4 Network Service general description

3GPP48.016Base Station System (BSS) - Serving GPRS Support Node (SGSN) interfaceGeneral Packet Radio Service (GPRS)Network serviceRelease 17TS

4.1 Overview

The position of the Network Service within the protocol stack of the Gb interface is shown in figure 4.1.1.

NOTE: BSSGP, L1, LLC, MAC, RELAY, RLC are outside the scope of the present document, refer to 3GPP TS 43.060 for further details.

Figure 4.1.1: Position of the NS within the Gb interface protocol stack

The Network Service performs the transport of NS SDUs between the SGSN and BSS. The services provided to the NS user shall be:

– Network Service SDU transfer. The Network Service entity shall provide network service primitives allowing for transmission and reception of upper layer protocol data units between the BSS and SGSN. The NS SDUs are transferred in order by the Network Service, but under exceptional circumstances order may not be maintained.

– Network congestion indication. Congestion recovery control actions may be performed by the Sub-Network Service (e.g. Frame Relay). Congestion reporting mechanisms available in the Sub-Network Service implementation shall be used by the Network Service to report congestion.

– Status indication. Status indication shall be used to inform the NS user of the NS affecting events e.g. change in the available transmission capabilities.

The Network Service entity is composed of an entity dependent on the intermediate transmission network used on the Gb interface, the Sub-Network Service, and of a control entity independent from that network, the Network Service Control. There is a hierarchical relationship between both entities. This is reflected in figure 4.1.2. The detailed communication mechanisms between both entities is an internal matter for the Network Service and is not further standardized.

Figure 4.1.2: Internal architecture of the Network Service

The Sub-Network Service entity provides a communication service to Network Service Control peer entities. The Network Service Control peer entities use the Sub-Network Service for communication with each other. The peer-to-peer communication accross the Gb interface between remote Network Service Control entities is performed over Network Service Virtual Connections (NS-VCs). An NS-VC is a virtual communication path between Network Service Control peer entities.

The Network Service entity provides a communication service to NS user peer entities: the peer-to-peer communication between remote NS user entities is performed over BSSGP Virtual Connections (BVCs). A BVC is a virtual communication path between Network Service user peer entities. A Network Service Entity communicates with only one peer Network Service Entity.

Addressing across the Gb interface is further detailed in the rest of the present document.

The Network Service Control entity is responsible for the following functions:

– NS SDU transmission: The NS SDUs shall be transmitted on the NS-VCs. The NS SDUs are encapsulated into Network Service Control PDUs which in turn are encapsulated into Sub-Network Service PDUs.

– Load sharing: The load sharing function distributes the NS SDU traffic amongst the available (i.e. unblocked) NS-VCs of a group.

– NS-VC management: A blocking procedure is used by an NS entity to inform an NS peer entity when an NS-VC becomes unavailable for NS user traffic. An unblocking procedure is used for the reverse operation. A reset procedure is used between peer NS entities in order to set an NS-VC to a determined state, after events resulting in possibly inconsistent states of the NS-VC at both sides of the Gb interface. A test procedure is used to check that an NS-VC is operating properly between peer NS entities.

4.2 Addressing

The purpose of this sub-clause is to describe the addressing principles on the Gb interface in a generic way, i.e. irrespective of the exact configuration of the Gb interface and of the exact nature of the intermediate transmission network, when present. Therefore, this sub-clause provides an abstract description of the addressing principles. These principles are then applied to real networks in sub-clause "Sub-Network Service protocol".

In this sub-clause, addressing is considered in the general case where an SGSN is connected to several BSSs via an intermediate transmission network and in the specific case where a BSS is connected to several SGSNs within one or more pool areas, see 3GPP TS 23.236. Point-to-point physical connections may also be used, addressing in this special case can be derived from the general case.

4.2.1 Network Service Virtual Link (NS-VL)

An SGSN and a BSS may use different physical links for connecting to each other (e.g. because of intermediate equipment or transmission network). Each physical link is locally (i.e. at each side of the Gb interface) identified by means of a physical link identifier. The exact structure of the physical link identifier is implementation dependent.

