the latency of collectives. To mitigate both problems, we propose a lightweight, fully-connected inter-FPGA network called OPTWEB. Since all end-to-end separate communication paths are statically established using onboard optics, raw block data can be transferred with simple link-level synchronization. Once each source FPGA assigns a communication stream to a path by its internal switch logic between memory-mapped and stream interfaces for RDMA, a one-hop transfer is provided. Since each FPGA performs input /output of RDMA between all FPGAs simultaneously, multiple RDMAs efficiently form collectives.The OPTWEB network provides 0.71-sec start-up latency of collectives among multiple FPGA cards with onboard optics. The OPTWEB network consumes 57.7% of ALMs for aggregate 800-Gbps interconnect.
- 24 March 2021
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Computers
- Vol. 70 (6) , 849-862
Open AccessOPTWEB: A Lightweight Fully Connected Inter-FPGA Network for Efficient Collectives
Modern FPGA can be equipped with many high-bandwidth network I/Os enabled by onboard optics. Some dozens of tightly-coupled FPGAs form an emerging computing platform for distributed processing. However, a conventional indirect packet network imposes an unacceptably large amount of the logic for handling such high-bandwidth interconnects on an FPGA. Besides an indirect network, another approach builds a direct packet network. Existing direct inter-FPGA networks have a low-radix network topology. However, the low-radix network has the disadvantage of a large diameter and large average shortest path length that increases the latency of collectives. To mitigate both problems, we propose a lightweight, fully-connected inter-FPGA network called OPTWEB. Since all end-to-end separate communication paths are statically established using onboard optics, raw block data can be transferred with simple link-level synchronization. Once each source FPGA assigns a communication stream to a path by its internal switch logic between memory-mapped and stream interfaces for RDMA, a one-hop transfer is provided. Since each FPGA performs input /output of RDMA between all FPGAs simultaneously, multiple RDMAs efficiently form collectives.The OPTWEB network provides 0.71-sec start-up latency of collectives among multiple FPGA cards with onboard optics. The OPTWEB network consumes 57.7% of ALMs for aggregate 800-Gbps interconnect.
- 1 January 2021
- Published by Optica Publishing Group
- p. Th5H.1
Accelerating Parallel Sort on Tightly-Coupled FPGAs enabled by Onboard Si-Photonics Transceivers
We illustrate a case for a tightly-coupled lightweight fully-connected inter-FPGA optical network with Si-photonics transceivers. Counting sort of in-memory column-oriented data using eight custom FPGA cards can be 5.3 times faster than that with eight CPU-servers.
- 1 January 2021
- Published by Optica Publishing Group
- p. Th5H.1
Accelerating Parallel Sort on Tightly-Coupled FPGAs enabled by Onboard Si-Photonics Transceivers
We illustrate a case for a tightly-coupled lightweight fully-connected inter-FPGA optical network with Si-photonics transceivers. Counting sort of in-memory column-oriented data using eight custom FPGA cards can be 5.3 times faster than that with eight CPU-servers.
- Open Access
- 15 January 2020
- Published by Association for Computing Machinery (ACM)
- p. 81-91
Wavelength-routing interconnect "Optical Hub" for parallel computing systems
- 15 January 2020
- Published by Association for Computing Machinery (ACM)
- p. 81-91
Open AccessWavelength-routing interconnect "Optical Hub" for parallel computing systems
- 22 December 2014
- Published by Institute of Electrical and Electronics Engineers (IEEE) in Journal of Lightwave Technology
- Vol. 33 (6) , 1223-1229
First Demonstration of Athermal Silicon Optical Interposers With Quantum Dot Lasers Operating up to 125 °C
We previously proposed a photonics-electronics convergence system to solve bandwidth bottleneck problems among large-scale integrations (LSIs) and demonstrated a high bandwidth density with silicon optical interposers at room temperature. For practical applications, the interposers should be usable under high-temperature conditions or rapid temperature changes so that they can cope with the heat generated by the mounted LSIs. We designed and fabricated athermal silicon optical interposers integrated with temperature-insensitive components on a silicon substrate. An arrayed laser diode (LD) chip was a flip-chip bonded to the substrate. Each LD had multiple quantum dot layers with a 1.3-μm lasing wavelength. The output power was higher than 10 mW per channel up to 100 °C. Silicon optical modulator and germanium photodetector (PD) arrays were monolithically integrated on the substrate. The modulators were structured as symmetric Mach-Zehnder interferometers, which were inherently insensitive to temperature and wavelength. The phase shifters composed of p-i-n diodes were stable against temperature change under a constant bias-current condition. The PD photocurrent was also temperature insensitive, and the photo-to-dark current ratio was higher than 30 dB up to 100 °C. We achieved error-free data links at 20 Gbps and a high bandwidth density of 19 Tbps/cm 2 operating up to 125 °C without adjusting the LDs, modulators, or PDs. The interposers are tolerant of the heat generated by the mounted LSIs and usable over the extended industrial temperature range without complex monitoring or feedback controls. The bandwidth density is sufficient for the needs of the late 2010s.
