Help us improve your experience.

Let us know what you think.

Do you have time for a two-minute survey?

 
 

Hierarchical Class of Service for Network Slicing

Understanding Network Slicing

Network slicing is the partitioning of a physical network into multiple logical networks. Each logical network is called a slice. On virtue of being a logical network, a slice is a designated set of network resources, such as interfaces, firewall filters, policers, virtual output queues, schedulers, shapers, traffic control profiles etc. to carry traffic.

Slice Domain

A set of connected physical nodes such as routers and switches (along with their links) that participate in network slicing is called a slice domain. The slice domain has ingress nodes, transit nodes, and egress nodes. Ingress and egress nodes are located at the borders of the slice domain. The ingress nodes receive traffic into the domain and may classify them before forwarding them to the transit nodes. The egress nodes forward traffic out of the slice domain, and before doing so, may classify the packets.

Figure 1: Slice DomainSlice Domain

Slice Selector

By definition, a slice selector is information in the packet header. The information is used by the boundary nodes and/or transit nodes of a slice domain to classify and/or process packets. There are various options to encapsulate/identify/designate a slice selector in the packet. As an example, a Service Label in the packet header can be used as a slice selector. If defined, this label is allotted a position in the packet header and is checked at this position by firewall filters to determine/designate slices. Similarly, there are several other options as depicted in the following figure.

Figure 2: Slice Selector TypesSlice Selector Types

Workflow for Creating Slices

Slices as entities are created by specifying them under network slicing hierarchy under services. Then these slices are used to steer packets and to manage traffic destined to slices.

Hierarchical class of service for slices

You define a traffic control profile for a slice under a physical or aggregated Ethernet interface. Note that you can define traffic control profiles for multiple slices under a physical or aggregated Ethernet interface. See slice (CoS Interfaces).

See Hierarchical Class of Service (CoS) Queuing for Slice Per-hop-behavior to understand how slices (as part of a hierarchy) are used to control traffic.

Packet steering

Packet steering is the process of marking/matching packets to/from slices. Packets can be steered using firewall filters (firewall steering) and/or routing policy (route steering).

Firewall steering

  • A firewall filter can be used at the ingress node to mark matched packets as belonging to slices using the “slice” action. See slice (firewall filter action).

  • A firewall filter can also be used at the transit node to match slice packets using the “slice” match condition. See slice (firewall filter match condition). The packet can then be marked to another slice if required by the firewall filter or a policer applied to this packet etc.

  • Packets that are not marked/matched to to/from slices are processed as non-slice traffic.

Route steering

  • An export policy can be used at the ingress and/or transit node to mark matched routes as belonging to slices. See slice (export routing policy action).

  • The export policy can also attach a firewall filter to the route. The firewall filter is used to typically apply a policer to the packets matching the route on the ingress side. This firewall filter is not attached to any interface. Rather, it is part of the forwarding information of the route. See filter (export routing policy action).

  • The slice and/or firewall filter will be part of the route’s next-hop forwarding information. See show route extensive expanded-nh to view slices and/or firewall filters attached to routes.

  • Packets that do not match routes with slice information, are classified as non-slice traffic. Packets that match routes with no slice information, are also classified as non-slice traffic.

To summarize, the following are the configurations that are to be enabled before creating slices, can be used to create slices, or manage packets belonging to slices.

Hierarchical Class of Service (CoS) Queuing for Slice Per-hop-behavior

In hierarchical CoS, packets are classified at various levels. It could be at the port level, followed by the logical unit level, and then at the queue level. This means that packets are passing through a hierarchy. At every stage of the hierarchy, packets are being classified, shaped, scheduled etc.

In the context of network slicing, a slice also becomes part of the hierarchy. Shapers, schedulers, and traffic control profiles can be applied to the slice. Just as queues are made available to logical units, queues are made available to slices as well.

As the following figure shows, the queues (labeled BA1, BA2, BA3, BA4) are made available to the slice. The queues map to forwarding classes (FCs). Based on behavioral aggregate classifiers, packets are classified into FCs, and subsequently into a corresponding queue (BA1 or BA2 or BA3 etc.).

Figure 3: Hierarchical QueuingHierarchical Queuing

As Workflow for Creating Slices describes, you configure network slicing using a combination of firewall filters and Class of Service (CoS) configuration. See slice (CoS Interfaces) to read on how to enable slice(s) under any interface using CoS configuration.