Each physical link supports one or more Network Service Virtual Links (NS-VLs). Each NS-VL is supported by one physical link if the Frame Relay Sub-Network is employed. For an IP sub-network, the NS-VL is mapped to an IP endpoint. The exact nature of the NS-VL depends on the intermediate network used on the Gb interface. In the general case of an intermediate transmission network, the NS-VL is used to access the intermediate network. Communication means internal to the intermediate network are outside the scope of the present document. The NS-VLs may alternatively be used end-to-end between the BSS and SGSN, in case of a point-to-point configuration on the Gb interface.

Each NS-VL may be identified by means of a Network Service Virtual Link Identifier (NS-VLI). The significance (i.e. local or end-to-end) and the exact structure of the NS-VLI depends on the configuration of the Gb interface and on the intermediate network used. For example, in the case of a Frame Relay network, the physical link is the FR bearer channel, the NS‑VL is the local link (at UNI) of the FR permanent virtual connection (PVC) and the NS-VLI is the association of the FR DLCI and bearer channel identifier.

4.2.2 Network Service Virtual Connection (NS-VC)

In order to provide end-to-end communication between the BSS and SGSN irrespective of the exact configuration of the Gb interface, the concept of Network Service Virtual Connection (NS-VC) is used. The NS-VCs are end-to-end virtual connections between the BSS and SGSN. In the case of a FR sub-network, at each side of the Gb interface there is a one-to-one correspondence between NS-VCs and NS-VLs. When employing an IP-sub-network one NS-VL may serve several NS-VCs (see figure 4.2.2.1).

Figure 4.2.2.1: IP sub-network relationship between NS-VCs and NS-VLs

For example, in the case of a Frame Relay network, the NS-VC is the FR permanent virtual connection (PVC).

Figure 4.2.2.2 shows the relationship between NS-VCs and NS-VLs for a Frame Relay sub-network. In the case of an IP network, the NS-VC is given by a pair of IP endpoints at the BSS and SGSN. While Figure 4.2.2.1 illustrates a configuration with only one NSE, multiples NSE in either the BSS or SGSN are allowed.

At the BSS, the IP endpoints for each NSE shall not be shared among NSEs connected to the same SGSN. Howewer, an IP endpoint at the BSS may serve multiple NSEs when each of the NSEs is connected towards different SGSNs. At the SGSN, an IP endpoint may serve multiple NSEs; (i.e. IP endpoints may be shared among NSEs).

Figure 4.2.2.2: Frame relay sub-network relationship between NS-VCs and NS-VLs

Each NS-VC is identified by means of a Network Service Virtual Connection Identifier (NS-VCI) having end-to-end significance across the Gb interface. An NS-VCI uniquely identifies an NS-VC within an SGSN and within a BSS.

The establishment of an NS-VC includes the establishment of physical links, see 3GPP TS 48.014, and of NS-VLs.

In the case of an FR sub-network NS-VCs and NS-VLs are permanently established by means of administrative procedures, NS-VCIs are allocated by administrative means as well. The mapping of NS-VCIs on NS-VLIs and on physical link identifiers is held in non-volatile memory.

When employing an IP sub-network the NS-VCs and NS-VLs may be established by means of administrative means (static configuration) or by auto-configuration procedures (dynamic configuration).

4.2.3 Network Service Virtual Connection Group

The Network Service Virtual Connection Group groups together all NS-VCs providing communication between the same peer NS entities. One NS-VC group is configured between two peer NS entities. This grouping is performed by administrative means. At each side of the Gb interface, there is a one-to-one correspondence between a group of NS-VCs and an NSEI. The NSEI has an end-to-end significance across the Gb interface.

4.2.4 BSSGP Virtual Connection (BVC)

The Network Service provides communication paths between remote NS user entities. These communication paths are called BSSGP Virtual Connections (BVCs). Each BVC is used to transport NS SDUs between NS users.

A Network Service Entity provides one or more BVCs between peer NS user entities. Each BVC is supported by one group of NS-VCs. Each group of NS-VCs supports one or more BVCs. The NS entity maps between BVC and the related NS-VC group.

