1 3GPP Radio Resource Control in Practice Antonio Barbuzzi, Pekka H. J. Peral¨ a,¨ Gennaro Boggia, and Kostas Pentikousis

Abstract—3GPP-standardized networks have been evolving at being important for standard performance metrics such as a very fast pace over the last decade. Cell capacities increased throughput and delay, RRC is critical for energy efficiency, more than an order of magnitude, latencies have become consid- because many of its features were introduced to increase erably smaller, and worldwide deployment has changed the way people access services. In this evolution, efficient mechanisms for mobile phone battery life. This is an area of active research radio resource control (RRC) have played a key role. In this [1], but it is beyond the scope of this article. paper we will review the RRC state transition model and follow This paper is an extended version of [2] and contributes its development from its early stage backwards compatibility towards a better understanding of 3GPP/UMTS networks in and integration with GPRS at the outset of UMTS Terrestrial practice in three distinct ways. First, we present a com- Radio Access Network (UTRAN) deployments to state-of-the art HSPA-enhanced networks. This paper also overviews recent prehensive overview of the 3GPP standards with respect to work on empirical measurements from networks that study RRC. We summarize succinctly the evolution path and key the “theory and practice” of RRC state transitions. Finally, technical choices made along the way until we reach the we present our 3G Transition Triggering Tool (3G3T), and use latest system releases. Second, after introducing 3G3T (3G it to study empirically network configuration parameters that Transition Triggering Tool), which can be used to trigger RRC prompt RRC transitions. Our results come to the aid of fully understanding the behavior of RRC state transitions and lead us state transitions, to measure one-way delay, and to determine from “theory” to “practice” on RRC mechanisms. the delay due to paging procedures and/or channel setup, we present its extended capabilities. Third, we use 3G3T in a Index Terms—Radio Resource Control, 3GPP, EDGE, UMTS, HSPA, LTE, Traffic Measurements. comparative study carried out in four live UMTS networks in three different European countries. Our study is, to the best of our knowledge, the most extensive in the published literature I.INTRODUCTION concentrating on studying RRC state transitions. It is also the Effective radio resource management is a cornerstone in most up-to-date as we empirically study all currently-deployed all wireless communication systems. In 3GPP networks, in 3GPP releases (i.e., Rel. 99, 05, and 06). We show how the particular, Radio Resource Control (RRC) evolution has been operator network settings may differ drastically from each instrumental in delivering increased throughput and lower other and that one cannot always infer network behavior based delays. In simple terms, RRC manages all things related on the available literature. 3G3T is so far the only tool useful with wireless resource allocation, effectively implementing to bridge this gap, especially since operators are typically not the lower part of the protocol stack. Due to the scarcity of willing to share this type of information. radio resources, wireless connections need to be managed with The rest of this paper is organized as follows. Section II great care and accuracy. Wireless resource usage optimization provides a comprehensive review of the evolution of RRC in is tightly coupled with end-user experience and the overall /3G, and beyond networks. Section III reviews previous capacity of the system. For this reason, we will present work on simulation-based evaluation of RRC state transitions the fundamentals of RRC in 3GPP networks shortly below, and presents our empirical evaluation of this transitions in live emphasizing the role of system evolution. We will focus on the networks. Finally, Section IV concludes the paper summariz- role and importance of RRC state transitions and explain the ing the main points. inner workings both in theory (as specified in the standards) and in practice (as we measure state transitions in live, public networks). To obtain the latter, we developed a purpose-made II.RADIO RESOURCE CONTROL EVOLUTION tool which can accurately detect and measure state transitions. In cellular networks, RRC refers to the functional block RRC manages both active and idle connections. The former (including the associated signaling) that essentially controls carry user traffic. Idle connections, in contrast, do not carry the lower part of the protocol stack, that is, Radio Link user traffic, but they provide global reachability for mobile Control, and the . devices, a pillar for the success of 3GPP networks. Besides RRC is used to initialize connections, reserve resources, establish and maintain connectivity, and (at the end of A. Barbuzzi is with EURECOM, 2229 route des crłtes, Sophia-Antipolis, France. e-mail: [email protected]. This work was done when he connection lifetime) release all associated radio resources was at“DEE - Dip. di Elettrotecnica ed Elettronica”, Politecnico di Bari. shared between users and network. RRC handles broadcast G. Boggia is with “DEE - Dip. di Elettrotecnica ed Elettronica”, Politecnico system information, manages radio bearers, controls paging di Bari, v. Orabona, 4 - 70125, Bari, Italy. e-mail: [email protected]. K. Pentikousis is with Huawei Technologies European Research Centre, and RRC connection mobility. Finally, RRC is responsible Carnostr. 