Real-time batch processing at a GPU-based edge with a passive optical network

Yukito Onodera; Yoshiaki Inoue; Daisuke Hisano; Naoto Yoshimoto; Yu Nakayama

doi:10.1364/JOCN.476116

Journal of Optical Communications and Networking
Vol. 15,
Issue 7,
pp. 404-414
(2023)
•https://doi.org/10.1364/JOCN.476116

Real-time batch processing at a GPU-based edge with a passive optical network

Yukito Onodera, Yoshiaki Inoue, Daisuke Hisano, Naoto Yoshimoto, and Yu Nakayama

Not Accessible

Your library or personal account may give you access

Get PDF
Email
Share
Get Citation
Copy Citation Text
Yukito Onodera, Yoshiaki Inoue, Daisuke Hisano, Naoto Yoshimoto, and Yu Nakayama, "Real-time batch processing at a GPU-based edge with a passive optical network," J. Opt. Commun. Netw. 15, 404-414 (2023)

Export Citation
- BibTex
- Endnote (RIS)
- HTML
- Plain Text
Citation alert
Save article

Abstract

In recent years, advances in deep learning technology have significantly improved the research and services relating to artificial intelligence. Real-time object recognition is an important technique in smart cities, in which low-cost network deployment and low-latency data transfer are key technologies. In this study, we focus on time- and wavelength-division multiplexed passive optical network (TWDM-PON)-based inference systems to deploy cost-efficient networks that accommodate several network cameras. A significant issue for a graphics processing unit (GPU)-based inference system via a TWDM-PON is the optimal allocation of the upstream wavelength and bandwidth to enable real-time inference. However, an increase in the batch size of the arrival data at the edge servers, thereby ensuring low-latency transmission, has not been considered previously. Therefore, this study proposes the concept of an inference system in which a large number of cameras periodically upload image data to a GPU-based server via the TWDM-PON. Moreover, we propose a cooperative wavelength and bandwidth allocation algorithm to ensure low-latency and time-synchronized data arrivals at the edge. The performance of the proposed scheme is verified through a computer simulation.

Full Article | PDF Article

More Like This

Passive Optical Network Based Mobile Backhaul Enabling Ultra-Low Latency for Communications Among Base Stations

Jun Li and Jiajia Chen
J. Opt. Commun. Netw. 9(10) 855-863 (2017)

Dependency-aware task cooperative offloading on edge servers interconnected by metro optical networks

Shan Yin, Wei Zhang, Yutong Chai, and Shanguo Huang
J. Opt. Commun. Netw. 14(5) 376-388 (2022)

Flexible low-latency metro-access converged network architecture based on optical time slice switching

Jialong Li, Nan Hua, Zhizhen Zhong, Yufang Yu, Xiaoping Zheng, and Bingkun Zhou
J. Opt. Commun. Netw. 11(12) 624-635 (2019)

Cited By

You do not have subscription access to this journal. Cited by links are available to subscribers only. You may subscribe either as an Optica member, or as an authorized user of your institution.

Contact your librarian or system administrator
or
Login to access Optica Member Subscription

Figures (13)

You do not have subscription access to this journal. Figure files are available to subscribers only. You may subscribe either as an Optica member, or as an authorized user of your institution.

Contact your librarian or system administrator
or
Login to access Optica Member Subscription

Tables (4)

You do not have subscription access to this journal. Article tables are available to subscribers only. You may subscribe either as an Optica member, or as an authorized user of your institution.

Contact your librarian or system administrator
or
Login to access Optica Member Subscription

Equations (18)

You do not have subscription access to this journal. Equations are available to subscribers only. You may subscribe either as an Optica member, or as an authorized user of your institution.