Each BVC is identified by means of a BSSGP Virtual Connection Identifier (BVCI) having an end-to-end significance across the Gb interface. The BVCI together with the NSEI uniquely identifies a BVC within an SGSN. The BVCI and NSEI are used on the Network Service-Service Access Point (NS-SAP) for layer-to-layer communication.

4.2.5 Use of Concepts on the Gb Interface when Intra Domain Connection of RAN Nodes to Multiple CN Nodes applies in the BSS

Figure 4.2.2.3: Use of Gb Concepts when Intra Domain Connection of RAN Nodes to Multiple CN Nodes applies in the BSS

For a pool area the BSS sets up several NSEs, and each of these NSEs goes towards different SGSNs. In this way the BSS have one NSE towards each of the connected SGSNs. Alternatively, several NSEs in the BSS are connected towards each of the SGSNs supporting the pool area.

One or more NS-VCs are set up between each of the NSEs in the BSS and the corresponding peer NSEs in the SGSNs. In an IP network, an NS-VC is identified by a pair of IP addresses and UDP ports at both the BSS and the SGSN. In a FR network, the identity of an NS-VC is unique within an NSEI.

4.3 Sub-Network Service functions

The Sub-Network Service functions of the Network Service shall provide access to the intermediate network (or to the peer entity in case of direct point-to-point configuration) by means of NS-VLs and shall provide NS-VCs between NS peer entities.

On each NS-VC, data are transferred in order by the Sub-Network Service.

When the Sub-Network Service entity detects that an NS-VC becomes unavailable (e.g. upon failure detection), or when the NS-VC becomes available again (e.g. after failure recovery), the Network Service Control entity shall be informed. Failures may occur due to protocol errors, intermediate transmission network failure, equipment or link failure or other reasons.

4.4 Load sharing function

4.4.1 Load Sharing function for the Frame Relay Sub-Network

4.4.1.1 Overview

The load sharing function distributes the NS SDU traffic among the unblocked NS-VCs of the same group on the Gb interface. The use of load sharing also provides to the upper layer seamless service upon failure by re-organizing the NS SDU traffic between the unblocked NS-VCs of the same group. The re-organization may disturb the order of transferred NS SDUs. The load sharing function should be implemented in both the BSS and the SGSN.

Load sharing applies only to NS SDUs, not to NS signalling such as NS-VC management PDUs.

4.4.1.2 Requirements on load sharing function

All NS SDUs to be transmitted over the Gb interface are passed to the load sharing function along with the Link Selector Parameter (LSP) , the BVCI and the NSEI. LSP and BVCI are used by the NS entity to select amongst the unblocked NS-VCs within the group, addressed by means of the NSEI, where to send the NS SDU. The mapping between LSP and NS-VC is based on an implementation dependent function that meets the following requirements:

– For each BVC and NSEI, the NS entity selects the NS-VC out of the group based on the LSP. This is an internal matter for the NS entity and it is not subject to further standardization.

– For each BVC and NSEI, NS SDUs with the same Link Selector Parameter shall be sent on the same NS-VC. Thus, the load sharing function guarantees that, for each BVC, the order of all NS SDUs marked with the same LSP value is preserved. In the event of a link failure and subsequent re-organization of the NS SDU traffic between the unblocked NS-VCs, the receiver may get out of order NS SDUs. Further actions implemented to prevent this error are outside the scope of the present document.

– Load sharing functions at the BSS and the SGSN are independent. Therefore, uplink and downlink NS SDUs for a subscriber may be transferred over different NS-VCs.

– A change in NS-VCs available for NS user traffic (i.e. one or more NS-VCs become blocked or unblocked) shall result in a re-organization of the NS SDU traffic amongst the unblocked NS-VCs of the same group.

– For a BVC, when there is no unblocked NS-VC of the group left between a BSS and a SGSN, the corresponding traffic is discarded by the NS at the sending side.

The Link Selector Parameter is locally used at the BSS and at the SGSN and shall not be transmitted across the Gb interface.