4, 10587 Berlin, Germany. e-mail: [email protected]. This of ciphering control, outer loop power control, and User work was done when he was at VTT -Technical Research Centre of Finland. Equipment (UE) measurement and reporting [3]. Table I P. Peral¨ a¨ is now with Accenture Ltd, Itamerenkatu¨ 1, FIN-00180 Helsinki, Finland. e-mail: [email protected]. This work was done when he summarizes the relevant RRC technical specifications (TS) was at VTT -Technical Research Centre of Finland. issued by 3GPP. In what follows, we will refer to this Table 2 presenting the evolution of RRC. Access (HSPA), the High Speed Downlink Shared CHannel (HS-DSCH) and the Enhanced DCH (E-DCH) may also be TABLE I used for downlink and uplink transmissions, respectively. In Cell FACH state, data are carried through common channels; RRC was first specified in the context of 3G/UMTS typically, the Random Access Channel (RACH) for uplink, networks. However, similar mechanisms for allocating and and the Forward Access Channel (FACH) for downlink. In managing radio resources were also defined for previous Cell PCH and URA PCH states, UEs listen to the Paging cellular systems, starting with the introduction of General Channel (PCH) and Broadcast Channel (BCH), but uplink data Packet Radio Service (GPRS) for Global System for Mobile transfer is not possible [3], [4]. Communications (GSM) networks in GSM Rel. 97 (see RRC state transition changed only marginally between Fig. 1). After all, the management of radio resources and Rel.’99 and Rel. 7. For example, in Rel. 6 a new method their efficient allocation according to user needs is at the core for a handover from GPRS packet transfer mode to RRC of a wireless packet switched network. connected mode was introduced. Rel. 7 introduced Contin- uous Packet Connectivity (CPC), which is an improvement FIGURE 1 of the scheduling for HSPA-related channels, but does not affect the state transition model. CPC consists of three main The GPRS radio resource procedures and operating modes building blocks: Discontinuous Transmission, Discontinuous are specified in 3GPP TS 04.08 (see Table I). There are Reception, and High Speed Shared Control CHannel (HS- two radio resource operating modes: packet idle and packet SCCH)-less operation [5]. Another important feature is the transfer (Fig. 2). In packet idle mode, the UE listens to possibility to schedule HSPA related channels in Rel. 5 and the Packet Broadcast Control channel and to the Paging Rel. 6, respectively. channel without allocated Temporary Block Flows (TBFs), RRC state transitions depend on Buffer Occupancy (BO) corresponding to unidirectional connections between the UE levels at the Radio Link Control (RLC) layer. Typically, a and the network (see TS 03.64, in Tab. I). If the UE is in idle transition from Cell DCH to Cell FACH takes place when BO mode and needs to transmit data, it switches to the packet is zero and a threshold for Dedicated CHannel (DCH) release transfer mode effectively asking for radio resource allocation timer is exceeded. The transition back to Cell DCH is carried through a TBF on one or more packet data physical channels. out if the BO level exceeds a threshold value for a set time, that The transition from transfer to idle mode (releasing any as- is, there are data waiting to be transmitted. Moreover, if the sociated TBFs) is regulated by the TBF-Timer, restarted every period of inactivity in Cell FACH is long enough (ranging 2- time data is received on the TBF radio bearer. According to 10 s) the UE may change its state to Cell PCH or URA PCH. the specifications, the TBF-Timer is set to 5 s. The number of The transition back to Cell FACH or Cell DCH is usually TBFs allocated is based on the capacity-on-demand principle. carried out if user activity is detected. The RRC mode may be GPRS was enhanced (Enhanced GPRS (EGPRS)) in GSM changed from connected to idle, if the inactivity timer triggers Rel.’99 with the introduction of the Phase 1 of Enhanced Data a transition, or the RNC is (over)loaded. In the latter case, the rates for GSM Evolution (EDGE). EGPRS slightly impacts the RRC connection is released [4]. radio resource and the (TS 04.60). Rel. 9 (released in 2009) and Rel. 10 (in 2011) do not introduce relevant changes in RRC. 3G/UMTS In 3G/UMTS networks, RRC is employed between the GPRS after Release 4 UE and the (RNC) in the access 3GPP specifications (Rel. 4 and later) describe the mainte- network. In all 3GPP releases (from Rel. 99 to Rel. 7), the nance and evolution of GPRS/EDGE standards. We now dis- RRC state transition model has remained largely unchanged cuss the relevant modifications introduced for radio resource (TS 25.331). Universal Mobile Telecommunication System allocation in light of the previous discussion on 3G/UMTS (UMTS) terminals have basically two operational modes: idle networks. and connected. Fig. 2.b illustrates RRC states and transitions In Rel. 4, the GSM EDGE Radio Access Network (GERAN) as per see TS 36.331. (TS 43.051) was adopted to complete the evolution of GSM standards towards UMTS. GPRS may use the Iu interface FIGURE 2 (in addition to the classic A/Gb interface) and the QoS architecture of UMTS. When the UE is powered on, it enters idle mode, i.e., Concerning the radio resource management, the original it attaches to the network but does not actively engage in long duration of TBF was shown in practice to deal poorly data transfers. When a RRC connection is established, the with bursty traffic, so that the Extended Uplink TBF (to terminal switches to connected mode and the UE can be in any maintain the uplink TBF during inactivity states) was added of the following four service states: Cell DCH, Cell FACH, in Rel. 4 (TS 44.060). Its counterpart for the downlink was Cell PCH, and URA PCH. In Cell DCH state, according already introduced in Rel. 97. to Rel.’99, user data are transferred through a dedicated The RRC protocol for GERAN Iu mode is the combination channel (DCH). If the network supports High Speed Packet of the GSM and UMTS RRC radio resource operating modes. 3

