Ye Ouyang and M. Hosein Fallah*

Int. J. Mobile Network Design and Innovation, Vol. 3, No. 2, 2009 93 Evolving core networks from GSM to UMTS R4 version Ye Ouyang and M. Hosein Fallah* Howe School of Technology Management, Stevens Institute of Technology, Hoboken, NJ 07030, USA E-mail: youyang@stevens.edu E-mail: hfallah@stevens.edu *Corresponding author Abstract: More than 217 UMTS licenses have been issued by June 2007. Mobile operators, especially those with GSM legacy networks, prefer UMTS R4 technique to evolve their existing 2G GSM networks. UMTS R4 technique provides a smooth path to bridge legacy TDM-based network to an IP-based soft-switched network. This paper describes the basic architecture and topology of UMTS R4 core network and introduces two options in network planning: flat structure or layered structure. To propose an evolution path, the paper then suggests a threelayer structure solution to seamlessly converge UMTS R4 core network with legacy GSM core network. The proposed solution approach achieves the all-ip vision and is capable of convergence with IMS and EPC. Keywords: GSM; universal mobile telecommunications system; UMTS; soft-switch; SS; core network; CN; circuit switch; media gateway; MGW; MSC server; MSCS; IP multimedia subsystem; IMS; evolved packet core; EPC; mobile network design. Reference to this paper should be made as follows: Ouyang, Y. and Fallah, M.H. (2009) Evolving core networks from GSM to UMTS R4 version, Int. J. Mobile Network Design and Innovation, Vol. 3, No. 2, pp.93 102. Biographical notes: Ye Ouyang is a PhD student in the Telecommunications Management Program at Stevens Institute of Technology. His research interest is in communications network technologies and services, focused on communications network planning, convergence, evolution and techno-economic analysis. He has extensive experience in planning, design and implementation of 2-4G networks. He worked in Starent Networks and ZTE Corporation, dimensioning the first nationwide GSM core network for Ethiopia and UMTS core network for Pakistan and Lybia. He holds an MS in System Engineering Management from Tufts University and an ME and a BE in Control Engineering and Information Engineering from Southeast University. M. Hosein Fallah is an Associate Professor of Technology Management at Stevens Institute of Technology in New Jersey. His research interest is in the area of innovation management with a focus on the telecommunications industry. Prior to joining Stevens, he was the Director of Network Planning and Systems Engineering at Bell Laboratories. He has over 30 years of experience in the areas of systems engineering, product/service realisation, software engineering, project management and R&D effectiveness. He holds a BS in Engineering from AIT and MS and PhD in Applied Science from the University of Delaware. 1 Introduction Over the past 20 years, the way people communicated, stayed informed and entertained has changed dramatically. The technical changes in mobile networks are always revolutionary, generation by generation, and the deployment of universal mobile telecommunications system (UMTS) is no exception. The transition from second-generation (2G) to 3G took several years and we expect the same for the transition from 3G to 4G. The requirements of smooth transition drive mobile operators to look for strategies and solutions that will enhance their existing GSM networks, while addressing their 3G deployment requirements, and will not be a forklift upgrade from legacy facilities. Radio access domain is a primary concern of the UMTS deployment strategy, as it is closely coupled with the mobile operators most valued asset: spectrum. However, equally important, the core network (CN) is also playing an essential role in enhancing mobility, service control, efficient use of mobile network resources and a seamless evolution from 2G to 3G/4G. Therefore, the network evolution calls for a migration to a soft-switch (SS) CN with a flat, all-ip and simplified architecture and open interfaces which interwork with non 3rd Generation Partnership Project (3GPP) mobile networks. Copyright 2009 Inderscience Enterprises Ltd.

