Always-on connectivity is required to provide push services, like push e-mail and push-to-talk, from the server to the terminal. The always-on capability also improves the end user performance of interactive services by avoiding setup delays. It is not practical to maintain the physical radio connection all of the time, but the network must be able to maintain a logical connection without physical resource allocation.
An always-on application can preserve its transport layer connections if it can keep the same Internet Protocol (IP)address despite its mobility. The IP address in a Universal
Mobile Telecommunications System (UMTS)netw ork is allocated as part of the Packet
Data Protocol (PDP) context during GPRS session management
Based on this background I discuss the signaling procedures and the delay in cell change from HSPA to GPRS during data transmission.
The signaling procedures are shown in the illustration below;
Once the RNC observes that HSPA coverage becomes weak, it requests the UE to make measurements of potential second-generation (2G)target cells. Once the best 2G target cell is identified, the RNC instructs the UE to move onto a specific 2G cell using the Cell
Change Order message.
Once the UE has moved onto the 2G cell, it must read the 2G system information messages including routing and location area codes and the network mode of operation. The network mode of operation defines whether or not combined location area and routing area updates are possible. The combined location and routing area update mode of operation requires a Gs interface between the Mobile Switching Centre
(MSC) and the 2G SGSN. The UE sends a Routing Area Update Request to the 2G SGSN. This triggers the Serving Radio Network Subsystem (SRNS)con text transfer procedure. The SRNS Context Request message informs theRNC to start buffering and not to send any further downlink data to the Node B. The procedure is completed by the 2G SGSN sending the SGSN Context Acknowledge message. This message indicates that the 2G SGSN is now ready to start receiving data belonging to the UE’s packet-switched connection. The 3G SGSN then sends the SRNS Data Forwarding Command to the RNC. This message instructs the RNC to start tunnelling data towards the 2G SGSN. The 2G SGSN forwards the data to the UE via a 2G radio network. The UE location is then updated so that data are sent directly from the GGSN to the 2G SGSN. The 2G SGSN sends a Routing Area Update Accept message and the UE confirms this with an associated Routing Area Update Complete message. Lets take the example below indicating an inter-system cell change recorded by a UE for a UDP-based Service.
As observed from the table above, once the UE was moved to GPRS it took 1.7 sec to read the 2G system information and initiate the location area update procedure. The location area update procedure takes about 3.5 sec. The UE then releases its channel and initiates a routing area update. The routing area update together with UE-related delays take 7 sec. The first User Datagram Protocol (UDP)packet is received on the 2G system before the routing area update procedure is complete. User plane data are sent as soon as the UE’s location is updated within the packet-switched core network. The overall inter-system handover delay is 12.3 sec in this example when measured the Layer 3 signalling and 10.7 sec when measured from the end-to-end transport layer in the case of UDP. The figure below will illustrate these delays graphically;
The inter-system handover delay presented above is for a UDP-based service, such as streaming. The delay between the last item of user plane data sent on the 3G system and the first item of user plane data sent on the 2G system can be greater for a TCP-based service due to Transmission Control Protocol (TCP) retransmission timeout (RTO).
There are a number of ways to minimize an inter-system break. Introducing the Gs interface between the MSC and the 2G SGSN allows the use of combined location area and routing area updates. Measurements have shown that this reduces inter-system handover delay by approximately 5 sec. This represents a significant percentage of the overall inter-system handover delay. The time consumed by the UE in reading 2G system information also represents a considerable part of inter-system handover delay. The delay can be minimized by broadcasting the relevant system information more often in a 2G system.
