The SR OS Hybrid OpenFlow switch supports statistics retrieval using the OpenFlow protocol. There are two types of statistics that can be collected:
Statistics for SR OS H-OFS logical ports
Logical port statistics are available for RSVP-TE and MPLS-TP LSP logical ports. The non-zero statistics are returned as long as an LSP has statistics enabled through an MPLS configuration.
Zero is always returned for logical port statistics for SR-TE LSPs when LSP statistics are not supported on SR-TE LSPs. The statistics can be retrieved regardless of whether an OF switch uses the specified LSP. The returned packet/bytes values are an aggregate of all packets/bytes forwarded over the LSP.
Statistics are not available for any other logical ports encodings.
Statistics for SR OS H-OFS flow table
Flow table statistics can be retrieved for one or more flow table entries of an H-OFS. The returned packet/bytes values are based on ACL statistics collected in the hardware. An OpenFlow controller can retrieve statistics either directly from hardware or from the ACL CPM-based bulk request cache. The ACL cache is used when processing an OpenFlow statistics multi-part aggregate request message (OFPMP_AGGREGATE), or when an OpenFlow statistics multi-part flow message request (OFPM_FLOW) is translated to multiple flow table entries (a bulk request). When an OpenFlow multi-part flow statistics request message (OFPM_FLOW) is translated to a single flow table entries request (a single entry request), the counters are read from hardware in real time.
A mix of the two methods can be used to retrieve some flow table statistics from hardware in real time while retrieving other statistics from the cache. See Filter policy statistics for more information about ACL cache and ACL statistics.
When the auxiliary channel is enabled, the switch sets up a dedicated auxiliary channel for statistics. See OpenFlow switch auxiliary channels for more information.
Operational Notes:
Flow table statistics displayed through the CLI debugging tools (tools>dump>open-flow>of-switch) are read in real time from hardware. However, to protect the system, executing CLI debugging tool commands within 5 s returns the same statistics for any flow that had its statistics read from hardware within the last 5 s.
When retrieving flow table statistics at scale, Nokia recommends to either use bulk requests, or to pace single entry requests to obtain the balance between stats real-time accuracy and CPM activity.