HQoS

What Is HQoS?

Hierarchical Quality of Service (HQoS), within the DiffServ model, adopts a hierarchical queue scheduling mechanism to ensure bandwidth allocation for various services, catering to the needs of users. Unlike conventional QoS, which operates with single-level traffic scheduling, HQoS implements multi-level queue-based scheduling. This enables differentiation of traffic from different users and services, resulting in refined QoS guarantees.

HQoS vs. QoS

HQoS is an extension of conventional QoS within the DiffServ model, rather than an independent QoS solution. It incorporates hierarchical scheduling on top of conventional QoS to achieve more precise traffic management.

Within the QoS DiffServ model, packets are classified into various queues based on mappings between their priorities and the device's internal priorities. The scheduler then determines the order in which packets from different queues are transmitted, utilizing scheduling algorithms like priority-based SP scheduling or weight-based DRR/WRR/WDRR/WFQ scheduling.

Conventional QoS: One-Level Scheduling Mechanism

Assuming data, voice, and video packets are allocated to separate queues based on their priorities (high, medium, low) using the SP scheduling algorithm. The packets are then transmitted in order of priority, with voice packets given the highest priority, followed by data and video packets.

Conventional QoS operates on a per-port basis, allowing differentiation of service priorities but not individual users. Consequently, multiple types of traffic with the same priority from different users are grouped into a single port queue (PQ), leading to competition for resources. This limitation results in an inability to provide differentiated services to different users.

HQoS: Hierarchical Scheduling Mechanism (Two-Level Scheduling as an Example)

Consider the following scenario: data, voice, and video services are needed by both average consumers and VIP users. HQoS introduces a two-level scheduling mechanism: user-based scheduling and service-based scheduling. User-based scheduling prioritizes the transmission of packets from VIP users, while service-based scheduling focuses on prioritizing important services within each user. This approach is similar to conventional QoS's one-level scheduling but offers enhanced capabilities.

Once HQoS is implemented on a port, traffic management can be visualized. The available port bandwidth is allocated to different users, taking into account their specific requirements. Subsequently, within each user, the service bandwidth is further allocated to different services based on their individual needs.

HQoS employs hierarchical scheduling to differentiate services and users, prioritizing the processing of services for VIP users and high-priority services. It also allocates guaranteed bandwidth to accommodate the needs of various services and users. This implementation enables more precise and refined traffic management compared to conventional QoS.

HQoS Application

HQoS finds its application when conventional QoS falls short of meeting network traffic management demands, offering differentiated QoS guarantees for key users and services, thereby enhancing the user experience. It is crucial to ensure that the network devices utilized support HQoS for effective traffic processing.

HQoS is extensively employed in various scenarios. For instance, in enterprise campus networks, it caters to the diverse service requirements of campus users. Similarly, on ISP edge devices, HQoS is utilized to address the bandwidth needs of access users with different service levels. The design and deployment of the HQoS traffic model vary according to industry and application scenarios, necessitating customization based on specific network requirements.

To illustrate the application of HQoS, let's consider a scenario involving home broadband services. A carrier introduces a residential broadband package in a housing complex, providing a total bandwidth of 100 Mbit/s. The package includes VoIP, IPTV, and high-speed Internet (HSI) services. Each family within the compound can utilize VoIP for calls, IPTV via set-top boxes (STBs) for television programs, and access the Internet through PCs, tablets, and mobile phones.

HQoS can be deployed as follows:

• Classify the three types of services and regulate traffic scheduling and bandwidth allocation among them.

• Treat each family as an individual user and allocate a dedicated bandwidth of 100 Mbit/s to each user, employing rate limits to enforce the allocation.

• In this example, households within each building can be grouped together as a user group, implementing three-level scheduling. The total bandwidth available to the user group is the cumulative bandwidth of all 16 households, which can be shared among them.

HQoS ensures efficient bandwidth allocation for different services within a household and facilitates bandwidth allocation among households. Additionally, it supports bandwidth sharing to optimize resource utilization.

HQoS Scheduling Structure

The HQoS hierarchical scheduling framework typically utilizes a tree scheduling structure. It can employ two levels of schedulers to establish a three-layer scheduling structure, such as service-user two-level scheduling. Alternatively, three levels of schedulers can be employed to establish a four-layer scheduling structure, which can include user group-based scheduling in addition to service-user two-level scheduling. Furthermore, an additional level of scheduling can be added to establish a five-layer scheduling structure. For example, if multiple user groups access the network through sub-interfaces, scheduling based on sub-interfaces for user group access can be added to the existing service-user-user group three-level scheduling. This allows for the allocation of guaranteed bandwidth to each user group and facilitates bandwidth sharing among the user groups.

It is important to note that the HQoS scheduling structure is closely tied to the capabilities of the hardware being used. Different vendors and chips may offer varying levels of support for scheduling levels, scheduling algorithms, and configuration models.