94 Y. Ouyang and M.H. Fallah Mobile operators are looking for a best network structure to maximise quality of service for users and to minimise the impact on legacy networks. Therefore, a new challenge for the UMTS operators is: how to most efficiently and smoothly evolve their legacy CN to UMTS and IP multimedia subsystem (IMS)? With this question, several considerations need to be addressed: Simplified topology A simplified and flattened CN with a possible reduction of network entities (NEs) involved in service processing and data transport enhances the performance of UMTS. Evolved backhaul With the deployment of UMTS, the transport backhaul becomes a key consideration that many resources are achieving after the fact. It is critical to deploy a CN solution that is flexible enough to offer smooth migration from centralised (longer backhaul) to distributed (shorter backhaul) CN nodes. Enhanced performance Obviously, the intent of UMTS is to improve the performance and efficiency of the legacy GSM network. In order to realise the full potential of UMTS, it will be important to deploy an appropriate CN structure that can meet the demands generated by increased mobile services and a growing subscriber base, including increasing network capacity requirements, thousands of call volumes per second and significant throughput. Smooth migration When mobile operators upgrade their networks to UMTS or further to IMS, they need to ensure compatibility of the new network with legacy facilities. This requires the UMTS CN structure to avoid a forklift upgrade and address 2G/3G network requirements, while at the same time, being used for evolution to IMS or evolved packet core (EPC) network. Furthermore, the term SS initially came from the definition of next generation networks (NGN) with a pure IP vision. NGN has three separations and one common characteristics: separation of call control and media bearer via H.248 and signalling transport (SIGTRAN) separation of all features from call control via session initiation protocol (SIP) separation of subscriber database [home location register (HLR) or HSS] from service logic via diameter common service logic for all service mechanisms with subscriber portability. The first separation of call control and media bearer via H.248 and SIGTRAN, applied in UMTS Release 4 (R4), can be achieved by dividing the mobile switching centre (MSC) into mobile switching centre server (MSCS or MSS) and media gateway (MGW). This is also the physical embodiment of SS in UMTS R4. The other three separations and the common service logic will be achieved in UMTS R5 network with the introduction of IMS. In UMTS R4, the separation of control from bearer achieves the all-ip vision of NGN and moves the time division multiplexing (TDM) portion into the edge of the network. That is why, from technical aspect, most mobile operators who are operating GSM networks select UMTS R4, but not R99 as the target network to evolve their legacy facilities. Hence, it is very important to study the convergence of UMTS R4 CN with legacy GSM CN. To answer the question posed earlier and based on four considerations for network evolution and three characters of NGN, we propose a network architecture of circuit switched (CS) domain for mobile operators for evolving their legacy networks to UMTS R4 CN. The proposed architecture of UMTS CN consists of three possible layers: local network, tandem network and gateway network. Section 2 will give an overview of UMTS CN; Section 3 summarises the current network structure; Sections 4, 5 and 6 describe our proposed UMTS CN structure layer by layer; and finally, Section 7 presents a summary and conclusions. 2 UMTS CN architecture As discussed by Britvic (2004) and Vrabel et al. (2007), the UMTS network consists of three primary portions: CN, radio access network (RAN) and user equipment (UE). The RAN provides all of the functions related to the radio network. CN is the heart of the mobile communication networks. It processes all the voice and data services in the UMTS core system and also implements the switching and routing functions with external networks. CN provides capabilities to achieve the essential network functions such as: mobility management, call and session control, and billing and security. Logically, the CN can be further classified into CS domain and packet switched (PS) domain. Shalak et al. (2004), Mishra (2003), Konstantinopoulou et al. (2000), Harmatos (2002) and Hoikkanen (2007) proposed several solutions to plan GSM and UMTS core networks. The solution proposed in this paper is focused on CS domain only to evolve the legacy CNs since the PS domain mainly stays the same in the network topology and the composition of NEs. From 3GPP TS 23.002, 3GPP TS 25.401, 3GPP TS 25.415 and Neruda and Bestak (2008), 3GPP has defined many versions for UMTS standard, from the include R99 to R4, R5, R6, R7 and R8. R99 is the first version of UMTS. CN in R99 is also composed of two domains: CS domain and PS domain, both of which remain the same as in GSM network in topology and NEs. From GPP TS 29.415, 3GPP TS 25.413 and 3GPP TS 29.414, there are changes in RAN: node-bs and radio network controllers (RNCs) are introduced to replace or co-exist with base stations (BTSs) and base station controllers (BSCs).

