UP de-registration is triggered in the following scenarios:

• Graceful de-registration from UP—In this scenario, Control-Plane-Group association is removed with User-Plane-Service CLI. The IP addresses are released at sessmgr level on CP.

• UP connection failure from CP—This scenario occurs either because of miss of heartbeat from UP to CP, or because UP restarts and CP is communicated about it. When UP restarts, it implies that the reception of a new Restart-Counter at CP of the specified UP.

After the UP de-registration is triggered, the VPNMGR task on CP validates the identity and address of UP with the information available in the VPNMGR database. In case of mismatch, VPNMGR shows the failure message. In case of match, the validation is successful. On successful validation, VPNMGR takes all the assigned and unassigned chunks from both IPv4 and IPv6 pools from the specified UP.

Whether the UP has some used or all unused IPs, VPNMGR starts a 2-minutes timer before carrying out forceful de-registration of the UP. During forceful de-registration, all IP addresses are deleted from VPNMGR database locally, session entries are removed, and all the chunks are placed to the main address pools at CP.

Hold Timer is configured per pool for IPv4 dynamic pools. Static pools and IPv6 pools are not considered. If Hold Timer is not configured, the IP address moves from Free to Used state when allocated, and returns to Free state upon session release. When Hold Timer is configured in the pool, a released IP address is moved to the Hold state. For the Hold Timer duration, the IP address remains in the Hold state and can be reused if the same subscriber attaches again. While in Hold state, the IP address is not assigned to any other subscriber. After Hold Timer expiry, the IP address moves to Release state and is reused only when all free IP addresses in the pool are exhausted.

For dual-stack PDN connections using the IPv4 address-hold-timer, the IPv6 session affinity anchors the IPv6 component to the same UP that previously hosted the IPv4 session, overriding the standard UP selection algorithm during reassociation. This ensures successful allocation of both IPv4 and IPv6 addresses.

Note	IPv6 session affinity ensures the allocation of IPv6 address provided there are free IPv6 addresses with the userplane however, it does not guarantee allocation of same IPv6 address on reassociations.

Address hold timer on IPv4 pools is supported only for one PDN connection (single or dual stack) per subscriber. When a subscriber has multiple PDN connections, address hold timer cannot be used for more than one PDN. In other words, in case of a multi-PDN call, only one of the allocated IP addresses will get into Hold state when the call is disconnected.

When UP deregistration occurs, all IP addresses in Hold state move to Free state. This happens because the UP details (UP ID and related memory) are not preserved at the CP. This change might result in the IP address being reused by a different subscriber. VPNMgr recovery and ICSR are also supported for Hold addresses.

Address State Change

Following call flow describes the address state change with Hold Timer configured.

Following call flow describes the address state change without Hold Timer configuration.

Configuring Hold Timer

Use the following configuration to enable Hold Timer feature in CUPS.

configure 
   context context_name 
      ip pool pool_name address-hold-timer seconds 
      end 

NOTES:

• pool_name : Specifies the logical name of the IP address pool. pool_name must be an alphanumeric string of 1 through 31 characters.

• When the feature is enabled, and an active subscriber is disconnected, the IP address is held or considered still in use, and isn't returned to the free state until the address-hold-timer expires. This enables subscribers who reconnect within the length of time specified (in seconds) to obtain the same IP address from the IP pool. seconds is the time in seconds and must be an integer from 0 through 31556926.

Use the show ip pool address pool-name pool_name CLI command to check the status of all IP addresses in a pool. It also shows the remaining hold time for the held addresses.

You can configure 600 IP pools per UP group in a single context at CP. Also, 2000 IPv4 and 256 IPv6 pools can be configured per context in CP which can be distributed among various UP groups with upper limit of 600 pools per UP. The functionality includes:

• UP group can have a maximum of 600 IP pools for all possible combinations of pool type.

• Pools can be either static, dynamic, or combination of both.

• Pools can be all IPv4, IPv6, or combination of both.

• Out of 600, a UP group can have a maximum of 256 IPv6 pool (context level limitation that is same as ASR5500). All 600 pools can be IPv4.

• If more than 600 IP pools are configured in a UP group, then it can't be determined as to which 600 pool/pool chunks will be allocated to a UP.

• The CP maintains count of routes that are installed at UP. If it exceeds 6000 pool routes (context level limitation that is same as ASR5500), then no new chunk is allocated to UP even if it reaches the threshold for overuse. Similarly, if new IP pool is dynamically allocated and 6000 pool routes are already installed, then no new chunk is allocated from that pool even if pool count is less than 600 for that UP.

As part of this feature, the dynamic IPv4 and IPv6 pool count is replaced with total IPv4 and IPv6 pool count in the show ip user-plane verbose CLI command. Also, the output of the CLI command is enhanced to display Total Pool Kernel Routes and Max Pool Kernel Routes fields.

In CUPS architecture, the CP has all the IP Pool configurations in PDN/IP context. In compliance with 3GPP standards, the UP registers with CP by Sx Association Request/Response procedure.

