Fast Link Recovery via PTP-synchronized Nanosecond Optical Switching

A. Mehrpooya; D. Simeonidou; K. Bardhi; S. Shen; S. Yan; V. Yokar; Y. Teng; Z. Wu

REVIEW 3 major objections 6 minor 3 cited by

Fast Link Recovery via PTP-synchronized Nanosecond Optical Switching

T0 review · 3 major / 6 minor · reviewed 2026-08-11 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read The paper demonstrates that PTP time synchronization can coordinate commercial optical switches with nanosecond precision, enabling optical link recovery in 2.7 ms and scheduled recoveries in about 12.7 ms.

desk verdict A plausible optical-switching demo undercut by an unexplained 2.7 ms recovery path and an abstract/body discrepancy; the underlying jitter result is probably real but the paper needs a serious revision. read the letter →

arxiv 2412.13778 v1 pith:DPBK23WV submitted 2024-12-18 eess.SY cs.SY

V. Yokar , A. Mehrpooya , Y. Teng , S. Shen , Z. Wu , K. Bardhi , S. Yan , D. Simeonidou This is my paper · ORCID

classification eess.SYcs.SY

keywords PTPsynchronizationopticalcircuitswitchinglinkrecoverynanosecondSDNcontrollerFPGAtimedatacenternetworks

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper tries to establish that Precision Time Protocol (PTP), the packet-based Ethernet time-synchronization standard, can coordinate multiple commercial optical switches tightly enough for practical network operations, removing the need for dedicated synchronization clock paths. It reports that two off-the-shelf 1x4 optical switches, each driven by an FPGA and a GPS-disciplined OpenTimeCard, switch in about 8-10 ns with a synchronized switching-window jitter of 62 ns when the PTP traffic is given priority by a PTP-enabled Ethernet switch. The same testbed demonstrates optical link recovery: after a photodiode detects a power drop, both switches reroute to a backup link in 2.7 ms in the instantaneous-recovery path, and in about 12.7 ms when the recovery is scheduled through the SyncNet SDN controller. If these measurements hold, fast optical protection no longer requires phase-locked, dedicated synchronization hardware, which would make nanosecond-scale coordinated optical switching easier to deploy in data centers.

What carries the argument

The load-bearing mechanism is a PTP-disciplined 1PPS timeline shared by two switching nodes. Each node has a Raspberry Pi Compute Module 4 as a local agent, an OpenTimeCard that locks to GPS/GNSS time and emits a pulse-per-second signal, and an FPGA that drives the optical switch. The SDN controller (SyncNet) communicates target port and timestamp to each agent over UART, so the switches reconfigure simultaneously when their local clocks reach the scheduled time. Because the synchronization rides on ordinary Ethernet PTP rather than on dedicated phase-locked clock paths, the mechanism is what makes coordinated nanosecond switching and millisecond-scale recovery possible.

What would settle it

Instrument the instantaneous-recovery experiment so that the SDN controller logs whether it issued the reconfiguration commands. If the controller is in the data path, the end-to-end recovery time should exceed the paper's own 10 ms controller delay plus switching time; if the controller is bypassed, a sub-3 ms measurement with both switches reconfiguring confirms the claimed path.

Watch

Extended reading notes

Core claim

The central claim is that PTP-based time distribution can serve as the synchronization plane for nanosecond-scale optical switching across multiple nodes. In the authors' setup, two 1x4 optical switches are driven by FPGAs that each receive a pulse-per-second signal from an OpenTimeCard; the SyncNet SDN controller computes its time offset to each agent and sends timestamped port-reconfiguration commands, so the switches act together when their local clocks reach the designated instant. The paper validates this with a jitter of 62 ns over a half-hour measurement, compared with 105 ns when using a standard Ethernet switch, and a switching rise time around 10 ns. It then applies the mechanism to link recovery: instantaneous recovery is measured at 2.7 ms, while scheduled recovery takes about 12.7 ms, including a 10 ms SDN-controller delay. The authors present this as the first demonstration of PTP-based time synchronization integrated into network management for coordinated nanosecond optical switching.

Load-bearing premise

The 2.7 ms instantaneous recovery figure assumes that the failure-detecting FPGA can trigger both optical switches directly, without waiting for the 10 ms SDN-controller delay that the paper itself measures in the scheduled path, but the paper never states where that direct-trigger path is implemented.