Evolving core networks from GSM to UMTS R4 version 95 Figure 1 UMTS R4 CN architecture (see online version for colours) R4, the second version of UMTS technique, introduces SS technology into CS domain in which bearer function is separated from control function. The MSC in GSM/UMTS R99 version is split into two NEs in UMTS 4 version: MSCS and MGW. As a result, the logical separation of traffic bearer and call control is achieved based on this physical division. It also evolves the CS domain to an all-ip structure and enables voice and SIGTRAN to be separated. Compared to a single TDM bearer mode in R99, CS domain in R4 supports various bearer modes: IP, ATM and TDM. From Figure 1, UMTS R4 CS domain consists of three NEs: the MSCS (or MSS) or gateway mobile switching centre server (GMSCS); MGW or gateway media gateway (GMGW) and visitor location register (VLR) which is physically integrated in MSCS. The HLR is a common entity that both CS domain and PS domain can access. UMTS R4 PS domain includes such NEs as serving GPRS support node (SGSN) and gateway GPRS support node (GGSN). Below is a short description of the NEs existed in CN domain of UMTS R4 CN. The core of the CN in UMTS R4, MSCS is a functional entity that implements mobile call service, mobility management, handover and other supplementary services. Due to the philosophy of separation of control function from bearer function in UMTS CN, it is actually the MGW that establishes call routes between mobile stations (MSs) via interface Mc. The MSCS also serves as an interface between UMTS and circuit switching networks such as public switched telephone network (PSTN) and integrated services digital network (ISDN). Furthermore, it also manages SS7, auxiliary radio resources and mobility management between RNS and CN. In addition, to establish call routes to MSs, each MSCS needs to function as a GMSCS. An MGW in UMTS R4 implements bearer processing functions between different networks. It implements UMTS voice communication, multimedia service, CS domain data service and interworking between PSTN and UMTS and between 3G and 2G networks. It also supports GSM and UMTS radio networks as well as all existing interfaces with legacy network elements. 3 Current network structure The flat (full meshed) structure is mostly selected in building the legacy 2G CNs by mobile operators whose network size is not large. However, the flat structure will no longer fit in the new environment with growing traffic volume and more new NEs. So, there are two options in planning the UMTS R4 networks: flat structure or layered structure. Figure 2 Flat structure of switching network (see online version for colours) The flat structure enables each NE to connect with every other NE in the network either physically, for example, through leased lines, or logically, for example, through T1 or E1. The flat structure (full meshed structure) is similar to the point to point structure in the internet, bypassing the tandem routes and communicating directly between two NEs. The flat structure possesses a feature of high redundancy and simple connectivity. But its network

96 Y. Ouyang and M.H. Fallah topology will become more and more complicated when the network keeps expanding. For example: if 51 2G visiting MSCs are distributed in the legacy GSM network, there are at least 51 * 50 = 2,550 routes to be configured to deliver the signalling message or carry the TDM-based traffic for the non-local calls. With a flat structure, the number of connection for the least route is given by: Least route number = N ( N 1) (1) where N is the number of network elements in the network. The layered structure does not require direct links between all the network elements, but provides some tandem elements (class 2 4 switches) in the tandem layer to connect all the local exchanges in the local access layer. In this scheme, the traffic or signalling routing between the NEs takes place either directly if they are connected or indirectly through the tandem NEs (3GPP TS 25.415). The layered structure simplifies the network topology and reduces the link resources. In a UMTS R4 CN, tandem mobile switching centre server (TMSCS) or call mediation node (CMN) can be built in a tandem layer to converge and forward the signalling messages such as bear independent call control (BICC) message or integrated services digital network user part (ISUP) message between two MSCSs. Similarly, tandem MGW, if needed, may also be provisioned to forward the traffic between any two visiting MGWs. For example: if a pair of TMSCS or CMN are built in the tandem layer t to connect the 51 local MSCS in a