During the registration process, the CP finds out all the APNs which are being served by the particular UP, and the associated Pool configuration in each APN. The CP allocates some of the IP chunk resources to a particular UP and sends over the Sx Association Update Request/Response procedure. This information is sent to PDN/IP context instance at UP.

After UP registration is successful, the PDN/IP instance initiates sending of IP chunk resource information to the UP from the Pool. This IP chunk resource information is sent to the UP on Proprietary/Custom IE on Sx Association Update Request/Response message. The PDN/IP instance at UP announces the BGP routes, on per chunk basis, which is received from the CP.

Each UP, which is registered with the CP, is identified using "Peer Id" and the Node ID.

For efficient utilization of IP resources, the CP allocates IP resources to UP on need basis. And so, it supports replenishment and withdrawal procedures for IP chunk resources.

Based on the threshold logic in CP, it monitors the usage of IP resources in each UP on pool-level basis. If the overall IP chunk usage of the UP crosses certain threshold, the CP sends additional IP chunk resources to the UP.

If certain IP chunks in the UP are not utilised, and idle for certain duration, the CP withdraws those IP chunk resources from respective UPs. For details, see Configuring Percentage of Chunks Per Pool section.

Feature Description

For static IP address allocation, the SessMgr requests for specific IP address. The VPNMgr searches for that specific IP address. If the chunk is already allocated to a particular UP, then the VPNMgr allocates that address and responds to the UP which serves the call. For static IPv4v6 call, the requested IPv4 and IPv6 address might belong to different UPs and therefore, success of IPv4v6 can't be guaranteed unless there's only one UP per UP Group. So, for successful static IPv4v6 call, only one UP per UP group can be configured. For one UP per UP Group use case, pool chunking isn't recommended as only one UP uses that pool, and the entire pool can be allocated to the UP rather than in chunks. Also, there are certain use cases to contain one APN to one UP. To support both these use cases, an option to not chunk the pool in CUPS architecture is introduced.

Without the no-chunk-pool functionality, if number of usable addresses are less than the chunk size, then minimum of two chunks were configured.

With no-chunk-pool functionality, a pool can be configured without being chunked. The entire pool is allocated to the UP that is first to request for the pool.

Note	The no-chunk-pool functionality is recommended only for a setup with one UP per UP Group. It's not recommended for multi-UP per UP Group.

How it Works

The no-chunk-pool functionality includes:

• When a pool is configured as no-chunk-pool, then pool itself is considered as a chunk and the entire pool is allocated to the UP that is first to request for the pool.

• No-chunk-pool can be public, private, or static.

• No-chunk-pool can be configured within VRF.

• For multi-UP per UP Group, the entire dynamic no-chunk-pool is allocated to the UP that is first to do Sx-association.

• For multi-UP per UP Group, the static no-chunk-pool is allocated in round-robin algorithm among currently servicing UP.

• For multi-UP per UP Group, the dynamically added new pool can get allocated to any UP in UP Group and can’t be deterministically known.

Configuring No-chunk-pool

Use the following configuration to enable no-chunk-pool functionality.

configure 
   context context_name 
      cups enable 
         ip pool pool_name ip_address/subnet_mask no-chunk-pool 
         ipv6 pool pool_name prefix ip_address/length no-chunk-pool 
   exit 

The no-chunk-pool can be identified from the output of the following CLI commands if the "total-chunks" field displays 1 (one) for that particular pool.

• show ip pool-chunks pool-all

• show ipv6 pool-chunks pool-all

In CUPS architecture, the strategy to manage static IP pools differs from dynamic pool managemnent. Static IP pools are broken down into "static-chunks" similar to how dynamic pools are chunked. However, these static chunks are not distributed to the UPs and remain at the CP until a UE requests for the first static address in a certain Static-IP-Chunk during session creation.

The CP selects the UP using the round-robin algorithm and the entire Static-IP-Chunk, to which the requested static address belongs, is assigned to the selected UP. Therefore, whenever any UE requests static addresses (IPv4 or IPv6) from that chunk, the UE is assigned that UP.

Note

Within dynamic pools, "allow static" is not supported. IPv4v6 static PDP is not supported with multiple UPs in a UP Group. For the static IPv4v6 PDNs to be successful, both IPv4 and IPv6 addresses must be on the same UP. Only way to ensure this is to have a single UP in the UP group. For the multi-PDNs on same APN to be successful, with one PDN as static and the other as dynamic, both addresses must be on same UP. Only way to ensure this is to have a single UP in the UP group. In case of static IP pool, address is already decided by UE and so, the benefit of UP selection does not remain.

In CUPS architecture, during the establishment of sessions, UP selection happens among the registered UP. There are various ways to select UP. In earlier releases, Round-Robin Algorithm based UP selection was supported. Currently, least connection User Plane selection algorithm is supported.

The following functionalities are supported as part of the IP Pool Management feature.

• IPv4, IPv6 Public, and private pool-based IP address allocation.

• IPv4 static type address allocation.

• Session Manager recovery and VPN Manager recovery for active calls types.

• CP to CP Interchassis Session Recovery (ICSR) support.

• Hold-timer for IPv4 pools.