The new protocol is a compromise between the adaptation LTE of the GERAN Iu features to the UTRAN Iu mode and the need for backward compatibility [6]. New procedures, such In Long Term Evolution (LTE), introduced in Rel. as RRC connection and mobility management, were added to 08, the RRC state model was simplified. As illustrated the GERAN Iu mode and others were modified, including han- in Fig. 2.a, there are two main states: RRC IDLE and dover and paging. Two operation modes were defined: RRC- RRC CONNECTED. There is no need for additional states idle and RRC-connected, equivalent to the ones for UTRAN in LTE, since there is no concept of common or dedicated Iu (see Fig. 2.a). After the connection establishment, when channels. In RRC Idle the terminal monitors PCH to detect the mobile enters the RRC-connected mode, three states are incoming calls. For Earthquake and Tsunami Warning System defined: RRC-Cell Shared, RRC-Cell Dedicated, and RRC- (EWTS) capable UEs, EWTS notifications are received, and GRA PCH (TS 43.051). system information changes are detected. Furthermore, the UE carries out neighboring cell measurements and possible cell re-selections. On the other hand, in RRC CONNECTED Basically each state differs in the information on mobile state unicast data can be transferred in both downlink and position and in the type of allocated channel. In particular, uplink, while also Discontinuous Reception and Transmission the mobile position can be known on cell or on GERAN (DTX/DRX) may be utilized. In this state, also network con- Registration Area (GRA) level (i.e, a specified set of cells anal- trolled handovers may be used with optional Network Assisted ogous to UTRAN Registration Area (URA) within UTRAN). Cell Change (NACC) to GERAN. The UE monitors PCH In RRC-Cell Dedicated state, one or more dedicated physical and EWTS similarly to RRC idle state, but additionally UE subchannels in the uplink and downlink are allocated to UE, also monitors control channels associated with the shared data which can have also one or more shared physical subchannels. channel in order to determine if data is scheduled for it. The The UE position is known by GERAN at cell level. In RRC- UE provides channel quality feedback to the network,performs Cell Shared, the mobile location is known at cell level, but, needed neighboring cell measurements and reports the mea- unlike RRC-Cell Dedicated state, it has no allocated physical surements. For more details see TS 33.331 and [5]. channels. In RRC-GRA PCH state, no physical subchannel is allocated to UE and its location is known at GRA level. In this state, the UE monitors the paging channel and initiates III.RRCSTATE TRANSITION EVALUATION a GRA update procedure upon GRA change. Any activity leads to a transition either to the RRC-Cell Shared or RRC- To the best of our knowledge, past literature mostly fo- Cell Dedicated state. There is a complex interaction between cused on analytical modeling and simulation studies of RRC. RRC and RLC/MAC layer for radio resource management, pri- Previously published work aimed primarily on testing RRC marily due to legacy reasons. The MAC state machine has four management policies with regard to blocking and dropping states. In MAC-Idle state no shared or dedicated basic physical rates as well as evaluating the overall system capacity and subchannel are allocated. RRC can be in RRC-Cell Shared, investigating the energy consumption of mobile device [7]-[9]. RRC-GRA PCH state, or RRC-Idle mode. In the MAC- RRC state transitions can be modeled in an analytical way Shared state, at least one TBF is allocated, but no dedicated using a discrete time Markov chain; see for example [7]. The physical subchannels are allocated. The corresponding RRC transition probabilities are calculated as the probabilities that state is RRC-Cell Shared. In MAC-DTM state, the Mobile the idle times are longer than the timers used. Of course, the Station (MS) has one or more dedicated physical subchannels use of any analytical model presents a number of obvious limi- allocated and one or more shared physical subchannels. The tations and such model is used mainly for energy consumption corresponding RRC state is RRC-Cell Dedicated. In MAC- studies. Dedicated state, the MS uses a dedicated physical subchan- The popular ns-2 network simulator (see nel, but no shared physical subchannels. RRC is in RRC- www.isi.edu/nsnam/ns/) has also been employed in simulation Cell Dedicated state. As a simple rule, MAC layer deals with studies (see [8]). According to the authors, four generic states assigning, maintaining, and releasing of shared channels, while are common among the different packet switched cellular RRC deals with the dedicated channels. air interfaces (GRPS, EDGE, WCDMA, CDMA2000). Each technology is modeled by mapping its own RRC states to In Rel. 6, the RRC is mostly unchanged; the major im- the generic ones. The states are differentiated according to provement in GERAN is introduced in the physical layer with the available bandwidth, defined transitions, latencies of the the Flexible Layer One (FLO) (see TR 45.902), that allows transitions and delays. Specific channels features are not physical layer configuration along call setup. This change was modeled, and both uplink and the downlink channels are needed to support QoS in IP Multimedia Subsystem (IMS), for considered reliable. example, for future real time multimedia traffic: pre-existing CARMA (Comparative analysis of radio resource manage- GERAN radio bearers were dedicated for a given service and, ment strategies) [9] is another discrete event simulator used thus, were not enough flexible to support new ones, differently for studying the performances of different RRC management from UTRAN. FLO is a further step towards the process of strategies. CARMA can simulate only Packet Switched (PS) unifying GERAN to UTRAN, to enable seamless provision of services, the downlink channel and the various timeouts and services over the two radio access technologies. setup latencies can be changed in a convenient way. 4