UMTS R4 network, there are at least 2 * 51 + 1 = 103 links (< 51 * 50 = 2,550 links) configured to forward the BICC signalling messages. With a layered structure, the least route number is calculated as follows: K N K = 1 Least route number = K N + 1 K = 2 (2) K N + K + K ( K 3) 2 K 3 where N is the number of network elements in the network (tandem elements excluded). K is the number of tandem elements. Figure 3 Layered structure of switching network (see online version for colours) Table 1 compares the flat with layered structure. Based on the same traffic, the layered structure, compared to flat structure, saves the link resources for local exchanges via traffic converging and forwarding. The flat structure has a lower CAPEX due to no investment on the tandem network elements. However, the reduced CAPEX does not guarantee to offset its higher cost of OPEX. Table 1 Comparison between flat and layered structure Characters Flat network Layered network Preferred Network processing capability Number of links Data configuration and maintenance CAPEX OPEX Depends on subscriber number or traffic model Depends on subscriber number or traffic model N * (N 1) K * N + K + K * (K 3)/2 Heavy work load; one element revised, all others revised No investment on tandem layer, but link budget is higher Maintenance scale larger; higher OPEX Easy to operate; individual revisal More CAPEX on tandem network elements, saves link budget Simplified structure; lower OPEX No difference Layered structure Layered structure Not clear Layered structure 4 The architecture of local network 4.1 Integration mode The integration mode in Figure 4 has been widely applied in GSM CNs. It strictly complies with the administrative division. All the NEs achieve localised deployment. The MSC, HLR, short message centre (SMC) and BSC are distributed at the same physical location. In signalling transmission, ISUP protocol is adopted. The signalling messages delivered between MSCs; mobile application part (MAP) is delivered between MSC and HLR and between HLR and SMC. In voice transmission, TDM-based E1 or T1 is selected to carry traffic between MSCs. Figure 4 Integration mode of local network (see online version for colours)

Evolving core networks from GSM to UMTS R4 version 97 In introducing R4-based NEs into the legacy local network, the easiest way is to directly replace the current 2G NE (MSC) with the new 3G NEs (MSCS and MGW). We do not need to modify the legacy network topology, but only need to replace the NEs in the network and allocate more link resources to accommodate increased traffic in 3G phase. However, it does not help achieve the all-ip target in evolving the legacy networks since the transmission medium is still based on TDM not IP or ATM in the integrated mode. 4.2 Detached mode An alternative is detached mode which centralises the MSCS while distributes the MGW locally. The MSCS is detached with MGW in the end layer and MSCS up into the tandem layer. Since MGW and MSCS are newly deployed into the legacy network, it is more convenient, compared to the existing links in legacy network, to achieve IP connections between new MGW and MSCS. Consequently, the centralised deployment of MSCS enables the wireless carriers to first set up a new IP-based private network for the interface Mc between MSCS and MGW. If detached mode is adopted, the evolution to all-ip actually starts from interface Mc. As per Figure 5, two options are available for the interface Mc between MSCS and MGW: IP/ATM over E1/T1 or IP/ATM private network. If current TDM transmission resources are still sufficient, it is suggested to select IP/ATM over current E1/T1 transmission network as an interim step before the IP/ATM private network is available for the interface Mc. Meanwhile, IP/ATM over E1/T1 also does not impact the ongoing development of IP/ATM private network to interface Mc. networks in the end layer. Through interface Nc, the MSCSs communicate with each other via ISUP messages carried by TDM links. The interface Mc between MSCS and MGW is the only portion that has achieved the IP transport via the newly built IP private network which may extend to interface Nc or Nb according to the respective plans of wireless carriers. Figure 6 shows the topology of TDM option which achieves IP transport in interface Mc. Figure 6 Figure 7 TDM option of detached mode (see online version for colours) IP option of detached mode (see online version for colours) Figure 5 Private IP network for interface Mc between MSS and MGW (see online version for colours) There are three options for voice bearer in the detached mode: TDM, IP or ATM bearer. The wireless carriers make their decisions to select a bearer medium for their networks by considering such factors as current TDM resources, physical bearer preference, the schedule to deploy IP/ATM private network for MSCS and MGW, CAPEX and OPEX. With TDM option, MSCSs are moved upward to locate in the tandem layer, while MGWs are distributed into local With IP or ATM option, MSCSs are centralised, while MGWs are distributed to deploy respectively. Due to the availability of IP or ATM bearer, the MSCSs are able to apply BICC to substitute ISUP protocol, in which a circuit identification code (CIC) is specific to TDM, in interface Nc via the IP private network. Defined by ITU-T Q1901

