4.1 Scenario 1: EPS with LTE Deployment with Existing 3GPP Installations

This scenario covers the situation where an operator has an existing GSM or WCDMA radio network and core network installation. In many cases, the operator will also have commenced the deployment of mobile broadband through the use of HSPA technology. The services in this instance can therefore be pure IP services, or based on the IMS, or some other proprietary mechanism. The key focus for such an operator is therefore to enable end-users to have access to higher bandwidth but also to ensure that the existing services available to GSM or WCDMA end-users are still accessible to the LTE subscribers where applicable.

A common deployment scenario is when an operator has an existing HSPA mobile broadband installation up and running. Spectrum is made available for LTE and the operator is then faced with the choice of using spectrum for building an increasingly dense 3G network, or to start deploying LTE in parallel in order to provide support for high-bandwidth data services. In this case, LTE can be rolled out in order to provide high-speed data services in ‘hotspot’ areas, for example city centres, airports or any other places that might require an operator to provide extra high bandwidth. Running LTE in parallel in this fashion allows an end-user to always fall back to HSPA coverage, which may offer up to, say, 40 Mbit/s per user. In this manner, the end-user can enjoy relatively high-speed coverage through a wide-area HSPA network combined with, say, up to 100 Mbits/peak rate in LTE hotspots.

Another possible deployment is where the operators of existing 3GPP accesses have not decided on HSPA and instead decided to roll out LTE-based mobile broadband network. The typical operators in these cases have GSM/GPRS and/or WCDMA networks deployed currently and thus deploying EPS with LTE. The same reasoning can be the deterministic factor for such deployment scenario, to restate: LTE can be rolled out initially in order to provide high-speed data services in ‘hotspot’ areas, for example city centres, airports or any other places that might require an operator to provide extra high bandwidth. And then the operator has a choice of either continuing with LTE over a wider coverage, that is for the rest of the operator’s network, or to rely on the existing deployed network (for example HSPA), deploy HSPA with EPC.

In terms of devices, most initial LTE devices are data centric in the form of modems for laptops (e.g. wireless data cards and dongles), but over time there will also be a gradual increase in the number of handheld terminals that support the LTE radio access.

In the initial phase of the roll out of LTE, it is natural for the operator to keep the LTE and the 2G/3G network infrastructures separate and introduce EPS in a step-wise manner. The main reason for this is to ensure that the existing revenue-generating infrastructure is not affected in the first phase of LTE deployment.

So, how would an operator decide on the mechanism to roll out LTE and EPC in parallel with their existing 2G or 3G network infrastructure? Initially the operator would need to assess where in their network they would want to improve the broadband capacity or where they wish to offer even high-peak bandwidth to their end-users. The operators would need adequate number of eNodeBs and Mobility Management Entities (MMEs) to handle the control plane signalling and mobility management for Intra and Inter Radio Access Technology (RAT) and also an adequate number of Serving Gateways and PDN Gateways. Since EPS is an all-IP network, an appropriate IP network infrastructure needs to be in place, for example Routers, DNS servers, FireWalls, etc. Additional decisions that would be needed is the protocol choice for S5/S8 reference point and whether dynamic policy control and charging would be deployed, since these decisions also increase the number of entities that need to be deployed. There are now several additional decisions that they need to take in order to deploy the network, but nothing out of the ordinary than what needs to done for any new infrastructure deployment.

Firstly, do operators need to separate the PDN gateway for the new subscribers or not? An operator may have the option of replacing existing GGSNs with PDN GWs because PDN GWs will support the GGSN functionality through the Gn–Gp interfaces from the SGSN. These interfaces use the GPRS Tunnelling Protocol (GTPv1) which allows interworking with legacy SGSNs. The PDN connectivity in the PDN GW in this case should be similar or equal to the PDN connectivity offered by the GGSN in order to ensure that the resident services on the existing 2G/3G network are available to the LTE subscribers during mobility to and from 2G/3G access. Of course, operators may choose not to supply some of the existing 2G/3G services to the LTE subscribers; for example, dedicated PDNs for a particular company might not be prioritized in the first phase of the LTE roll out. Another option is to deploy new PDN GWs in parallel with existing GGSNs, allowing existing terminals to continue using GGSNs as anchor points, while LTE-capable terminals would rely on the PDN GWs. Selection between these anchor nodes would be handled by the SGSNs, as previously discussed in Chapter 3 and which will be further described in, Section 9.2.

