Priority QoS

Priority QoS

Definition

Priority QoS schedules, queues, and allocates resources according to service priority, with higher priority services receiving bandwidth and time slots first. LTE/5G implements this through priority levels of QCI/5QI (1–9), ARP (Allocation and Retention Priority), and pre-emption priority. During congestion, lower-priority services may be downgraded, discarded, or pre-empted to ensure critical services continue to receive service.

Priority Levels and ARP

Typical priority from high to low: signaling (QCI 9), voice (QCI 1), video call (QCI 2), real-time gaming (QCI 3), real-time data (QCI 4), video streaming (QCI 5–6), regular data (QCI 7–8), Internet of Things (QCI 9, etc.). ARP includes priority level, pre-emption capability, and pre-emption vulnerability, used for access control and resource pre-emption decisions.

Relationship with Related Mechanisms

Priority QoS is reflected in traffic scheduling as scheduling weight; in packet drop strategy as WRED thresholds; and in differentiated services as DSCP priority. IoT services are usually configured with lower priority, yielding to high-value services like voice and video during congestion, but critical commands can request high-priority bearer service.

SimtoLink's IoT SIM cards can configure different priority QoS based on the type of service when establishing bearers. Bulk reporting for meters, sensors, etc., uses low priority; scenarios sensitive to latency, such as in-vehicle and industrial control, can apply for high priority. The SimtoLink platform, during critical operations such as OTA upgrades and remote command issuance, combines priority QoS to ensure execution success rates in congested networks.

ARP Decides Whether a Bearer Can Preempt, and Whether It Can Be Preempted

Priority QoS is not only who stands further forward in a queue. It also includes, when resources run out, who may preempt an existing bearer and who must give way. A QCI or 5QI states the service class: the priority number, whether the bit rate is guaranteed, the delay budget, and the packet-loss target. A smaller number usually means a higher scheduling weight. ARP has three further fields: priority, whether it may preempt, and whether others may preempt it. A bearer can have a high scheduling priority and still fail to be established at admission if it has no preemption capability. In the usual QCI table, IMS signaling is 5, the highest band; conversational voice is often 1; default best-effort data is often 9, the lowest priority. Reading 9 as “the highest” reverses the whole configuration.

These identifiers act on the scheduling and admission the operator reserves for this bearer. If a switch behind the enterprise network does not recognize the matching mark, the priority disappears once the traffic leaves the cellular network. Priority also cannot turn a narrowband link into a wideband one. It only decides who loses first when there is not enough to go around.

High Priority Has to Exist in the Subscription First

If a device or a platform temporarily marks a socket as “urgent,” and the APN subscription has no matching QCI and ARP, the core will not establish that bearer. The traffic still rides the default bearer. The condition for use is that the operator has already configured the target QCI for this APN, that the firewall and the bandwidth management devices keep the same DSCP, and that the application really separates critical packets and bulk packets onto different bearers or different marks. Configure only one side, and the other side still forwards as default.

A guaranteed-bit-rate bearer reserves resources. It suits a continuous voice or control flow. An occasional small command can often use a non-guaranteed bit rate plus a higher priority. Reserving too much crowds everyone else out. Bulk IoT reporting should sit at a lower priority, and alarms and a remote emergency stop at a higher one. If both share the default bearer, the priority field has nothing to distinguish. Extended 5QI classes also need support on both the device and the network. Configure them only in the core, with nothing in the device capability, and session setup is rejected or falls back to the default.

Reversing the Numbering Direction and the Preemption Switches

Besides treating QCI 9 as the highest, it is common to turn ARP “may preempt” and “may be preempted” both on or both off. Both on, and services of the same class preempt each other, so bearers are rebuilt over and over. Both off, and a new high-priority service cannot get in, while an old low-priority service holds the resource and will not release it. Another mistake is to write the QCI only in a module AT command. A capture still shows the outer IP on the default DSCP, which means the core did not rewrite by bearer. The enterprise queue policy cannot see a difference at all.

Acceptance cannot look only at delay in an idle period. Priority shows up under congestion. Under an artificial load or a real peak, compare loss and delay for alarm packets and for bulk reports, and check that the QCI and ARP in the bearer-setup signaling match the request. If the signaling already says 9, no amount of “urgent” marking at the application changes the air-interface scheduling order.

When the preemption switches are set so that services of the same class preempt each other, bearers are released and rebuilt over and over, and the signaling delay is worse than queuing. When both sides forbid preemption, a later high-priority session is rejected at admission and the old low-priority session keeps the resource. Compare setup success and queuing delay for the two classes during congestion. In an idle period they look equally fast. If an exit device clears the differentiated-services mark, the enterprise queue cannot see the cellular priority, and alarms and upgrades still jam on the same link. Take the mark once from the air-interface scheduling record and once from the enterprise exit. If the two do not match, fix the mapping first. Equal speed in an idle period does not prove that priority is in effect. Compare queuing delay again under congestion. If the bearer already fell onto the default class in the setup signaling, later marks will not raise its scheduling order.


Related Links

Return to the communications knowledge base