98 Y. Ouyang and M.H. Fallah Series Q and ITU-T Q1902.1 to 5 Series Q, BICC is developed to be interoperable with any type of bearer. It has no knowledge of the specific bearer technology which is referenced in the binding information (Cho and Kim, 2008). Either IP or ATM option achieves the non-tdm (IP or ATM) transport in interface Mc, Nc and interface Nb which enables the MGWs from different local networks to deliver the voice traffic via IP/ATM transport. The only exception exists in interface E between the MGW and legacy 2G MSC, which only allows TDM bearer for voice delivery, but does not support the evolution to IP or ATM bearer. Figure 7 shows the topology with IP option which achieves IP transport in interface Mc, Nc and Nb. Table 2 Networking characters CAPEX and OPEX R4 evolution QoS Summaries of the integration and detached mode Integration mode Integrated deployment with TDM bearer Typical SS architecture in local layer; high integrability Change to IP interfaces; huge modifications to the current network topology No difference with existing 2G network Detached mode with TDM option Detached deployment with TDM bearing voice Separation control from bearer; balanced capability in local level; optimised resource distribution Change to IP interfaces Not much difference with existing 2G network 5 The architecture of tandem network Detached mode with IP/ATM option Detached deployment with IP bearing voice Higher cost in initial investment, but benefits for the long-term One step evolution Differentiated service (DiffServ); multiprotocol label switching (MPLS) The tandem network is responsible for converging and forwarding the voice traffic and signalling messages between two visiting MGWs, two MSCSs or two MSCs. As mentioned in Section 3, the tandem network can be organised into either a flat structure when the network size (represented by the number of NEs in the network) is small or a layered structure if the network volume keeps expanding. In addition, another factor impacting the tandem network structure is the operation and maintenance (O&M). It is suggested that the mobile operators estimate the allowable tolerance of flat networking structure from both the O&M aspect and network size aspect. Based on flat structure in Figure 8, any two MSCSs or visiting MGWs have direct connection. There is no longer an actual tandem layer existing in the network. However, the visiting MSCSs and MGWs in the end layer play the tandem function as well. The layered structure is preferred if either the network size or the O&M load exceeds the threshold of flat structure. An appropriate opportunity to separate the tandem layer from end layer is at the time of building the IP-based SS MGW in the legacy network. The tandem NEs are advised to be provisioned with the deployment of IP (soft-switching) based MGW at the end layer. The CMN in Figure 9 relays BICC protocol. From Van Deventer et al. (2001), the CMN may be useful in a large-scale BICC network with a large number of interface serving nodes (ISNs), where the CMN would route the BICC messages. In this paper, the ISN denotes MGWs. Therefore, it is concluded that, with the independent tandem NEs provisioned in tandem layer, the signalling links between MSCS and CMN (or TMSCS) are available to deliver BICC messages for the long distance (non-local) call triggered by a soft-switching MGW in the local network. Tandem MGWs are also built with CMNs or TMSCS to forward the IP voice media stream between two visiting MGWs. CMN can be co-configured with TMSCS in markets with relatively fewer soft-switching MGWs or with lower traffic in the local layer. CMN can also be independent from TMSCS when the number or the traffic of MGWs keeps growing. Based on the independent structure that CMN separates from TMSCS, the independent CMN is only responsible for relaying BICC message, while TMSCS is responsible for delivering ISUP messages only. To achieve this independent