4.4.2 Load sharing function for the IP Sub-Network

4.4.2.1 Overview

The load sharing function distributes the NS SDUs and SNS PDUs traffic among the available IP endpoints on the Gb interface. The use of load sharing also provides to the upper layer seamless service upon failure by re-negotiating the NS SDU traffic between the remaining IP endpoints. The re-negotiation may disturb the order of transferred NS SDUs. The load sharing function shall be implemented in both the BSS and the SGSN.

The load sharing function for the IP sub-network determines the local IP endpoint locally and the remote IP endpoint based on weight information provided by the peer NSE. Each NSE shall use an implementation dependent load sharing function to determine the local IP endpoint associated with all NS SDU traffic related to an MS. The remote IP endpoint shall initially be determined by an implementation dependent load sharing function that distributes the traffic in equal proportion to the relative weights assigned to the peer NSE’s endpoints. These relative weights are assigned by the peer NSE for both signalling traffic and data traffic and will be referred to as signalling weight and data weight respectively. (These relative weights are communicated using the SNS-CONFIG, SNS-ADD, and SNS-CHANGE-WEIGHT PDUs. Also, the remote IP endpoint can be changed by the peer NSE via the Resource Distribution Function (refer to sub-clause 4.4a.).

4.4.2.2 Selection of the local IP endpoint

The LSP is used by an NS entity to select amongst the available local IP endpoints for the given NSEI, from which the NS SDUs are sent. The association between an LSP and a local IP endpoint is based on an implementation dependent function that meets the following requirements:

– For each NS-SDU, an NS entity selects a local IP endpoint based on the LSP for sending the NS-SDU to the peer NSE. If the LSP has not been associated with a local IP endpoint, then a local IP endpoint shall be selected by a method that is implementation dependent. This is an internal matter for the NS entity and it is not subject to further standardisation.

– Once a local IP endpoint is selected for the LSP, the NSE should maintain a linkage between the LSP and the local IP endpoint so that NS SDUs with the same Link Selector Parameter shall be sent from the same local IP endpoint. The NSE may disassociate the MS from the last used local IP endpoint when an MS makes a cell change.

– If a local IP endpoint that has been mapped to an LSP is taken out of service, a new local IP endpoint shall be selected and associated with the LSP.

– Selection of the local IP endpoint occurs locally at each NSE.

– If there is no available local IP endpoint, the corresponding traffic is discarded by the NS at the sending side.

The Link Selector Parameter is locally used at the BSS and at the SGSN and shall not be transmitted across the Gb interface.

4.4.2.3 Selection of the remote IP endpoint

A remote IP endpoint is an IP address and UDP port that the peer NSE has made known to the local NSE (e.g. using the Configuration procedure). In addition to being used to select the local IP endpoint, the LSP is also used by the NS entity to select amongst the available remote IP endpoints for the given NSEI, to which the NS SDUs are sent. The selection of the remote IP endpoint for an LSP is initially based on the signalling weight and data weight of the remote IP endpoints. Also, the Resource distribution function can be used by the peer NSE to initially request or subsequently change the remote IP endpoint selection for an LSP.

4.4.2.3.1 Selection of remote IP endpoint for data traffic

For each NS-SDU, an NS entity selects a remote IP endpoint based on the LSP for sending the NS-SDU to the peer NSE. If the LSP has not been associated with a remote IP endpoint, then a remote IP endpoint shall be selected for the LSP according to the data-weights assigned to the peer NSE’s IP endpoints. The remote IP endpoints shall be selected in equal proportion to the data-weights assigned to the peer NSE’s endpoints. Data-weights are assigned by the peer NSE and have a value range of 0 to 255. A data weight of 0 assigned to an IP endpoint indicates that the load sharing function shall not initially associate this remote IP endpoint to an LSP. However, if an LSP has been previously associated with a remote IP endpoint, NS SDUs associated with this LSP shall be sent to this remote IP endpoint regardless of its data weight, i.e., even when the data weight has a value of 0. (This association of LSP to the IP endpoint with a data weight of 0 may have been requested by the remote NSE via the Resource Distribution Function.)