The Serving GW may or may not be integrated with the PDN GW. Even in the cases where they are integrated, however, there will be situations where a subscriber will use the Serving GW on one node and the PDN GW on another node. For example, when a user is roaming, they will use the Serving GW in the roaming network and the PDN GW of their home network in case of Home Routed Traffic (see ‘Architecture Overview’ Chapter 3 for more on the various roaming models supported in EPS). Another example where a Serving GW and PDN GW may be located in separate nodes is when a subscriber connects to multiple service networks (e.g. Corporate network and IMS) via separate APNs leading to 2 different PDN GWs.

GTP may be used on the interfaces between the PDN GW and the Serving GW; S5 is the interface used internally, while S8 is the interface used while roaming. GTP is selected in this example to illustrate further as it is already well established for roaming in 2G and 3G networks.

Secondly, an operator will need to decide whether to deploy stand-alone MMEs, or combined MMEs and SGSNs. Note that such deployment of combined node is more natural for operators with existing 2G/3G networks and there are some benefits to such deployment option.

Essentially, therefore, an operator is most likely to add LTE as a stand-alone network, get some LTE terminals into their network and then start offering LTE access to subscribers. Subscribers can then request the operator to upgrade their subscription, which an operator will do through updating the end-users subscription in the Home Subscriber Server (HSS) and/or with a new UICC card. This means that when an end-user accesses the network via GSM or WCDMA, the instruction can be given to the SGSN that it needs to select a PDN GW, rather than a GGSN. The simplest way of doing this is through using different Access Point Names (APNs) depending on if the terminal is LTE capable or not. There are also features which allow the SGSN to select a Serving GW/PDN GW instead of GGSN if the terminal is LTE capable or select a GGSN if the User Equipment (UE) is only 2G/3G capable, based on information from the UE. The use of DNS and other terminal capabilities for such selection is further described in Chapter 9. Use of DNS and other terminal capabilities for such selection can be found in Part III – Selection Functions. When an LTE subscriber starts in GSM coverage but roams into LTE coverage, they will need to have their PDN connection established via a PDN GW, rather than the legacy GGSN.

The operator now has a functional but fairly stand-alone LTE network with EPC in parallel with a 2G/3G packet core. Note that if the operators have not upgraded to a Rel-8 EPS network for 2G/3G networks, then Inter RAT handover will only work in the direction from LTE towards the 2G/3G network. When more and more of the subscribers are migrated by the operator to allow LTE access as well, this leads to a situation that is administratively inefficient for the operator to maintain two separate core network architectures (i.e., one with Gn/Gp based core network with SGSN/GGSN and its associated infrastructure and the other being the upgraded EPC architecture) for the different radio network accesses.

As long as the nodes are installed within the deployed network, there should be little problem in maintaining the parallel networks unless an operator needs to update them regularly; two parallel networks should function perfectly well. There may, therefore, be a number of GGSNs that will remain installed within an operator’s network for some time. At some point, however, it may become more economically viable for the operator to migrate to a full EPS architecture and decommission the remaining core network infrastructure originally installed for the 2G/3G networks. However, it may be possible to upgrade the existing GGSNs to PDN GWs, in which case, the migration path will be a relatively easy process.

The S4-SGSN, first defined in Rel-8, has been specified to handle the situation where handover from and to LTE is required as well as provide functions developed for the EPC in order to connect via a Serving GW and a PDN GW. Essentially, the S4-SGSN provides a control plane interface between the S4-SGSN and the MME and a user plane interface between SGSN and Serving GW. The S4-SGSN also enables Idle Mode Signalling Reduction (ISR) function (as described in detail in Part III – Session Management and Mobility,, Section 6.4), which reduces signalling overhead over the air between the UE and the network as a user moves between 2G/3G and LTE radio networks. The capability to support ISR and support of handover both to and from 2G/3G access are the essential differences between an S4-SGSN and a Gn-SGSN.

