]> China Unicom

Beijing100000 China xubh15@chinaunicom.cn

China Unicom

Beijing100000 China liuc131@chinaunicom.cn

China Unicom

Beijing100000 China renj30@chinaunicom.cn

China Unicom

Beijing100000 China wency15@chinaunicom.cn

Centec

Suzhou215000 China chengw@centec.com

Centec

Suzhou215000 China wangjj@centec.com

Centec

Suzhou215000 China zhanggy@centec.com

Centec

Suzhou215000 China yangyt@centec.com

TSV TSVWG queue management AQM queue depth congestion Active Queue Management (AQM) manages queue depth by controlling packet admission at the point of enqueue. This document specifies a complementary queue management behavior that operates on packets already admitted to a queue: when the queue depth exceeds a configurable threshold, the maximum time a packet may remain queued before being dropped is reduced. When congestion subsides, the duration reverts to its base value. This per-queue dropping behavior helps manage queue depth during congestion by releasing resources from queues where packets have been waiting longest, complementing AQM without modifying its signaling semantics.

Introduction recommends the deployment of Active Queue Management (AQM) to manage queue depth and reduce latency in network devices. AQM operates at the point of enqueue: it decides whether to admit, drop, or ECN-mark an arriving packet based on current queue state. establishes that queuing and dropping are conceptually separate operations that work in series. AQM procedures such as CoDel and PIE focus on controlling queue depth by signaling congestion to transport endpoints. A related but distinct concern is managing packets that have already been admitted to a queue and have been waiting for transmission. Each queued packet has a maximum queuing duration — the upper bound on how long it may remain in the queue before being dropped. In many implementations, this duration is a fixed value that does not vary with queue state. During sustained congestion, packets continue to occupy queue resources for the full duration even when their transmission is unlikely, delaying the release of resources needed by other queues. This document specifies an adaptive queue management behavior that addresses this concern: when the queue depth exceeds a congestion threshold, the maximum queuing duration for packets in that queue is reduced, enabling faster resource release. When congestion subsides, the duration reverts to its base value. This behavior operates within the dropping framework of and complements, rather than replaces, AQM. The behavior is specified in terms of externally observable characteristics — the per-queue maximum queuing duration and the queue-depth thresholds that govern its value — rather than any particular implementation. This follows the approach used for the Differentiated Services per-hop behaviors in and , which likewise define node-local forwarding behavior by its observable effect. The behavior is independent of, and may be deployed alongside, AQM algorithms such as CoDel and PIE ; describes the considerations when both are active.

Scope This document defines a per-queue dropping behavior. It does not define new on-the-wire protocol elements, packet formats, or congestion signals. The behavior is local to the device implementing it and is transparent to endpoints. This behavior MUST NOT be used as a trigger for ECN marking . Dropping of queued packets due to duration expiry is local queue management, not congestion signaling.

Requirements Language The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "NOT RECOMMENDED", "MAY", and "OPTIONAL" in this document are to be interpreted as described in BCP 14 when, and only when, they appear in all capitals, as shown here.

Terminology This document uses terminology consistent with and .

Maximum Queuing Duration:

The upper bound on the time a packet may remain in a queue before being dropped. This bounds the per-packet queuing delay at the device.

Base Duration:

The maximum queuing duration during normal (non-congested) operation.

Reduced Duration:

The shorter maximum queuing duration applied when the queue depth exceeds the congestion threshold.

Congestion Threshold:

The queue depth at which the maximum queuing duration transitions from base to reduced.

Restoration Threshold:

The queue depth at which the maximum queuing duration reverts from reduced to base. Set lower than the congestion threshold to provide hysteresis.

Problem Statement Section 3 of discusses the importance of managing queue depth. When a device serves multiple output queues with different traffic classes , sustained congestion in one queue has broader effects:

• Packets unlikely to be transmitted continue to occupy queue resources for the full base duration, contributing to unnecessarily deep queues.
• Lower-priority queues may hold resources longer than higher-priority queues require, delaying service for priority traffic.
• Deep queues on congested ports can reduce resource availability for queues on other ports.

