How It Works

Chunk size defining and route aggregation is done by allocating continuous IP chunks to a UPF and publishing a single or aggregated route records as per the subnet of the continuous chunk. This functionality works in the following process:

  • Smaller chunk sizes, such as 4K, in consecutive fashion are defined.

  • During allocation, IPAM allocates a complete group to a UPF and single route record, as per chunk group range, is sent to UPF.

  • The chunk groups have 2, 4, or 8 chunks within it. This way, one group consists of 8K, 16K, or 32K size, and accordingly the route record of that size is sent to the UPF.

  • Node manager instances work on the smaller IP chunks only.

  • Upon an initial UPF registration, chunks for the remote node manager are also allocated from the same group.

  • When the chunks start getting freed up, they return to the free pool only when all the chunks within the group are freed.

  • To further optimize the routes sent to the UPF, reserve-contiguous-groups option can also be enabled. With this SMF upfront reserves multiple continuous chunks-groups (as per the max-session-size configured for the UPF/vDNN) for the UPF and push only one route record toward the UPF. The option reserve-contiguous-groups configuration must be enabled along with the complete address pool configuration. When enabled, the option to disable it, is not supported. If the user wishes to change the pool behavior, they must clear the subscribers associated with the pool, delete the address pool configuration, and re-configure it without the option.

  • The cache-split and dp-split values must be the same when the chunk-groups are to be enabled for the address-pool. In case the dp-split and cache-split values are different then the node manager pulls more chunks.