UMTS Forum Report 33, November 2003

1. Introduction

The purpose of this study is to provide a first estimate of the offered traffic characteristics for UMTS/3G network, in particular how these translate into uplink and downlink requirements. When reviewing all mobile aggregate traffic, 2G traffic, which is outside the scope in this report, should also be taken into account. Whilst there are clearly many unknown factors, this study provides a reasonable picture of the offered traffic of future 3G services.

Most of the mobile services studied do not currently exist. These results are therefore dependent upon the service assumptions made. A sensitivity analysis determined which assumptions had significant impact of the aggregate results. Those assumptions are presented.

This study builds upon previous market analysis work completed by the Forum, using the services category framework shown in Figure 1.1 (1). This framework includes all anticipated possibilities for 3G services. The framework cleanly segments the services by user segment, type of functionality, and connectivity. Its specific design is broad enough to include new individual service concepts while at the same time, eliminate double counting of revenue and subscribers.

Figure 1.1. 3G services framework with its six service categories.


Source: UMTS Forum and Telecompetition, Inc., Report 13, September 2000.

The use of this framework along with the underlying subscription forecasts from previous UMTS Forum market studies prevents double counting of traffic volume and ensures consistency with market forecasts using a representative mix of service types for study.

(1) The UMTS Forum services category framework was originally presented in Report 9, and used as the basis for service forecasts in Reports 9, 13, and 17.

1.1 Service Traffic Characteristics Defined

Service traffic characteristics describe the unrestrained end user traffic offered to the UMTS/3G network; not considering network imposed asymmetry or other hardware and software limitations or remedies (such as traffic caching). Service traffic characteristics therefore refer to the expected nature of the traffic offered to the network, not the actual traffic characteristics over the air interface. No impediment to the build up of traffic is considered (such as the non-availability of devices or spectrum). The traffic loads are based on forecasted traffic in 2010, after networks have been deployed for more than five years.

Traditionally, the examination of service traffic for the purposes of spectrum calculations has only considered the technical characteristics of a particular application or service. For example, the spectrum requirements for voice have simply considered the speech coder characteristics (i.e. data rate), the spectrum efficiency of the modulation scheme and network and the predicted offered traffic. This was possible in the past, because circuit-switched voice service is symmetric and only the bulk traffic was taken into account.

Mobile multimedia services, however, introduce new challenges, such as traffic asymmetry, driven by the wide variety of multimedia-based activities available to the user. For example, web browsing typically has much more traffic coming to the user (downlink) than from the user (uplink). Telecompetition's ATIVA Research Tools, which analyse the "propensity to buy" for any given service enables the determination of unrestrained traffic demand, taking into account a number of variables, which are explained below. Thus, the methodology adopted incorporates this variety of activities and service variables in a way that relates them to their forecasted market demand.

Specifically, in this report, service traffic characteristics includes all traffic that end users would offer the network based upon baseline location and subscriber profiles plus seven service variables and states (2), shown in Figure 1.2. This is the unconstrained traffic offered to the network. Many other factors may affect the actual network traffic characteristics - both technical (e.g., traffic shaping, required overhead) and market-oriented (e.g., pricing plans).

The first step in this study was to develop a structured way to consistently analyse each service category. The structured approach chosen is based on nine service variables and their associated states. These variables capture the most significant attributes of mobile service that impact the traffic loads and asymmetry. The nine service variables are:

- Location profile
- Subscriber profile
- Connectivity Type
- Market segment
- Media / Activity Type
- Network cases
- Contact type
- Terminal type
- Transmission mode
- Service level

These service variables and their states are shown in Figure 1.2 and further described in Section 2.

Figure 1.2. Service variables and associated states used to analyse traffic characteristics.

For each of the six service categories, relevant states for each service variable shown in Figure 1.2 are analysed. This analysis resulted in a large number of permutations of the service variable states - each representing a potential individual service. After identifying the permutation that represented reasonable realistic services, related assumptions included the following:

- File sizes for non-real time sessions for the uplink (UL) and downlink (DL).
- Asymmetry of relevant service permutations (UL/DL).
- Data rates - customer expectation of speed over the air interface (UL and DL).
- Session frequency and duration.
- Busy hour characteristics and traffic distribution.
- Subscriber adoption of individual services.

Total offered traffic includes the average aggregate offered traffic over all service categories.

The total offered traffic has to be considered separately for uplink and downlink because it is possible that the traffic asymmetry for one service will be offset by another service.

The volume and proportion of traffic related to the first two service variables (location and subscriber profiles) was determined by using the ATIVA Research Tools, which uniquely forecasts the propensity to buy (or use) such services, based on large, detailed and extremely well qualified social data. This is described further in Section 8.

(2) The volume and proportion of traffic for each of the service variables was determined by using the ATIVA Research Tools, which forecasts the propensity to buy (or use) such services, based on large, detailed and well qualified social data, as described in Section 8.

1.2 Study Scope

The study considers the following:

- Six service categories.
- Up to 288 service permutations within each service category.
- Two market segments within each service category.
- Unconstrained offered traffic only based on future market demand (at saturation).
- Average aggregate daily and busy hour traffic.
- Average 3G subscriber demographic profile and service demand in a Western European country.

The study results address:
- Expected asymmetry per service per market segment
- IP session duration
- Busy hour offered traffic, including identification of the busy hour
- Number of subscriptions per service
- Total traffic per subscription (uplink + downlink)
- Total traffic per country and per service category
- Total traffic per 3G subscriber
- IP sessions ("call attempts") per subscription
- Service level

The following general methodology is used in this study (3):

- Choose the UMTS Forum service forecasts from a representative Western European country as the baseline for determining subscriber and subscription levels.
- Use the service forecasts for the year 2012 to project the anticipated traffic offered to the network once 3G has reached a mature subscriber penetration level.
- Using location and subscriber profiles, determine service subscription and per-subscription frequency of use and session duration for each service category to develop individual subscriber offered traffic volumes.
- Analyse each service category based on the seven service variables and states. Exclude states that don't apply.
- Estimate the traffic volume for each service permutation.
- Develop traffic distributions and service asymmetry for each market segment and service category.
- Calculate traffic load, busy hour and aggregate asymmetry.
- Test sensitivity of service assumptions to determine the most critical traffic assumptions.
- Aggregate service level traffic and busy hour loads to determine overall traffic characteristics.

(3) This study analyses traffic characteristics for a representative Western European country. Thus traffic characteristics and estimates are presented on a total country basis and are not analysed in more granular detail, such as by cell site or specific metro area. Because it is recognised that network engineering requires this granular level of detail, the study also provides the data on a per subscription or per subscriber basis.