HQoS
Traffic is scheduled based on interface bandwidth, allowing differentiation of traffic based on service levels. However, it is difficult to differentiate services based on users. Therefore, traditional QoS is applicable to the core layer, but does not apply to the access layer.
Traffic of multiple services from multiple users cannot be managed and scheduled simultaneously.
To address the preceding problems, Hierarchical Quality of Service (HQoS) is introduced, which can differentiate and schedule user traffic based on service priorities. HQoS uses multiple levels of queues to further differentiate service traffic, and provides uniform management and hierarchical scheduling for multiple transmission objects such as users and services. HQoS enables existing switches to control internal resources, providing QoS guarantee for VIP users while reducing network construction costs.
HQoS Scheduling
HQoS implements hierarchical scheduling based on queues. Currently, switches support flow queues (FQs) and subscriber queues (SQs). HQoS uses a tree structure, with FQs as leaf nodes and SQs as root nodes. Packets on an interface are first sent to leaf nodes, then scheduled, and finally sent out through root nodes. In addition, packets can be further scheduled. For example, packets can be scheduled in port queues (PQs). Switches also support the mapping between FQs and PQs to schedule the same service from different users, as shown in Figure 1-18.
When HQoS scheduling is implemented on the WAC, the WAC also supports subscriber group queues (GQs) in addition to FQs and SQs. Figure 1-19 demonstrates the HQoS scheduling process on the WAC side.
FQ
Based on the DiffServ model, an HQoS-enabled device places packets into FQs based on mapped internal priorities to differentiate services. Each user has eight FQs corresponding to eight service priorities: BE, AF1, AF2, AF3, AF4, EF, CS6, and CS7. Priority queuing or Weighted Fair Queuing (WFQ) can be configured for FQ scheduling. Each FQ supports Weighted Random Early Detection (WRED) and traffic shaping, ensuring high bandwidth and preferential scheduling of high-priority services.
SQ
SQs are used to differentiate users, and are defined differently in HQoS on the wired and wireless sides:
- SQ in HQoS on the wired side: An SQ usually refers to a VLAN or VPN, and users are classified using ACLs. Each user has an SQ, which is the aggregation of eight FQs. Traffic shaping can be configured to limit the total bandwidth of each user.
- SQ in HQoS on the wireless side (WAC): Each user with the HQoS attribute authorized by the RADIUS server is assigned an SQ when the user goes online, and other users are assigned the same SQ, that is, the default SQ of the corresponding AP queue. WDRR scheduling is used for SQs of both VIP users and common users, and PQ scheduling is used between SQs of VIP users and SQs of common users. If network congestion occurs, VIP users' service traffic is preferentially processed. This means that common users' packets with the highest priority are scheduled only after VIP users' packets with the lowest priority are scheduled.
PQ
Similar to FQs, the eight PQs correspond to eight service types. Priority queuing or Weighted Deficit Round Robin (WDRR) can be configured for the eight PQs. Each PQ supports WRED and traffic shaping. Switches support the mapping between FQs (BE, AF1, AF2, AF3, AF4, EF, CS6, and CS7) and PQs (BE, AF1, AF2, AF3, AF4, EF, CS6, and CS7). The mapping allows the switches to flexibly send service traffic in an FQ to a PQ.
AP Queue (GQ)
Each AP corresponds to a GQ. That is, traffic from all users associated with an AP is placed into the same GQ. A GQ can be bound to multiple SQs, but an SQ can be bound to only one GQ. WDRR scheduling is used between AP queues.
Target Port
Target ports are physical interfaces through which outgoing data is sent out. Traffic shaping can be performed for each target port after scheduling of FQs, SQs, and PQs is performed.