With the introduction of mobile broadband, it is expected that there will be an exponential increase in data traffic, this has a significant effect on the cost of a network in terms of scalability – in essence, an operator will need to ensure that their networks are also configured in the most cost-effective manner possible. Take the SGSN, for example, which is responsible for handling control plane signalling and maintaining session data whilst also keeping track of the UE; this implies that the SGSN should be scalable for processing, handling security associations and the states of millions of users. The User Plane nodes, meanwhile, for example Serving GW, need to be scaled to match the expected data volumes. The PDN GW is another example of a user plane node; this may also need to be scalable for Deep Packet Inspection (DPI) in addition to data volumes. The PDN GW also needs to interface the Policy and Charging Control (PCC) infrastructure in order to download filters for Quality of Service (QoS) control and therefore also needs some additional control plane capacity for this.

It is likely, therefore, that an operator will want to avoid having the SGSN heavily involved in the user plane as it is responsible for signalling and does not have any responsibility for the handling of subscriber traffic. An operator wanting to avoid the cost of upgrade to scale the SGSN in order to handle the vast amounts of data traffic that a mobile broadband network brings, such operator will naturally want to remove this ‘hop’ in their network as it will reduce their overall cost.

For operators with an existing WCDMA/HSPA installations, it is possible for them to bypass the SGSN for the user plane and take a subscriber’s traffic directly from the Radio Network Controller (RNC) into the GGSN in case of non-roaming scenarios for existing packet core before EPC. With S4-SGSN it is possible to do a similar bypass, but in this case the subscriber traffic is taken directly to the Serving GW/PDN-GW and as Serving GW provides the mobility anchor for 3GPP subscribers, the bypass of user plane traffic from SGSN is supported for roaming and non-roaming scenarios via Serving GW in EPC (in case of GPRS, only non-roaming bypass is supported).

This allows for an operator’s upgrades to the SGSN to be directly related to the upgrades required for the control plane. Through bypassing the SGSN for user plane traffic, the operator will no longer need to match the control plane and user plane capacity to one another. Through implementing them in separate nodes, it is possible to scale them independently.

In this deployment scenario, though upgrading of the HSS is definitely desirable for LTE subscribers, it is not mandatory due to the fact that both 3G UICC (SIM card) and the existing HSS may be reused for LTE subscribers but with the use of Interworking Function (IWF) where mapping of functions as well as protocols (Diameter to Mobile Application Part (MAP) and vice versa) may be required. The use of IWF may have its limitations in addition to protocol mapping during critical time-consuming signalling like the Attach/Registration procedures, but it can be useful for operators in this scenario where the HSS upgrades may not be possible in a timely fashion. Different vendors may have different solutions for the support of both MAP and Diameter towards the User Management infrastructure.

4.2 Scenario 2: LTE and EPS for Greenfield Operators

If we consider a pure Greenfield scenario where the operator starts LTE and EPC without having to consider any existing deployed networks and interworking, then issues arise which are purely related to roaming considerations and interworking aspects as well as the use of GRX (GPRS Roaming Exchange, this is the Roaming network deployed by operators who have roaming agreements for GPRS, more information available at: www.gsmworld.com) networks that an operator’s environment requires them to use. Also the need for national roaming is likely due to that LTE coverage may be assumed to be spotty at the beginning of the deployment. The Greenfield operator may hence need to rely on partners within the same country for offering wide area service coverage for its subscribers.

An operator in this scenario will start with the LTE access network connected to an EPC. As the operator does not have any pre-existing requirements, they are in a position to go with all the ‘bells and whistles’ an EPS can provide. A ‘fully fledged’ IP-based network would need to include the properly configured DNS infrastructure for 3GPP networks and MMEs, Serving GWs, PDN GWs (combined or stand alone), HSS and other entities depending on the operator’s choice such as Equipment Identity Registers (EIRs), Policy Control and Charging Rules Functions (PCRFs). Of course here we are only talking about the core EPS functions, while it is assumed that operators will also deploy the necessary entities that are required to support charging, security and legal intercepts as well as routers, firewalls and the IP backbone as demanded for an IP-based network. The users will have the UICC (SIM cards) and appropriate handsets enabling the higher level of security that has been designed for EPS. This is the simple part of the operator’s deployment scenario, but it gets interesting when the operator starts to consider LTE roaming as well as supporting, for example, 3G/LTE-capable subscribers as inbound roamers to their network. In such scenarios, operators need to consider supporting the necessary GRX network requirements, which may require support of SS7-based MAP protocols to/from HSS (via IWFs) as well as necessary GTP variants running on the GRX networks. When considering Greenfield operators, it can be assumed that they will not deploy 2G/3G radio access networks and thus they would not need to support any SGSN or GGSN functionality except what is required for roaming. But their own subscribers shall be able to roam into operators’ networks where LTE/2G/3G are supported or only 2G/3G are supported. Assuming the roaming support requires that it is mobile broadband only, since LTE is not capable of supporting circuit-switched services, the roaming considerations are only for Packet Core data traffic, the Greenfield operator must be able to handle queries to and from these networks over roaming interfaces that may require support of SS7-based protocols as well as pre-Rel-8 protocols. There should not be any other significant requirements on the Greenfield operators, and while it may not be as simple as it may sound to enable this type of roaming, it is very much feasible and one of the key interworking supported in 3GPP systems as can be seen from the analysis of Scenario 1.