Contact your librarian or system administrator
or
Login to access Optica Member Subscription

Variable	Definition
$I$	Set of PONs connected to the edge
$i$	PON identifier, $i \in I$
$J$	Set of ONUs; $\| J \|$ is the number of ONUs per PON
$j$	ONU/camera identifier, $j \in J$
$W$	Set of wavelength channels
$w$	Wavelength channel identifier, $w \in W$
$G$	Set of available time-domain groups
$g$	Group identifier, $g \in G$
$σ$	Time interval between groups
$τ$	Cycle length
$t_{c, g}$	Arrival time of the $g$ th group
$L_{p o n}$	Uplink bandwidth per channel in the PON
$d_{i, j}^{p r p}$	Propagation delay of the $j$ th ONU in the $i$ th PON
$d_{i, j}^{s e r i a l}$	Serialization delay of the $j$ th ONU in the $i$ th PON
$d_{i}^{s n i}$	Delay between the edge and the OLT in the $i$ th PON
$d_{i, j}^{u n i}$	Delay between the $j$ th ONU and the camera in the $i$ th PON
$d_{i, j}^{t o t a l}$	Total delay of the $j$ th ONU in the $i$ th PON
$d_{i, j}^{m a x}$	Maximum delay of the $j$ th ONU in the $i$ th PON
$s_{i, j, c}$	Transmission start time of the $j$ th ONU in the $i$ th PON
$ρ_{i, j}$	Image size from the $j$ th camera/ONU in the $i$ th PON
$ω_{i, j}$	Image generation interval of the $j$ th camera in the $i$ th PON
$δ_{i, j}$	Offset time of the $j$ th camera in $i$ th PON
$t_{i, j, c}^{i m g}$	Image generation time of the $j$ th camera in the $i$ th PON
$t^{r e f}$	Reference time
$d_{i, j}^{q u e u e}$	Queuing delay
$T_{l i m i t}$	Time requirement for inference task

Variable	Definition
$ω_{i, j}^{'}$	Estimated image generation interval of the $j$ th camera in the $i$ th PON
$δ_{i, j}^{'}$	Estimated offset time of the $j$ th camera in the $i$ th PON

Parameter	Value
Number of PONs $\| I \|$	10
Number of ONUs $\| J \|$	4–64
Propagation delay	5 µs/km
Data size $ρ$	2.2 MB
Interval image $ω$	$\frac{1}{3} s$
Link speed $L_{p o n}$	10 Gps
Number of wavelength channels $\| W \|$	4
Processing time of the edge server $σ$	150 ms
Preprocessing time $γ$	50 ms
Time requirement $T_{l i m i t}$	250 ms
Maximum batch size	140.8 MB

Variable	Definition
$I$	Set of PONs connected to the edge
$i$	PON identifier, $i \in I$
$J$	Set of ONUs; $\| J \|$ is the number of ONUs per PON
$j$	ONU/camera identifier, $j \in J$
$W$	Set of wavelength channels
$w$	Wavelength channel identifier, $w \in W$
$G$	Set of available time-domain groups
$g$	Group identifier, $g \in G$
$σ$	Time interval between groups
$τ$	Cycle length
$t_{c, g}$	Arrival time of the $g$ th group
$L_{p o n}$	Uplink bandwidth per channel in the PON
$d_{i, j}^{p r p}$	Propagation delay of the $j$ th ONU in the $i$ th PON
$d_{i, j}^{s e r i a l}$	Serialization delay of the $j$ th ONU in the $i$ th PON
$d_{i}^{s n i}$	Delay between the edge and the OLT in the $i$ th PON
$d_{i, j}^{u n i}$	Delay between the $j$ th ONU and the camera in the $i$ th PON
$d_{i, j}^{t o t a l}$	Total delay of the $j$ th ONU in the $i$ th PON
$d_{i, j}^{m a x}$	Maximum delay of the $j$ th ONU in the $i$ th PON
$s_{i, j, c}$	Transmission start time of the $j$ th ONU in the $i$ th PON
$ρ_{i, j}$	Image size from the $j$ th camera/ONU in the $i$ th PON
$ω_{i, j}$	Image generation interval of the $j$ th camera in the $i$ th PON
$δ_{i, j}$	Offset time of the $j$ th camera in $i$ th PON
$t_{i, j, c}^{i m g}$	Image generation time of the $j$ th camera in the $i$ th PON
$t^{r e f}$	Reference time
$d_{i, j}^{q u e u e}$	Queuing delay
$T_{l i m i t}$	Time requirement for inference task

Variable	Definition
$ω_{i, j}^{'}$	Estimated image generation interval of the $j$ th camera in the $i$ th PON
$δ_{i, j}^{'}$	Estimated offset time of the $j$ th camera in the $i$ th PON

Abstract

Cited By

Figures (13)

Tables (4)

Equations (18)

Journal of Optical Communications and Networking