Examples of equal proportion selection:

– If IP endpoint (A) has data weight=5 and endpoint (B) has data weight=10, then endpoint (B) will be selected for initial association with an LSP twice as often as endpoint (A).

– If IP endpoint (A) has data weight=10 and endpoint (B) has data weight=10, then endpoint (A) and endpoint (B) will be selected for initial association with an LSP on an equal basis.

– If IP endpoint (A) has a data weight=0, then IP endpoint (A) will not be selected for initial association with an LSP. However, IP endpoint (A) may be associated with an LSP via the peer NSE’s use of the Resource Distribution Function.

Once a remote IP endpoint is selected for the LSP, the NSE should maintain a linkage between the LSP and the remote IP endpoint so that NS SDUs with the same LSP shall be directed to the same remote IP endpoint. If a remote IP endpoint associated with an LSP is taken out of service, then another remote IP endpoint shall be selected according to the data-weights assigned by the peer NSE and associated with the LSP. The association of an LSP to a remote IP endpoint can be changed by the peer NSE using the Resource Distribution function (refer to sub-clause 4.4a).

The BSS may disassociate the LSP from the remote IP endpoint when an MS makes a cell change or when the MS context is deleted. The SGSN shall associate an MS with the last used remote IP endpoint as long as the SGSN has location information for the MS on cell level.

The Link Selector Parameter is locally used at the BSS and at the SGSN and shall not be transmitted across the Gb interface.

4.4.2.3.2 Selection of remote IP endpoint for signalling traffic

Outgoing BVCI=0 NS-SDUs, or SNS PDUs shall be sent to a remote IP endpoint according to the signalling weight assigned by the peer NSE. The sending NSE shall distribute these messages in equal proportion to the signalling weights assigned to the peer NSE’s IP endpoints. The sending NSE may send the BVCI=0 NS‑SDUs and all SNS PDUs from any IP endpoint.

NOTE: Load sharing endpoint selection does not apply to NS-SDUs or SNS PDUs subject to specific conditions stipulated elsewhere in the present specification, such as SNS-SIZE and SNS-CONFIG PDUs that need to be sent to an SGSN pre-configured endpoint, or any acknowledgement / response that needs to be sent back to the endpoint having originated the corresponding request.

Examples of equal proportion selection:

– If IP endpoint (A) has signalling weight=5 and IP endpoint (B) has signalling weight=10, then IP endpoint (B) will be selected as a signalling IP endpoint twice as often as IP endpoint (A).

– If IP endpoint (A) has signalling weight=10 and IP endpoint (B) has signalling weight=10, then IP endpoint (A) and IP endpoint (B) will be selected as a signalling IP endpoint on an equal basis.

– If IP endpoint (A) has a signalling weight=0, then IP endpoint (A) will not be selected as a signalling IP endpoint.

If there is no available remote IP endpoint, the corresponding traffic is discarded by the NS at the sending side.

The Link Selector Parameter is locally used at the BSS and at the SGSN and shall not be transmitted across the Gb interface.

4.4a Resource distribution function

The resource distribution function is only available for data traffic when an IP sub-network is used. This function allows the BSS or SGSN to explicitly change the IP endpoint at which it receives NS SDUs. The BSS or SGSN may choose not to initiate any request for change in IP endpoints for any MS or MBMS session and rely on the underlying load sharing function to properly distribute NS user traffic. However, if the BSS or SGSN receives a request to change the IP endpoint at which NS SDUs are received for an MS or MBMS session, then the higher layers shall be informed of this change by an indication in the Link Selector Parameter. That is, if the SGSN (BSS) receives a request to change the remote IP endpoint for an MS or MBMS session then subsequent downlink (uplink) data for the mobile or MBMS session shall be sent to that remote IP endpoint indicated by the request.

The Resource Distribution Function overrides the Load Sharing function for the selection of the remote IP endpoint. The Resource Distribution Function overrides the data weight of a remote IP endpoint (i.e. the Resource Distribution Function can request that data be sent to an IP endpoint that has a data weight=0).