structure, extra signalling links and routes are configured between the new CMN and visiting MGWs in local networks. The MGWs from different local networks, but under the same MSCS have direct connections. The MGWs from different local networks and under different MSCS communicate with each other via the tandem MGW in the tandem layer. Below is a summary for the scenarios in which signalling travels through CMN node. If the called party registered in the IP-based MGW in market B while the calling party belongs to the IP-based MGW in market A, the signalling messages will be forwarded via the independent CMN 2 in market B. The signalling routes follow this path: local MSCS in market A to CMN 1 in market A to CMN 2 in market B to local MSCS in market B. The voice traffic goes through this way: local MGW in market A to tandem MGW to local MGW in market B. The red and blue curve in Figure 9 denotes the signalling and traffic path respectively for this scenario. If the called party belongs to the TDM-based MSC in market B while the calling party registered in the IP-based MGW in market A, the routing will be pointed to TMSC server 2 which handles ISUP messages. The signalling routes follow this path: local MSCS in market A to CMN 1 in market A to TMSCS 2 in market B to local TDM MSC in market B. The voice traffic goes through this way: local

Evolving core networks from GSM to UMTS R4 version 99 MGW in market A to tandem MGW to local MGW in market B. If the calling party registered in IP-based MGW in market B while the called party registered in IP-based MGW in market A, the signalling routes follow this path: local MSCS in market B to CMN 2 in market B to CMN 1 in market A to MSCS in market A. The voice traffic goes through this way: local MGW in market B to tandem MGW to local MGW in market A. If the calling party registered in IP-based MGW in market B while the called party registered in TDMbased MSC in market A, the signalling routes follow this path: local MSCS in market B to CMN 2 in market B to TMSCS 1 in market A to local TDM MSC in market A. The voice traffic goes through this way: local MGW in market B to tandem MGW to local MSC in market A. Figure 8 Flat structure of tandem network (see online version for colours) Figure 9 Tandem network structure (see online version for colours)

100 Y. Ouyang and M.H. Fallah 6 The architecture of the gateway network In 2G, 2.5G and R99 phases, the gateway structure at CN side is not so complicated that the gateway NEs such as gateway mobile switching centre (GMSC) stands at the border of network to exchange MAP message with HLR and ISUP message with visiting MSC in the CN or exchange ISUP and telephone user part (TUP) messages with PSTN side. Meanwhile, regarding the voice transmission, GMSC is also the gateway to exchange the TDM-based G.711 voice stream between GSM side and PSTN side. Gateway NEs in UMTS R4 network, split into GMSCS and GMGW, and are also physically distributed at the border of the CN to achieve the functions of signalling conversion and traffic transition between PLMNs, between PLMN and PSTN, between PLMN and IMS or between PLMN and NGN. How to deploy the gateway NEs which interconnect with NGN, IMS and PSTN network, to some extent, decides whether or not the wireless carrier is able to achieve fixed mobile convergence (FMC). Take the interconnection between UMTS and NGN/PSTN as an example: in Figure 10, a pair of GMSCS and GMGW provisioned at the border of UMTS CN connects with a pair of SS and PSTN switch at NGN side via a back to back format in which the medium to carry traffic and transit signalling is still TDM-based E1 or T1. Therefore, this option does not achieve the all-ip structure on the gateway level. Figure 10 also displays the typical position of gateway NEs in UMTS and NGN networks. The alternative is to build an integrated soft-switch (ISS) gateway centre to achieve direct intercommunication between UMTS and NGN networks. Including ISS server and integrated MGW, the ISS gateway centre integrates and converges the gateway functions used to play by the individual gateway NEs such as GMSCS, GMGW and GSS distributed at the borders of UMTS and NGN network. This option helps the network actually achieve the IP structure in the gateway layer. Compared to the separated gateway structure in Figure 10, the integrated gateway centre in Figure 11 provides the integrated signalling process capability to support both IP and SS7 signalling, integrated media intercommunication capability to complete the conversion between multiple voice media streams such as G.711, AMR, G.729 and G.723, and integrated interconnection