A. RRC State Transitions in Practice Then, using 3G3T, sequences of packets were sent from one interface to other. Thus, our active measurement traffic While standards, in general, specify RRC operations, they patterns were injected into real operational cellular networks. are not sufficient to completely describe, in practice, the behav- It is important to note that in this case our focus was more ior of real networks: algorithms and the general methodology on the hidden technical properties of 3GPP networks than that regulate RRC state transitions are not explicitly defined by user behavior. Combining a study of typical user behavior any standard, even if most networks use the same procedures in 3GPP networks with the knowledge attained in this work and are customized through a range of - non disclosed - would be a natural next step. The typical used test scenario parameter settings according to specific design choices and is illustrated in Fig. 3. characteristics. The correct setting of these parameters becomes valuable FIGURE 3 for operators interested in their concurrent settings; researchers could use them for a realistic settings of analytical model or We now briefly describe the 3G3T working behavior using simulation tool. On the other hand, malicious users can also an example. use these values to overload the paging channel or to initiate Suppose we start with an UMTS connection in Cell DCH DCH starvation attacks. state and we want to test the respective timer in the downlink. Herein, we investigate on these, “non standard”, aspects 3G3T injects a packet sequence with slowly increasing inter- of RRC: the mechanisms regulating state transitions, the departure times (IDT), from the wired interface, through the parameters used by operators in real networks, the compliance Internet, to the UMTS interface. Fig. 2.b reports the diagram to standards of commercially available networks, and the for UMTS only state transition, to orient the reader across the interaction between user traffic and RRC mechanisms. For various transitions. As soon as the IDT increases, the inactivity this scope, we use the 3G3T, introduced in [10], [2], able period exceeds the release timer threshold, TDCH,off , config- to measure traffic over cellular networks and to discover ured by the operator, and a transition to Cell FACH state is RRC state transition parameters using active measurements carried out. The state transition is observed as a change in techniques. packet delay, since the average downlink delay in Cell DCH Through the description of 3G3T, we will guide the reader is significantly lower than in Cell FACH. The IDT of the first to the practical aspects involved in RRC operations, showing packet experiencing a larger delay is, on first approximation, how lower layers’ networks procedure, usually hidden to upper the value of the idle timeout that triggered the transition. layers, interact with user traffic. 3G3T can determine the Delving a bit deeper in 3G3T operation, we note that values of idle timers, RRC procedures’ delays, and can infer connections in Cell DCH state can use different Spreading the dynamics of transitions. The tool is based on the different Factor (SF) values (i.e., the ratio of the chips to the baseband delay characteristics in RRC states. For instance: high band- information rate) according to user bandwidth requirements. width/battery consumption states are characterized by lower Lower SFs correspond to higher available bandwidth and lower delays whereas channels with higher delays typically are as- delays; higher SFs correspond to lower bandwidth and higher sociated with lower bandwidth/power states. 3G3T exploits the delays. Therefore, in the Cell DCH state, packets experience fact that RRC state transitions are directly coupled with traffic different delays based on the used SF. The choice of SF can generated by UE, without channel preemption (due to higher be regulated by inactivity timers. priority traffic, for example) or cell reselection (due to ter- We tested four commercial HSPA networks, in three differ- minal mobility, for instance). Thus, we developed algorithms ent European countries; hereafter most remarkable results are to generate traffic patterns able to trigger state transitions, reported, to clarify the RRC mechanisms understanding. Along based on the following principles. Typically, transitions from the text, we do not distinguish results among the networks, resource hungry to less resource hungry states are triggered since their behavior is similar and just the numeric values of by an idle timeout, measuring the time from the last trans- the parameters change. We repeated the experiments with the mitted or received packet. State transitions from Cell DCH to new version of 3G3T tool, and, after six months, the network Cell FACH or from Cell FACH to Cell,URA PCH states are performances were substantially unchanged. good examples. Thus, the usage of resource hungry states is Fig. 4.a illustrates the one-way delay, d, for the downlink correlated with user activity. Since our earlier work [2], 3G3T channel as a function of the packet IDT for a real operational was evolved, based on what we learned from our experience. network, as measured by 3G3T. The figure clearly shows the It was reprogrammed according to a new modular design, transition from Cell DCH to Cell FACH in the downlink that allows us, for example, to exploit 3GPP TS 27.007 AT direction, effectively visualizing state transition dynamic, commands for automatically testing our inference on network values of the inactivity timer, and the procedure delays. status and then to identify if the network behavior differs from the one expected. FIGURE 4 With respect to the measurements reported in the rest of this paper, three identical setups were used in three different The value of the inactivity timer triggering the RRC tran- countries: a laptop with two network interfaces, one connected sition (as can be estimated from Fig. 4.a) is 2.6 s. RRC to public Internet via wired connection (Ethernet) and another transitions are not instantaneous, therefore a packet that ar- using a mobile packet