5G Flow Sampling and Filtering
Required License : FlowVUE license is required for Flow Sampling
5G flow sampling samples a configured percentage of 5G sessions. 5G flow sampling uses map rules to select subscribers and then forward a percentage of the packets to tool ports.
Pass rules are defined in flow sampling maps. Each rule contains some combination of SUPI, PEI, and GPSI numbers or patterns as a percentage to sample. The flow is sampled to see if it matches a rule. The percentage of the subscriber sessions matching each rule are selected.
Map rules specify the type of traffic to be flow sampled by that map. For each new session, map rules are evaluated in top-down order of decreasing priority. If there is a match, the indicated percentage of the subscriber session is either accepted or rejected. If accepted, the traffic is sent to the tool port or load balancing group specified in the map. If rejected, the traffic is dropped. If there is not a match to a rule, the traffic is passed to subsequent maps.
Flow sampling rules are configured in maps called flow sampling maps. Up to ten (10) flow sampling maps per GigaSMART group are supported. Each flow sampling map supports up to
5G flow sampling (rule-based flow sampling) is performed after 5G whitelist-based forwarding. So, flow sampling maps have a priority lower than forward list selective forwarding - forward (formerly whitelist) maps and higher than flow filtering maps.
Note: For 5G second level maps, a maximum of fifteen maps can be attached to a vport. For example, for the same vport you can have one forward list map and ten flow sampling maps, or ten forward list map, four flow sampling maps, and one flow filtering map. In addition, you can have a collector map, which is not counted.
In the flow sampling maps, the rules in the first map have a higher priority than the rules in the second, third, and subsequent maps. Within any single map, rules are evaluated in order.
Rules can be added to, deleted from, or inserted into a flow sampling map when the subtype selected for a Second Level map is Flow Sample. Suffix wildcarding, such as SUPI 100*, is supported in the flow sampling map rules.
Use the Add a Rule button in the Maps page to add a new flow sampling rule (a pass rule). Specify SUPI, PEI, or GPSI subscriber IDs, as well as the percentage of the flow to be sampled. The percentage is a range from 1 to 100%. Use 0% to drop sampled data.
A DNN pattern can be specified in a rule by itself.
For DNN, specify a pattern (a name) to match, for example, three.co.uk. Wildcard prefixes and suffixes are supported, for example, *mobile.com or *ims*. The pattern can be specified in up to 100 caseinsensitive alphanumeric characters and can include the following special characters: period (.), hyphen (-), and wildcard (*).
Use DNN to send traffic that matches a certain DNN pattern to specified tool ports, based on the sampling percentage.
When creating Flow Sampling rules on the Maps page, the first rule created has the highest priority and the priority of subsequent rules is in the order that they are added.
Flow sampling is applied for new subscribers. When a new rule is added to the rules in a flow sampling map, traffic will be sent to the port or load balancing group specified in the map.
When a rule is deleted from a flow sampling map, the session associated with the rule stays active. The traffic associated with the rule will not be reevaluated by subsequent maps.
Priority is set as per the order defined in the policy YAML file within the type.
When a flow sampling map is deleted, the priority of the remaining flow sampling maps will be reprioritized. For example, if the first flow sampling map is deleted, the second flow sampling map will increase in priority.
For the deleted flow sampling map, the traffic associated with the rules in the map will be reevaluated and then passed to subsequent maps. When a flow sampling map is re-prioritized, the existing sessions will be reevaluated according to the new priority of the map. The traffic associated with the rules in the map will be reevaluated and then passed to subsequent maps.
When the last flow sampling map is deleted, the traffic associated with the rules in the map will also be reevaluated before being passed to subsequent maps. But the traffic associated with the rules in maps that were not matched, will not be reevaluated because that traffic was already passed to subsequent maps.
The flow-ops report displays the flow sampling rule ID for sessions that have been accepted or rejected by the flow sampling map.
However, since rule IDs are not unique across maps, when there are multiple flow sampling maps, the flow-ops report is unable to identify the exact rule that the session matched. For example, with multiple flow sampling maps, each map can have a rule ID of 1. The rule ID will be identified in the flow-ops report, but not the map associated with it.
The sampling Percentage field in a map for 5G flow sampling, represents the percentage of subscribers that will be sampled (not the sessions).
The 5G correlation engine tracks all the subscribers and all of their sessions that it sees on the network. In this example, for those subscribers with an SUPI starting with the value 46*, the 5G correlation engine keeps a list of them and randomly selects 80% of those subscribers and sets them to be in the sample, which means that a tool port (or load balanced group) will see 100% of the packets for 100% of the sessions for those randomly selected 80% of subscribers.
For the other 20% of subscribers, the 5G correlation engine continuously tracks those subscribers through the network but does not send any packets to the tool port (or load balanced group).
When a session matches one of the configured flow sampling rules, it is either accepted for sampling or rejected.
If it is accepted, all packets belonging to that 5G session are sent to the tool port or ports specified in the flow sampling maps. If a subscriber is in the sample, then both the control plane packets, and the user-data plane packets are sent to the tools.
If it is rejected, all packets belonging to the session are dropped. If the subscriber is not in the sample, then neither the control plane packets, nor the user-data plane packets are sent to the tools.
Control plane and user-data plane traffic are treated the same. For a matching session, all the control plane and user-data plane traffic will be accepted. Otherwise, all the control plane and user-data plane traffic will be rejected and dropped.