Editorial extensions

If this is right

Optical link protection can operate in the low-millisecond range without dedicated synchronization wiring: 2.7 ms from power-drop detection to restored traffic.
Scheduled recovery provides a deterministic 12.7 ms maintenance window, letting operators plan link migrations or repairs at fixed times.
The same PTP plus FPGA plus SDN-agent building blocks should extend to more than two switches, since the timing relationship is controller-to-agent rather than pairwise agent-to-agent.
With sub-100 ns switching jitter, coordinated optical switches can support applications that need tightly aligned nanosecond-level windows, not just coarse circuit switching.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

If the instantaneous path genuinely bypasses the SDN controller, a natural extension is multi-hop protection: each FPGA could hold precomputed backup configurations and trigger on a shared PTP timestamp, making recovery time independent of controller round trips.
The 62 ns jitter figure invites a stress test with tighter optical switching windows or higher-order modulation, where the 8-10 ns switch edge rather than PTP jitter may become the limiting factor.
Since the lab relies on GPS/GNSS-disciplined clocks, deployment in GNSS-denied environments would need an alternative grandmaster source, such as a high-quality local oscillator distributing PTP; the paper does not address that path.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

What's actually new here is the combination: PTP time distribution, SDN control, and two commercial 1x4 optical switches, with a measured 62 ns jitter over half an hour and ~10 ns rise time. That part looks like a solid engineering result and the eye diagram plus the PTP-vs-Ethernet switch comparison is the strongest section. The recovery demo is a natural application and worth having.

The soft spot is the headline number. The 2.7 ms 'instantaneous' recovery is not supported by the text. Section 3 says the FPGA-based failure detection must notify the SDN controller, which then configures both switches. The same paragraph states that scheduled recovery costs 10 ms for the controller plus 2.7 ms for switching. If the instant path also goes through the controller, 2.7 ms is implausible unless the controller can reconfigure in microseconds; if it doesn't, the bypass mechanism (direct FPGA trigger, pre-provisioned fallback, etc.) is never described. There's no timing diagram, no component-latency breakdown, no state machine. That is load-bearing, because the abstract's 'instant network recovery within 2.7 ms' is the paper's main claim.

There is also a numbers mismatch: the abstract reports 8.4 ns switching and 100 ns jitter, while the body reports ~10 ns and 62 ns. That's the kind of sloppiness that makes a reviewer distrust the measurement chain. The work would also benefit from raw data or error bars, though the jitter measurement over half an hour is a direct observation, not a fit, so I don't doubt it exists.

The 'first integration of PTP into network management' claim rests on three references; that's a bit thin, but not fatal. Self-citation [2] is background only, so it doesn't bother me.

Who is this for? Network engineers and the optical switching community. As a proof-of-concept it's useful. The manuscript needs a major revision to explain the recovery path and reconcile the numbers. I'd send it to peer review with a clear request to address those two points, because the core demo is plausible and potentially reproducible.

Referee Report

3 major / 6 minor

Summary. The manuscript reports an experimental demonstration of PTP-synchronized optical switching using two 1×4 optical switches driven by FPGAs with OpenTimeCard time sources, an SDN controller (SyncNet), and a PTP-enabled Ethernet switch. It claims synchronization jitter of 62 ns (or 100 ns in the abstract) and switching time of 10 ns (or 8.4 ns in the abstract), and demonstrates link recovery on a testbed with a 32 Gbaud PM-16QAM signal. The paper reports 2.7 ms instantaneous recovery and 12.7 ms scheduled recovery.

Significance. If the claims are correct, this is a useful integration of PTP-based time synchronization with nanosecond-class optical switch control using commercial components, and the recovery demonstration addresses a practical network-operations problem. The paper's strengths are its direct experimental measurements, the explicit comparison of PTP-enabled and ordinary Ethernet switches, and the half-hour eye-diagram stability check. However, the contribution is primarily an experimental proof-of-concept; the inconsistencies described below currently prevent the reader from assessing the headline numbers.

major comments (3)

[Abstract vs §2.3, §4] The abstract reports an '8.4ns optical switching' and '100ns jitter at the switching edges', but §2.3 reports a rising edge (switching time) of 'around 10 ns' and a jitter of 'approximately 62 ns' for the optical switching window, and §4 repeats 62 ns and 10 ns. The 8.4 ns and 100 ns values do not appear anywhere in the body. Please reconcile these numbers and state precisely which quantity each refers to (switch rise time, fall time, PPS jitter, optical-window jitter), since the headline claims depend on them.
[§3, Fig. 3(b)] The instantaneous-recovery claim is not supported by the architecture as described. The text states that failure detection 'requires the FPGA-based failure detection to notify the SDN controller, which then configures both switches for recovery,' yet the instantaneous path is reported as 2.7 ms total. The same paragraph reports that scheduled recovery is 12.7 ms and explicitly includes 'a 10 ms for the SDN controller and an additional 2.7 ms overhead.' No mechanism is described that would let the instantaneous path bypass or overlap the SDN controller latency; a timing diagram or component-latency breakdown is needed to show how the 2.7 ms is achieved when the SDN controller is in the path. Without this, the headline 'instant network recovery' result is not interpretable.
[§3] The recovery-time measurement is not described in enough detail to be reproduced. It is unclear how the photodiode power-drop detection, FPGA signal processing, SDN notification, and switch reconfiguration are timed, what triggers the oscilloscope in Fig. 3(b), and whether the reported 2.7 ms includes the optical signal stabilization time after switching. Please provide a timing diagram and specify the exact events defining the start and end of the recovery interval.