AQM addresses queue depth by controlling packet admission. This document addresses the complementary problem: managing packets that have already been admitted and are occupying queue resources during congestion.

Behavior

Per-Queue Parameters Each output queue SHOULD have independently configurable parameters:

• Base Duration: maximum queuing duration during normal operation.
• Reduced Duration: maximum queuing duration during congestion. MUST be less than or equal to the Base Duration.
• Congestion Threshold: queue depth triggering transition to Reduced Duration.
• Restoration Threshold: queue depth triggering reversion to Base Duration. SHOULD be lower than the Congestion Threshold.

Queue Depth Threshold Detection The device monitors the queue depth of each output queue per Section 3 of . When the queue depth exceeds the Congestion Threshold, the queue enters the congested state. When it falls below the Restoration Threshold, the queue returns to the non-congested state. The hysteresis between thresholds prevents oscillation. Implementations MAY additionally require the threshold to be exceeded for a configurable minimum interval before transitioning (dampening).

Duration Adaptation When a queue enters the congested state, the maximum queuing duration for that queue MUST be set to the Reduced Duration. A packet whose time in the queue has reached or exceeded the currently effective maximum queuing duration SHOULD be dropped rather than transmitted. Because packets are dequeued in order, this condition is evaluated at the head of the queue; an implementation MAY also evaluate it elsewhere in the queue, but is not required to. When a queue returns to the non-congested state, the maximum queuing duration MUST revert to the Base Duration. Reversion SHOULD be applied after a configurable hold-down period to avoid oscillation. A packet dropped under this behavior MUST NOT have its drop substituted by an ECN mark : unlike AQM congestion signaling, a duration-expiry drop indicates that the packet is no longer useful to deliver, and an ECN mark would not convey that.

Interaction with AQM This behavior and an AQM algorithm act at different points: AQM acts at enqueue, controlling admission and congestion signaling, while this behavior acts on already-queued packets at dequeue. Both may drop packets based on time in queue. When both are enabled on the same queue, the Base Duration SHOULD be set no smaller than the target delay of the AQM algorithm, so that the Base Duration does not pre-empt normal AQM operation under light load and takes effect only as a congestion backstop. Operators SHOULD monitor the drop counters of both mechanisms () to detect compounded loss.

Per-Class Differentiation In devices supporting Differentiated Services , parameters MAY be differentiated by traffic class. Higher-priority classes SHOULD have longer Base Durations and less aggressive Reduced Durations. The mapping is a local policy decision.

Example Per-Class Differentiation

Telemetry Implementations SHOULD expose per-queue:

• Current mode (base or reduced).
• Current queue depth.
• Packets dropped due to duration expiry, differentiated by mode.
• Number of mode transitions.
• Notifications on mode transitions including queue identifier, new mode, queue depth, and timestamp.

Data Model Framework A future document may define a YANG data model: /profiles/name +--ro state +--ro mode enumeration {base, reduced} +--ro depth uint64 +--ro drops-by-duration yang:counter64 +--ro transitions yang:counter64 ]]>

Deployment Considerations

• Begin with conservative thresholds (80-90% of queue allocation).
• Monitor per-mode drop counters during initial deployment.
• For latency-sensitive traffic classes, configure longer Reduced Durations.
• When AQM is also enabled, monitor both AQM drop counters and duration-based drop counters to avoid compounding loss.

Security Considerations The security considerations of apply. Configuration MUST be restricted to authorized administrators via standard access control (e.g., ). Telemetry does not contain user traffic but may reveal congestion patterns; access SHOULD be restricted. Implementations SHOULD log mode transitions to detect anomalous conditions.

IANA Considerations This document has no IANA actions.

References Normative References Informative References

Acknowledgements The authors acknowledge the IETF's work on queue management which established the framework this document builds upon.