capability to provide multiple interfaces to different access networks such as TDM interface for PSTN and GSM network, ATM or IP interface for UMTS and NGN networks. As per the integrated structure shown in Figure 11, it is suggested that integrated the SS server in the gateway centre supports the multiple signalling protocol conversion function. SIP-I/T, as an extension of SIP protocol, is advised to apply between ISS server and SS. SIP I/T is also the basic protocol in IMS, so it helps the NGN side to converge with the IMS network provisioned from UMTS R5. On the other side between ISS and visiting MSCS, BICC protocol is applied to comply with the same protocol adopted in interface Nc between MSCSs in UMTS network. Therefore, the most important requirement on the ISS server is to support the protocol conversion between SIP I/T and BICC/ISUP. The integrated gateway is required to achieve the codec conversion between different voice formats and between different video formats. For example: it needs to support the conversion of G.711/G.729/G.723/AMR voice stream and H.263 video stream. Figure 10 Gateway NEs between UMTS and NGN network (see online version for colours)

Evolving core networks from GSM to UMTS R4 version 101 Figure 11 Integrated gateway structure for UMTS and NGN network (see online version for colours) 7 Conclusions and future work Mobile operators, especially those with GSM legacy networks, need to evolve their existing 2G GSM networks to an all-ip network. This transition is a process that needs to be managed effectively over a period of time. The paper first gave an overview of UMTS network including its architecture and topology, and then described two network structures for legacy network evolution: flat structure and layered structure. The pros and cons of the two structures are compared so that mobile operators can adopt an appropriate strategy to plan the architecture of their UMTS CN. Based on the theoretical considerations, the paper proposed a three-layer structural network for CS domain of UMTS R4 CN. A detail description of the architecture, topology and intercommunication of local layer, tandem layer and gateway layer is provided. The current literature is focused more on RAN and overlooks the CN. A lot of design philosophy and proposed architecture have been applied in the plan of UMTS radio network. However, not much effort, however, has been made on how to evolve UMTS CN. This may be explained by two facts that CN in either logical or physical structure is more complicated than RAN and the internal throughput or traffic in CN may vary by different vendors NEs. This paper explored both RAN and CN to provide a proposed solution for evolving a legacy network to an all-ip-based network with IMS and system architecture evolution (SAE) capable. The discussion of NGN, FMC or voice over IP (VOIP) eventually boils down to pure IP or all-ip, which is the vision of every wireless or wire line operator. The evolution from TDM to IP is a lengthy process, but never just a simple task of replacing the circuit-based NEs in the legacy network with new IP-based NEs. Consequently, forklift, as a radical way, is not an optimal strategy for the mobile operators to evolve their legacy networks. As discussed in this paper, the layered design philosophy does not mean to place the different NEs in a hierarchy in the network at once, but to help the mobile operators steadily transit from traditional circuit-based network to IP-based network step by step and layer by layer. The proposed architecture with its evolution path, as discussed in this paper, has been partially deployed in some tier 1 mobile operators in Asia and the Pacific area. We are continuing our study of converging with IMS and SAE networks. References 3GPP TS 23.002, Technical Specification Group Services and Systems Aspects; Network Architecture. 3GPP TS 25.401, Technical Specification Group Radio Access Network: UTRAN Overall Description. 3GPP TS 25.413, Technical Specification Group Radio Access Network: UTRAN Iu interface Radio Access Network Application Part (RANAP) Signaling. 3GPP TS 25.415, Technical Specification Group Radio Access Network: UTRAN Iu interface user plane protocols. 3GPP TS 29.414, Technical Specification Group Core Network and Terminals: Core Network Nb Data Transport and Transport Signaling. 3GPP TS 29.415, Technical Specification Group Core Network and Terminals: Core Network Nb Interface User Plane Protocols. Britvic, V. (2004) Steps in UMTS network design, Electro- Technical Conference, 2004. MELECON 2004. Proceedings of the 12th IEEE Mediterranean, 12 15 May, Vol. 2, pp.461 464.