data access card (HSPA), as in Fig. 3. rives during the state transition has to be buffered until 5 the Cell DCH related channels are torn down. Then, the Cell FACH state. Such dynamics suggests that the idle buffered packets are retransmitted as soon as the transition timer is started after we enter the Cell FACH state, and the is successfully completed and Cell FACH channels are set up nearly 6 seconds of overlapping account for the idle mode again. In short, during each measurement run, one packet that to Cell DCH transition delay, Cell DCH release timeout arrives during the transition is delayed more than the rest of and Cell DCH to Cell FACH transition delay. This behavior the packets. This packet contributes, together with a packet was first explained in [10]. Subsequently, we measured and from each repetition of the test, to form a “cloud” of packets reported similar behavior in three different UMTS networks (see the blue circle in Fig. 4.a). The average delay of these across three European countries (Italy, Austria, and Finland) packets minus the average delay in Cell FACH state can be [2]. Due to license issues with NemoOutdoor, the absence of used as a good estimate of the procedure completion time. transitions to URA PCH states could not have been verified In the tested network, the measured release timer threshold, in all the networks, but their similar behavior suggests that TDCH,off , for uplink was very close to the one measured this is not an isolated implementation case. In Fig. 5, we for downlink, so it is straightforward to conclude that the report threshold values and the procedure completion times thresholds are the same for both uplink and downlink. We with the bounds of the 90% confidence interval. It is evident repeat measurements on a different network. Fig. 4.b shows a that all the values are similar among the three networks, similar dynamic, but a series of additional steps are present: the except for the threshold for the release of the Cell FACH first steps are related to a change of SF while the last one cor- state, tidle. responds to the Cell DCH to Cell FACH RRC state transition. This behavior was verified using the proprietary NemoOutdoor FIGURE 5 measurement tool (www.nemotechnologies.com). In particular, NemoOutdoor was used to extract the value of As a further consideration, we should note that RRC state the SF values corresponding to each group of packets shown transition mechanisms, designed to both optimize network in the figure. resource usage and save mobile device battery power, become So far, we have focused on the transition from Cell DCH useful in presence of considerable background traffic, due to to Cell FACH, which is triggered by timers. On the other network scanning or worms. In fact, in such cases, inactivity hand, the transition from Cell FACH to Cell DCH is triggered timers are ineffective since they are reset by unwanted traffic. based on the Buffer Occupancy (BO) level. The threshold was This phenomenon was experimentally verified in one of the measured by using a traffic generator to send 56 bytes long tested network in both [10] and [2]. Some networks operators UDP packets. The transition back to Cell DCH was reached filter unwanted traffic and RRC mechanisms are more effec- with a transmission rate of 7 packet/s. In terms of Layer 3 tive. throughput, the threshold was approximately 4 kbit/s. 3G3T can also be used to dynamically measure transitions IV. CONCLUSION from Cell FACH to states with lower power consumption (i.e., Cell PCH and URA PCH, as well as transition to idle We presented the evolution of RRC mechanisms in mobile mode). Fig. 4.c illustrates an example, from a real network, cellular networks, from earlier GPRS systems to the latest based on our measurements. If we attempt to interpret the 3GPP technology, mentioning the upcoming LTE. We de- state transitions based solely on standards and theoretical de- tailed the evolution of standards, the step towards the process scriptions, we may misinterpret results. In particular, after the of unifying the radio layers GERAN to UTRAN and the expiration of the inactivity timer, the connection would switch implications in RRC. Furthermore, we practically showed from Cell FACH to either Cell PCH or URA PCH. However, how real networks behave and the differences with respect when correlating 3G3T measurements with NemoOutdoor, to the standards, analyzing RRC state transitions for real we find that the network did not carry out such transitions, commercial networks. The work can be fruitfully exploited for but rather RRC transitions directly to idle mode. Such a understanding in depth RRC mechanisms, helping operators transition from connected to idle mode requires the release of and researchers to improve the management and the evolution all RRC connections; therefore, each subsequent sent packet of present cellular networks and to propose new advanced requires a state transition to Cell DCH. This is a time wasting services. Future work involves further developments of 3G3T procedure requiring three-way handshake signaling and the as well as measurements in LTE networks. reestablishment of all related channels, which leads to long overall delays. As we saw before, we can take advantage of ACKNOWLEDGEMENTS such delay differences to better understand We thank the anonymous reviewers for their comments and inner workings. The difference between this delay and the suggestions. They significantly improved this paper. This work mean Cell DCH delay can be used as a good estimate for was supported by COST Action IC0703 and by projects Celtic the procedure completion time. It is worth mentioning that Easy Wireless II (Finalnd) and PS-025 (Italy, Apulia Region). such a big delay is observed by each initial packet starting a connection when the network is in idle state. Fig. 4.c shows an IDT overlap in the range 63 ÷ 69s REFERENCES with packets experiencing alternating high and low delays. [1] K. Pentikousis, “In search of energy-efficient mobile networking,” IEEE We found that packets with low delays are transmitted in Commun. Mag., vol. 48, no. 1, pp. 95–103, Jan. 2010. 6