EPS also provides a Greenfield operator the possibility to consolidate any other access networks via a common Packet Core and Policy architecture.

4.3 Scenario 3: LTE and EPS Deployment for 3GPP2 Operators

These operators would fall into the category of 2G/3G operators who have well-established CDMA networks with a substantial customer base and investments. Thus, depending on the strategy for the long-term commitment towards LTE and the strategy for existing CDMA networks, the migration timeline and deployment intensity may be quite wide-ranging and different across the 3GPP2 operator community.

Even though in a sense this kind of operator can be seen as Greenfield operator for LTE and EPC (e.g. 3GPP developed system), their requirements are not the same as the ‘pure’ Greenfield operators. One of the key aspect to consider is the strategy and plans for their existing CDMA networks and how ‘tightly coupled’ the LTE deployment will be with the evolution of the HRPD networks. In addition, the operator’s existing customer base and roaming partners need to be supported and a well thought out ‘full migration’ to LTE may well be the most convenient but not likely scenario at the beginning of LTE/EPC deployment. When it comes to LTE/EPC deployment, the 3GPP2 operators have exactly the same requirements as a Greenfield operators, that is the need for deployment of E-UTRAN, HSS, MME, Serving GW and PDN GW, as well as the security and charging entities and DNS and IP infrastructure. But in addition to these basic considerations, an incumbent 3GPP2 operator must also consider the interworking support for their existing HRPD networks as well as securing dual-mode terminals (CDMA and LTE) as well as deployment of UICC. Depending on the type of interworking, there are different requirements on the network. In the case of ‘tight interworking’ (also known as optimized handover support), the 3GPP2 operator needs to upgrade their HRPD access networks as well as deploy HSGW (replacing or upgrading existing Packet Data Serving Node (PDSN) components of their network). Also upgrades of the Authentication, Authorization and Accounting (AAA) infrastructure and support functions in its MME and Serving GW are needed including S101 and S103 interfaces that allow for optimized handover between these two networks. In the case of ‘loose interworking’ (also known as non-optimized handover), the HRPD Access network needs to be able to support broadcast of certain parameters as well as configuration of LTE and HRPD cell information in order to facilitate the handover efficiently. In this case there are no specific functions needed in the MME, but the PDSN (e.g. the Access GW for 3GPP2 HRPD system) needs to be upgraded to an HSGW including EPS-related functions. The AAA infrastructure used in 3GPP2 systems also has to be upgraded to EPS AAA infrastructure and the UICC would be required for the handset.

Depending on what kind of inbound and outbound roaming agreements will be supported, the operator may choose to deploy additional functions (like in the case of Greenfield operators) to enable roaming with 2G/3G networks.

It is expected that the operator’s 3GPP2- and 3GPP-based networks (i.e. LTE/EPC) would be running in parallel for an extended period of time and this timeline would vary from one operator to another operator depending on the operator strategies including whether they plan to use LTE as a hotspot mobile broadband access in strategic locations like city centres and major urban centres or if they plan a ubiquitous LTE network in the near term.

In the long run, it can be expected to be more efficient to run a common network providing full functionality for the operator instead of running two networks with different technologies in parallel for a long time. In addition, as EPS is being developed to provide a common packet core network for the future and for any access technology able to connect using one of the mechanism EPS provides, EPS also then opens up opportunities for further consolidating additional access networks via a common packet core.