• Busy-out (basic functionality) for IPv4 and IPv6 pools.

Following are the known limitations and restrictions of this feature for this release:

• The “allow-static” type pool configuration isn't supported.

• Configure the cups enabled CLI before you add a pool in IP context to enable IP Pool Management functionality in CUPS mode.

• IPv4v6 static PDP isn't supported with multiple UPs in a UP Group.

• The output of the following CLI commands displays all pools with maximum of 2048 chunks per pool:

  • show ipv6 pool-chunks up-id up_id

  • show ipv6 pool-chunks pool-name ipv6_pool_name

  • show ip pool-chunks up-id up_id

  • show ip pool-chunks pool-name ipv4_pool_name

• Upon UE reattach, CP needs to select the same UP session (as IP address is already advertised by that UP in the earlier session). Hence there is no UP load based selection or location based UP Selection possible.

• Hold timer value of 0 is not supported.

• Recovery of Hold timer is supported for up to 1000 address per session manager.

• Reload chassis results in the standby chassis losing the hold timer information.

• Any change to the Hold timer value also requires a Sx re-establishment like it happens for any other pool configuration.

Enabling IP Context for IP Pool Management

Use the following CLI commands to enable IP context for IP Pool management.

configure 
   context context_name 
      cups enable 
      end 

Configuring Custom Threshold Timer

Important

In 21.9 (mid-July) and later releases, the cups chunk-allocate-timer allocate_timer_seconds chunk-release-timer release_timer_seconds CLI command is deprecated, and replaced by cups chunk-threshold-timer threshold_timer_seconds and cups min-chunks-threshold-per-pool threshold_percent CLI commands.

There is a threshold timer for chunk redistribution among UPs. By default, for sending chunk into an over utilized UP, check is carried out every 60 seconds, and for removing chunk from an underutilized UP, check is carried out every 300 seconds. For custom threshold timer, use the following CLI commands:

configure 
   context context_name 
      cups chunk-allocate-timer allocate_timer_seconds chunk-release-timer release_timer_seconds 
      end 

NOTES:

• This is an optional configuration. If not configured, then by default the allocate threshold is 60 seconds and the release threshold is 300 seconds.

• Use the default cups chunk-allocate-timer chunk-release-timer CLI command to revert back the chunk-allocation and chunk-release timer to 60 and 300 respectively.

• If the release timer is configured to be less than the allocate timer, then it is overwritten with the value that equals to the allocate timer.

Configuring Chunk Threshold Timer

Use the following CLI commands to configure CUPS IP pool chunk threshold timer for a context.

configure 
   context context_name 
      cups chunk-threshold-timer threshold_timer_seconds 
      end 

NOTES:

• threshold_timer_seconds : Specifies the chunk threshold timer value in seconds, integer 30 to 300. Default = 60 seconds.

• Use the default cups chunk-threshold-timer CLI command to set the default value of 60 seconds.

• In releases prior to 21.9 (mid-July), allocation of new chunks to UP and release of chunks from underutilized UP use to occur based on allocation and release timers, respectively. With 21.9 (mid-July) and later releases, only single threshold timer exists, based on which the allocation and release of chunks occur periodically.

Configuring Percentage of Chunks Per Pool

Use the following CLI commands to configure minimum percentage of chunks per pool in a context.

configure 
   context context_name 
      cups min-chunks-threshold-per-pool threshold_percent 
      end 

NOTES:

• threshold_percent : Specifies the minimum chunks in percentage of 0 to 50. Default = 10.

• Use the default cups min-chunks-threshold-per-pool CLI command to set the default value of 10 percent.

• Chunks are released periodically only when free chunks with particular pools, at CP, are less than the percentage configured with this CLI command.

  • When minimum chunks equals to, or falls below, the configured percentage, a check is done to ascertain if there is any UP that has less than 50% utilization and has more than 2 free chunks. If there is, then one is taken back from each underutilized UP from that particular pool.

• Warning log is generated for: periodicity = chunk-threshold-timer; till minimum chunks in CP VPNmgr are restored.

• UP lockdown period on registration: For first five (5) minutes of a UP registration, no chunks are taken back from that UP and sent to another UP even if other UPs are in need of chunks.

For IP context in UP, there is no requirement for IP Pool configuration, or to use the cups enabled CLI command.

Use the following CLI commands to configure maximum number of User Planes expected to be functional in a system.

configure 
   context context_name 
      cups max-user-planes value 
      end 

NOTES:

• In releases prior to 21.25:

  cups max-user-planes value : The default value is 10.

  The maximum number of user-planes supported in a context and UP Group is 1000.

• In 21.25 and later releases:

  cups max-user-planes value : The value is in the range of 1-1000. The default value is 10.

  The maximum number of user-planes supported in a context is increased to 1000.

  Use this CLI command to tune the chunks that were initially allocated on Sx-association. It can't be used to restrict the addition of new UPs into a system.

• Use the default cups max-user-planes CLI command to revert back the maximum user-planes value to 10.

This section provides information regarding show commands and/or their outputs in support of this feature at CP.