minor comments (6)

[Fig. 2(a)] Fig. 2(a) contains the label 'GG ns' which appears to be a typo for '62 ns'.
[§2.3] The sentence 'By replacing the Ethernet switch with a PTP-enable ethernet switch, which gives priority to PTP traffic. In this case...' is a fragment and should be rewritten.
[§2.1] The acronyms 'CM4' and 'UART' are defined only implicitly; please define them at first use.
[§3] Section 3 states 'One Facebook Voyager Transponder emulates the original deployed network traffic'; please specify the data rate and whether any bit-error-rate or packet-loss measurement was made during recovery.
[§1] Reference [2] is a background citation from the same group; the introduction would benefit from a short comparison of the proposed approach with prior SDN-controlled optical TDM synchronization.
[§3] The phrase '10 ms for the SDN controller' should be '10 ms of SDN controller latency' for clarity.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity found: the paper reports direct measurements of switching jitter, switching time, and recovery latency rather than deriving predictions from fitted inputs or self-cited theorems.

full rationale

This is an experimental measurement paper, not a derivation chain. The central quantities, namely 62 ns switching-window jitter over a half-hour period, 10 ns switching rise time, 2.7 ms instantaneous recovery, and 12.7 ms scheduled recovery, are presented as oscilloscope and link-recovery measurements, not as outputs of a model whose inputs include those same quantities. The only self-citation, reference [2], is background motivation about synchronization challenges in SDN-controlled optical switching and is not load-bearing for any reported result. No parameter is fitted to a subset of data and then presented as a prediction of a closely related quantity, and no uniqueness theorem or prior-work ansatz is imported to force the choice of PTP-based synchronization. The observed inconsistencies, such as the abstract reporting 8.4 ns switching and 100 ns jitter while the body reports 10 ns and 62 ns, and the unexplained gap between the instantaneous recovery path and the 10 ms SDN controller delay attributed to scheduled recovery, are correctness, interpretability, and reproducibility concerns rather than circularity. Under the required evidentiary standard of exhibiting a specific reduction of a claimed result to its own inputs, no circular step can be identified.

Assumptions & free parameters 0 free parameters · 4 assumptions · 1 invented entities

The paper is an experimental demonstration, so it introduces no fitted parameters. It relies on several domain assumptions about the accuracy of the time source, the PTP-enabled switch's jitter performance, the optical switch's switching time, and the measurement setup. The only new entity is the SyncNet controller, which is a prototype without independent evidence.

assumptions (4)

domain assumption The OpenTime Card's 1PPS output is a faithful, accurate time reference derived from GPS/GNSS.
Section 2.1 relies on the commercial OpenTime card to generate the 1PPS that drives the FPGA switching timing; no independent verification of its accuracy is provided in this paper.
domain assumption The PTP-enabled Ethernet switch's priority queuing reduces synchronization jitter to the reported 56-62 ns range.
Section 2.3 reports a single half-hour measurement; the result is an observed behavior of a specific switch, not a general guarantee.
domain assumption The optical switch's switching time is about 10 ns as specified by the vendor, and the FPGA can drive it within that time.
Section 2.3 assumes the vendor-specified 10 ns rise time; the actual 8.4 ns versus 10 ns discrepancy in the paper is unexplained.
domain assumption The oscilloscope trigger from the master switch provides an accurate reference for measuring the slave switch's jitter.
Section 2.3 uses a master-triggered oscilloscope to measure the slave window; trigger jitter and measurement noise are not separated from synchronization jitter.

invented entities (1)

SyncNet SDN Controller
purpose: Software controller that integrates PTP timing into SDN to coordinate multiple optical switches for scheduled or instant reconfiguration.
Described in Section 2.2 and used in the demo; no external validation or published specification, only this paper.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Fast Link Recovery via PTP-synchronized Nanosecond Optical Switching." pith.science (2026). https://pith.science/paper/DPBK23WV

@misc{pith2026241213778,
  author       = {Pith},
  title        = {Pith review of: Fast Link Recovery via PTP-synchronized Nanosecond Optical Switching},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/DPBK23WV}},
  note         = {Machine review of arXiv:2412.13778}
}

read the original abstract

This paper proposes and validates a PTP-synchronized 8.4ns optical switching with a 100ns jitter at the switching edges. This approach is adopted and demonstrated for instant network recovery within 2.7ms and scheduled network recovery.