[2] P. H. J. Peral¨ a,¨ A. Barbuzzi, G. Boggia, and K. Pentikousis, “Theory and practice of RRC state transitions in UMTS networks,” in Proc. of IEEE GLOBECOM Workshops, Honolulu, HI, USA, Nov.-Dec. 2009. [3] J. Bannister, P. Mather, and S. Coope, Convergence Technologies for 3G Networks: IP, UMTS,EGPRS and ATM. John Wiley & Sons, 2004. [4] H. Toskala, HSDPA/HSUPA for UMTS. Chichester, West Sussex, England: John Wiley & Sons Ltd, 2006. [5] D. Beming, 3G Evolution - HSPA and LTE for . Burlington, MA, USA: Academic Press, 2008. [6] H. Garcia, GSM, GPRS and EDGE Performance: Evolution Toward 3G/UMTS. New York, NY, USA: Halsted Press, 2002. [7] J.-H. Yeh, C.-C. Lee, and J.-C. Chen, “Performance analysis of energy consumption in 3GPP networks,” in Wireless Telecommunications Sym- posium, 2004, May 2004, pp. 67–72. [8] A. Talukdar and M. Cudak, “Radio resource control protocol configura- tion for optimum web browsing,” in Vehicular Technology Conference, 2002. Proceedings. VTC 2002-Fall. 2002 IEEE 56th, vol. 3, 2002, pp. 1580–1584 vol.3. [9] C. Brosch and A. Mitschele-Thiel, “Adaptive radio resource manage- ment scheme for UMTS packet data,” Proc. 11th European Wireless Conference, Nicosia, Cyprus, Apr. 2005. [10] A. Barbuzzi, F. Ricciato, and G. Boggia, “Discovering parameter setting in 3G networks via active measurements,” IEEE Commun. Letters, vol. 12, no. 10, Oct. 2008. 7