4.4a.1 Requirements on resource distribution function

In an IP sub-network the BSS or SGSN may receive an NS SDU with a request to change the remote IP endpoint. The triggers for generating a request are implementation dependent and are not subject to further standardization. The behaviour on receipt of a request shall meet the following requirements:

– An NSE may change the IP endpoint at which the NS SDUs for an MS are received by setting the R-bit field to "1" (Request change flow) in the NS SDU Control Bits IE in the next NS SDU or in an NS SDU with no data.

NOTE: The BSSGP DL-UNITDATA (BSSGP UL-UNITDATA) and the BSSGP UL-MBMS-UNITDATA with an LLC-PDU Length Indicator set to 0 are given in 3GPP TS 48.018.

– An NSE shall only act on a request for change in IP endpoint if the new IP endpoint has previously been configured and is operational.

– For MBMS, the resource distribution function is only available at the BSS side. That is, it is only the BSS that can request a change of the IP endpoint to which the NS SDUs for an MBMS session are sent while MBMS data flows are only present in the downlink direction.

– Upon receiving a request for change in IP endpoint, an NSE shall notify the NS user entity of the change by an indication in the Link Selector Parameter (that is, the IP endpoint at the peer entity) and direct subsequent NS SDUs for the given MS or MBMS session to the new IP endpoint.

– The resource distribution functions are independent at the BSS and the SGSN. That is, each entity may independently choose the IP endpoint at which NS SDUs for an MS or MBMS session are received.

– An NSE shall obtain the new IP endpoint from the source IP endpoint (i.e. in the UDP/IP header) of the NS SDU in which the Request Change Flow bit was set. The new IP Endpoint is not explicitly transmitted over the Gb.

– The BSS shall associate an MS with the last used remote IP endpoint as long as the MS context exist in the BSS or until the MS makes a cell change.

– The SGSN shall associate an MS with the last used remote IP endpoint as long as the SGSN has location information for the MS on cell level.

– The SGSN shall associate an MBMS session with the last used remote IP endpoint for the duration of the MBMS session.

4.5 NS-VC management function

The NS-VC management function is responsible for the blocking / unblocking, reset and test of NS-VCs.

4.5.1 Blocking / unblocking of an NS-VC

The blocking / unblocking procedures shall not be used for an IP Sub-network.

When a condition making an NS-VC unavailable for NS user traffic is locally detected at the BSS or at the SGSN, the NS-VC shall be marked as blocked by the local NS entity and the remote NS peer entity shall be informed by means of a blocking procedure. The remote NS entity shall then mark the NS-VC as blocked and shall consider it as unavailable for NS user traffic.

A BSS or SGSN may block an NS-VC because of:

– Operation and Maintenance intervention at the Gb interface making the NS-VC unavailable for NS user traffic;

– equipment failure at a BSS or an SGSN entity;

– equipment or link failure on a BSS or an SGSN site;

– failure in the transit network; or

– other causes.

When the NS-VC becomes available again for NS user traffic, the NS entity which initiated the blocking procedure may inform the remote NS peer entity by means of an unblocking procedure. The remote NS entity shall then mark the NS-VC as unblocked and shall consider it as available for NS user traffic.

The blocking / unblocking procedures are further detailed in the rest of the present document.

4.5.2 Reset of an NS-VC

The reset procedure shall not be used for an IP Sub-network.

This procedure is used to reset one NS-VC to a determined state between remote entities. This procedure is performed:

– when a new NS-VC is set-up;

– after a processor re-start;

– after a failure recovery when the state of an NS-VC must be set to blocked and alive; or

– at any local event restoring an existing NS-VC in the dead state or in an undetermined state.

When a reset procedure is initiated, data in transfer may be lost.

4.5.3 Test of an NS-VC

The test procedure is used to check that end-to-end communication exists between peer NS entities on a given NS-VC. When end-to-end communication exists, the NS-VC is in the "alive" state, otherwise it is in the "dead" state. A dead NS-VC can not be in the "unblocked" state.