- 22 December 2014
- Published by Institute of Electrical and Electronics Engineers (IEEE) in Journal of Lightwave Technology
- Vol. 33 (6) , 1223-1229
First Demonstration of Athermal Silicon Optical Interposers With Quantum Dot Lasers Operating up to 125 °C
We previously proposed a photonics-electronics convergence system to solve bandwidth bottleneck problems among large-scale integrations (LSIs) and demonstrated a high bandwidth density with silicon optical interposers at room temperature. For practical applications, the interposers should be usable under high-temperature conditions or rapid temperature changes so that they can cope with the heat generated by the mounted LSIs. We designed and fabricated athermal silicon optical interposers integrated with temperature-insensitive components on a silicon substrate. An arrayed laser diode (LD) chip was a flip-chip bonded to the substrate. Each LD had multiple quantum dot layers with a 1.3-μm lasing wavelength. The output power was higher than 10 mW per channel up to 100 °C. Silicon optical modulator and germanium photodetector (PD) arrays were monolithically integrated on the substrate. The modulators were structured as symmetric Mach-Zehnder interferometers, which were inherently insensitive to temperature and wavelength. The phase shifters composed of p-i-n diodes were stable against temperature change under a constant bias-current condition. The PD photocurrent was also temperature insensitive, and the photo-to-dark current ratio was higher than 30 dB up to 100 °C. We achieved error-free data links at 20 Gbps and a high bandwidth density of 19 Tbps/cm 2 operating up to 125 °C without adjusting the LDs, modulators, or PDs. The interposers are tolerant of the heat generated by the mounted LSIs and usable over the extended industrial temperature range without complex monitoring or feedback controls. The bandwidth density is sufficient for the needs of the late 2010s.
- 5 June 2009
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Selected Topics in Quantum Electronics
- Vol. 15 (3) , 488-493
A Monolithic Wideband Wavelength-Tunable Laser Diode Integrated With a Ring/MZI Loop Filter
We propose a widely tunable and low insertion loss loop filter incorporating two ring resonators and an asymmetric Mach-Zehnder interferometer. A semiconductor optical amplifier, phase control section, and loop filter are monolithically integrated on one chip in the InP-InGaAsP material system. The tuning range of 39 nm with side-mode suppression ratio of over 30 dB was demonstrated. The maximum of the total heater power over the tuning range was 84 mW.
- 5 June 2009
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Selected Topics in Quantum Electronics
- Vol. 15 (3) , 488-493
A Monolithic Wideband Wavelength-Tunable Laser Diode Integrated With a Ring/MZI Loop Filter
We propose a widely tunable and low insertion loss loop filter incorporating two ring resonators and an asymmetric Mach-Zehnder interferometer. A semiconductor optical amplifier, phase control section, and loop filter are monolithically integrated on one chip in the InP-InGaAsP material system. The tuning range of 39 nm with side-mode suppression ratio of over 30 dB was demonstrated. The maximum of the total heater power over the tuning range was 84 mW.
- 15 October 2007
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Selected Topics in Quantum Electronics
- Vol. 13 (5) , 1104-1111
Integration of Functional SOA on the Gain Chip of an External Cavity Wavelength Tunable Laser Using Etched Mirror Technology
A gap-mirror technology for integrating a functional section onto a laser gain chip is presented. A semiconductor optical amplifier, implementing several functions, is integrated on the gain chip used in an external cavity laser, and demonstrated as a high-performance variable attenuator and data modulator at 2.5 Gb/s. Control of the mirror properties through the gap geometry is demonstrated, and further optimization is realized by a gap-filling technique, which increases the output power to above 100 mW across the C -band.
- 15 October 2007
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Selected Topics in Quantum Electronics
- Vol. 13 (5) , 1104-1111
Integration of Functional SOA on the Gain Chip of an External Cavity Wavelength Tunable Laser Using Etched Mirror Technology
A gap-mirror technology for integrating a functional section onto a laser gain chip is presented. A semiconductor optical amplifier, implementing several functions, is integrated on the gain chip used in an external cavity laser, and demonstrated as a high-performance variable attenuator and data modulator at 2.5 Gb/s. Control of the mirror properties through the gap geometry is demonstrated, and further optimization is realized by a gap-filling technique, which increases the output power to above 100 mW across the C -band.