Fig. 1. Temporal evolution of access capacities, radio resource control, and network topology in 3GPP standard releases.

TABLE I 3GPP RADIO RESOURCE CONTROL SPECIFICATIONS Required for Spec No. First Freeze Date Latest Freeze Date Document Title GSM Before Rel-4 TS 04.08 1990-01-25 2000-06-30 Mobile radio interface layer 3 specification GSM Before Rel-4 TS 03.64 1997-06-13 1999-12-17 General Packet Radio Service (GPRS); Overall description of the GPRS radio interface; Stage 2 GSM Before Rel-4 TS 04.60 1998-06-26 1999-12-17 General Packet Radio Service (GPRS); Mobile Station (MS) - Base Station System (BSS) interface; Radio Link Control / Medium Access Control (RLC/MAC) protocol GERAN-based systems TS 44.060 2001-04-06 2009-12-10 General Packet Radio Service (GPRS); Mobile Station (MS) - Base Station System (BSS) interface; Radio Link Control / Medium Access Control (RLC/MAC) protocol GERAN-based systems TS 43.051 2001-03-22 2009-12-10 GSM/EDGE Radio Access Network (GERAN) overall description; Stage 2 GERAN-based systems TR 45.902 2005-01-28 2009-12-10 Flexible Layer One (FLO) UTRAN-based systems TS 25.331 1999-10-13 2009-12-10 Radio Resource Control (RRC); Protocol specification E-UTRAN-based systems TS 36.331 2008-12-11 2009-12-10 E-UTRA Radio Resource Control (RRC); Protocol specification All technical specifications (TS) are available from www..org 8

Fig. 2. RRC state transition diagrams in 3GPP Rel. 8 (a) and UMTS (b). 9

Fig. 3. Schematic of the measurement methodology.

10 10 Cell_FACH Idle

1 1 delay [s]

0.1 delay [s]

0.01 0.1 0 2 4 6 8 10 12 14 16 10 20 30 40 50 60 70 80 90 100 IDT [s] IDT [s] (a) (b)

1 Cell_DCH Cell_FACH SF=32SF=8 SF=4 0.1 delay [s]

0.01 0 1 2 3 4 5 IDT [s] (c)

Fig. 4. Transitions from CELL DCH to Cell FACH for download in two different networks. In b) Transition from Cell FACH to less power consumption states. In c) Spreading Factor variation can be observed. 10

70 Net#1 Net#2 60 Net#3

50

40 [s] 30

20

10

0 TDCH, off TIdle DDCH, DL DDCH, UL

Fig. 5. Extracted threshold and procedure delays for the three tested networks.