102 Y. Ouyang and M.H. Fallah Cho, J-J. and Kim, N-P. (2008) A study of cost and network efficiency in BICC signaling protocol, Advanced Communication Technology, 2008. ICACT 2008. 10th International Conference, 17 20 February, pp.699 701. Harmatos, J. (2002) Planning of UMTS core networks, Personal, Indoor and Mobile Radio Communications, 2002. The 13th IEEE International Symposium, 15 18 September, Vol. 2, pp.740 744. Hoikkanen, A. (2007) Economics of 3G long-term evolution: the business case for the mobile operator, Wireless and Optical Communications Networks, 2007. WOCN 07. IFIP International Conference, 2 4 July, pp.1 5. ITU-T Q1901 Series Q: Switching and signaling, Specifications of signaling related to Bearer Independent Call Control (BICC), Bearer Independent Call Control protocol Corrigendum 1. ITU-T Q1902.1 Series Q: Switching and signaling, Specifications of signaling related to Bearer Independent Call Control (BICC), Bearer Independent Call Control protocol (Capability Set 2): Functional description. ITU-T Q1902.2 Series Q: Switching and signaling, Specifications of signaling related to Bearer Independent Call Control (BICC), Bearer Independent Call Control protocol (Capability Set 2): General functions of messages and parameters. ITU-T Q1902.3 Series Q: Switching and signaling, Specifications of signaling related to Bearer Independent Call Control (BICC), Bearer Independent Call Control protocol (Capability Set 2): Formats and codes. ITU-T Q1902.4 Series Q: Switching and signaling, Specifications of signaling related to Bearer Independent Call Control (BICC), Bearer Independent Call Control protocol (Capability Set 2): Basic call procedures. ITU-T Q1902.5 Series Q: Switching and signaling, Specifications of signaling related to Bearer Independent Call Control (BICC), Bearer Independent Call Control protocol (Capability Set 2): Extensions to the Application Transport Mechanism in the context of the Bearer Independent Call Control. Konstantinopoulou, C.N., Koutsopoulos, K.A., Lyberopoulos, G.L. and Theologou, M.E. (2000) Core network planning, optimization and forecasting in GSM/GPRS networks, Communications and Vehicular Technology, 2000. SCVT- 200. Symposium, pp.55 61. Mishra, A. (2003) Performance characterization of signaling traffic in UMTS core networks, Global Telecommunications Conference, 2003. GLOBECOM 03. IEEE, 1 5 December. Neruda, M. and Bestak, R. (2008) Evolution of 3GPP core network, Systems, Signals and Image Processing, 2008. IWSSIP 2008, June, pp.25 28. Van Deventer, M.O., Keesmaat, I. and Veenstra, P. (2001) The ITU-T BICC protocol: the vital step toward an integratedvoice-data multiservice platform, Communications Magazine, IEEE, May, Vol. 39, No. 5, pp.140 145. Vrabel, A., Vargic, R. and Kotuliak, I. (2007) Subscriber databases and their evolution in mobile networks from GSM to IMS, ELMAR, 2007, 12 14 September, pp.115 117.