Figures

Figures reproduced from arXiv: 2412.13778 by the authors.

**Figure 2.** Experimental results for (a) PTP-based synchronization through both Ethernet and PTP enabled switches (b) Jitter measurements over 1 half-hour period 2.2. SyncNet SDN controller To manage multiple devices, we propose SyncNet to integrate PTP-based time synchronisation into the SDN controller. Asshown in [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗

**Figure 3.** (a) Experimental setup of the network recovery system, (b) Link recovery time measurements. loss is detected using a photodiode, followed by signal processing in the FPGA. When the original link fails due to significant power loss, the optical signal must be switched to a backup link to restore the connection. To achieve recovery, both fast optical switches need to be configured simultaneously. The process requires… view at source ↗

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 3 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

A Novel Frame Identification and Synchronization Technique for Smartphone Visible Light Communication Systems Based on Convolutional Neural Networks
cs.CV 2025-06 reject novelty 4.0 of 10

A lightweight CNN trained on a custom augmented dataset classifies payload, data, and overhead frames in a smartphone screen-to-camera VLC link with about 98.74% accuracy, claimed to improve synchronization.
Channel characterization in screen-to-camera based optical camera communication
eess.IV 2025-06 conditional novelty 3.0 of 10

A two-smartphone optical link at 20 cm was demonstrated; the screen's angular emission was fitted to a Lambertian source with order m=1, and success rate decays with distance.
An Image Processing Based Blur Reduction Technique in Smartphone-to-Smartphone Visible Light Communication System
eess.IV 2025-06 reject novelty 3.0 of 10

A simple image processing pipeline (grayscale, contrast enhancement, binarization) is shown to reduce blur-related errors in smartphone-to-smartphone visible light communication.

Reference graph

Works this paper leans on

4 extracted references · 3 canonical work pages · cited by 3 Pith papers

[2]

Synchronization Algorithm for SDN-controlled All-Optical TDM Switching in a Random Length Ring Network,

V. Kamchevska, V. Cristofori, F. Da Ros, B. Guo, C. Jackson, A. M. Fagertun, S. Ruepp, R. Nejabati, D. Simeonidou, L. Dittmann et al., “Synchronization Algorithm for SDN-controlled All-Optical TDM Switching in a Random Length Ring Network,” in Optical Fiber Communication Conference. Optical Society of America, pp. Th3I–2

work page
[1]

P.Jouppi

N. P.Jouppi. TPU v4: An Optically Reconfigurable Supercomputer for Machine Learning with Hardware Support for Embeddings. [Online]. Available: http://arxiv.org/abs/2304.01433

arXiv
[3]

[Online]

ADVA delivers unprecedented precision in timing networks with optical tim- ing channel solution. [Online]. Available: https://www.oscilloquartz.com/en/newsroom/press-releases/ 20200929-adva-delivers-unprecedented-precision-in-timing-networks-with-otc-solution

work page
[4]

Timing solutions for the modern world,

“Timing solutions for the modern world,” https://www.timebeat.app/. (a) Time(ms) 12.7 ms Scheduled Instantaneous Switching Time(ms) -4 -2 (b) Link Recovery Time Measurements Instantaneous Switching Amplitude(V) Amplitude(V)

work page

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.

[1] [2]

Synchronization Algorithm for SDN-controlled All-Optical TDM Switching in a Random Length Ring Network,

V. Kamchevska, V. Cristofori, F. Da Ros, B. Guo, C. Jackson, A. M. Fagertun, S. Ruepp, R. Nejabati, D. Simeonidou, L. Dittmann et al., “Synchronization Algorithm for SDN-controlled All-Optical TDM Switching in a Random Length Ring Network,” in Optical Fiber Communication Conference. Optical Society of America, pp. Th3I–2

work page

[2] [1]

P.Jouppi

N. P.Jouppi. TPU v4: An Optically Reconfigurable Supercomputer for Machine Learning with Hardware Support for Embeddings. [Online]. Available: http://arxiv.org/abs/2304.01433

arXiv

[3] [3]

[Online]

ADVA delivers unprecedented precision in timing networks with optical tim- ing channel solution. [Online]. Available: https://www.oscilloquartz.com/en/newsroom/press-releases/ 20200929-adva-delivers-unprecedented-precision-in-timing-networks-with-otc-solution

work page

[4] [4]

Timing solutions for the modern world,

“Timing solutions for the modern world,” https://www.timebeat.app/. (a) Time(ms) 12.7 ms Scheduled Instantaneous Switching Time(ms) -4 -2 (b) Link Recovery Time Measurements Instantaneous Switching Amplitude(V) Amplitude(V)

work page