Terminal availability supporting multiple access technologies for one common core network would also be more desirable than multiple ones. Thus a consolidation of the networks also allow for simplification of the handsets and to focus more on attractive services and applications towards end-users. Availability of terminals supporting LTE as well as CDMA 1xRTT and HRPD technologies may also contribute to a prolonged maintenance period of the CDMA networks even after considered undesirable by a specific operator’s long-term goal.

But for this scenario, the operator also has to consider the migration path for their existing customer base and that may vary from one operator to another operator depending on the strategy of what to do with the circuit-switched service offerings and corresponding infrastructure. Some operators may choose LTE/EPC for a much wider deployment and also utilize the IMS and Multimedia Telephony (MMTel) for voice multimedia services, while an operator intending to keep the existing circuit switched services may rather use the 1xRTT infrastructure and use the CS Fallback mechanism to continue to provide the voice services this way. See Chapter 5 for more information on the realization of voice services for different operator scenarios. One may consider reading Part II – Services in EPS, in order to better understand the opportunities in a glance.

Note that depending on the protocol choice possible for the roaming interfaces, the complexity of the deployment may also increase for an operator.

4.4 Scenario 4: WiMAX and WLAN Operators

Even though it might be questioned whether operators operating exclusively in these access technologies would see any motivation to migrate towards EPC (without LTE), the benefits can be seen from a common core network as well as handover/session continuity perspective with existing and well-established access technologies with huge number of subscribers (over 3 billion and counting…). The EPS could also open up the possibility to get into roaming relationship that exists, for example, via GRX with international community. One may consider the scenario that an LTE operator establishes partnership with, for example, a WLAN access provider, connected via the EPC networks and providing indoor coverage in certain local network environment. Since hand­over/session continuity with IP and session preservation is supported, users can easily move in and out of such coverage while maintaining their sessions via EPC. Depending on the scenario, such non-3GPP access network providers may desire to establish a full EPC network using AAA infrastructure, Access Gateways and PDN Gateways (and optionally Serving GWs in a very specific situations). Alternatively, an operator may choose to deploy AAA infrastructure and Access Gateways to connect via an existing operator’s EPS networks, where special business relationship may be established with such access providers. This type of symbiotic relationship may benefit both operators business, depending on the market environment they are operating on (e.g. wide usage of WLAN in a certain city centre).

An additional key question is of course the availability of dual/multi mode handsets, providing support for both LTE (and/or other 3GPP accesses) and the non-3GPP access that the operator is using. In the case of WLAN this combination can be expected to be very common in terminals.

4.4 Scenario 5: Consideration for EPC-only Deployment with Existing 2G/3G Accesses

Existing 2G/3G operators may choose to deploy/upgrade to EPC without the necessity of deploying LTE. Some aspects of this scenario has been already touched upon in Scenario 1, and here we highlight a few of these aspects. Some of the main benefits have already been mentioned such as support for handover to/from LTE, support for handover with non-3GPP access networks, the ability to provide local breakout more efficiently, and the built-in support of optimized user plane traffic (also known as Direct Tunnel for 3G packet core) for both roaming and non-roaming scenarios. Other benefits include allowing IP-only networks thus avoid maintaining SS7 networks for the packet core, being prepared for support/inclusion of other non-3GPP access network connection (such as explained in Scenario 4), efficient network operations and maintenance and enhanced QoS and Policy Control and Charging support. Since the terminals that support existing 2G/3G procedures will be supported without any problems, operators can continue to serve their existing subscribers. In addition, the selection of the GW (GGSN vs. Serving GW/PDN GW) based on the terminal’s network capability (which indicates if the terminal can support LTE or not) can be used to divert the subscriber towards a Serving GW/PDN GW (when LTE capable) or towards a GGSN (when not LTE capable). As it becomes increasingly costly for vendors as well as operators to continue to maintain multiple tracks of architecture which leads to multiple tracks of products, it would serve the overall community to converge over time towards a minimal set of product variants

There are obviously other possible scenarios that we have not discussed here. The purpose of this section has been to explore some possible key scenarios. During 3G deployment, some interesting aspects came about such as incumbent national operators not getting the license for 3G operations. Such obstacles also led to the creation of new solutions in 3GPP, for example the Network sharing feature allowing an operator with a 3G license to share their radio network with an operator without a 3G license and thus able to provide services using 3G to their subscribers. LTE and EPC have been developed when keeping such scenarios in mind, meaning that radio network sharing as well as sharing of MMEs by multiple operators are supported in the standard specifications.