3GPP LTE-Assisted Wi-Fi-Direct: Trial Implementation of Live D2D Technology
2015; Electronics and Telecommunications Research Institute; Volume: 37; Issue: 5 Linguagem: Inglês
10.4218/etrij.15.2415.0003
ISSN2233-7326
AutoresAlexander Pyattaev, Jiří Hošek, Kerstin Johnsson, Radko Krkos, Mikhail Gerasimenko, Pavel Mašek, Aleksandr Ometov, Sergey Andreev, Jakub Sedy, Vít Novotný, Yevgeni Koucheryavy,
Tópico(s)Power Line Communications and Noise
ResumoETRI JournalVolume 37, Issue 5 p. 877-887 ArticleFree Access 3GPP LTE-Assisted Wi-Fi-Direct: Trial Implementation of Live D2D Technology Alexander Pyattaev, Alexander Pyattaev alexander.pyattaev@tut.fi Search for more papers by this authorJiri Hosek, Corresponding Author Jiri Hosek hosek@feec.vutbr.cz Corresponding Authorhosek@feec.vutbr.czSearch for more papers by this authorKerstin Johnsson, Kerstin Johnsson kerstin.johnsson@intel.com Search for more papers by this authorRadko Krkos, Radko Krkos krkos@phd.feec.vutbr.cz Search for more papers by this authorMikhail Gerasimenko, Mikhail Gerasimenko mikhail.gerasimenko@tut.fi Search for more papers by this authorPavel Masek, Pavel Masek xmasek12@phd.feec.vutbr.cz Search for more papers by this authorAleksandr Ometov, Aleksandr Ometov aleksandr.ometov@tut.fi Search for more papers by this authorSergey Andreev, Sergey Andreev sergey.andreev@tut.fi Search for more papers by this authorJakub Sedy, Jakub Sedy jakub.sedy@phd.feec.vutbr.cz Search for more papers by this authorVit Novotny, Vit Novotny novotnyv@feec.vutbr.cz Search for more papers by this authorYevgeni Koucheryavy, Yevgeni Koucheryavy yk@cs.tut.fi Search for more papers by this author Alexander Pyattaev, Alexander Pyattaev alexander.pyattaev@tut.fi Search for more papers by this authorJiri Hosek, Corresponding Author Jiri Hosek hosek@feec.vutbr.cz Corresponding Authorhosek@feec.vutbr.czSearch for more papers by this authorKerstin Johnsson, Kerstin Johnsson kerstin.johnsson@intel.com Search for more papers by this authorRadko Krkos, Radko Krkos krkos@phd.feec.vutbr.cz Search for more papers by this authorMikhail Gerasimenko, Mikhail Gerasimenko mikhail.gerasimenko@tut.fi Search for more papers by this authorPavel Masek, Pavel Masek xmasek12@phd.feec.vutbr.cz Search for more papers by this authorAleksandr Ometov, Aleksandr Ometov aleksandr.ometov@tut.fi Search for more papers by this authorSergey Andreev, Sergey Andreev sergey.andreev@tut.fi Search for more papers by this authorJakub Sedy, Jakub Sedy jakub.sedy@phd.feec.vutbr.cz Search for more papers by this authorVit Novotny, Vit Novotny novotnyv@feec.vutbr.cz Search for more papers by this authorYevgeni Koucheryavy, Yevgeni Koucheryavy yk@cs.tut.fi Search for more papers by this author First published: 01 October 2015 https://doi.org/10.4218/etrij.15.2415.0003Citations: 32 Alexander Pyattaev (alexander.pyattaev@tut.fi), Mikhail Gerasimenko (mikhail.gerasimenko@tut.fi), Aleksandr Ometov (aleksandr.ometov@tut.fi), Sergey Andreev (sergey.andreev@tut.fi), and Yevgeni Koucheryavy (yk@cs.tut.fi) are with the Department of Electronics and Communications Engineering, Tampere University of Technology, Finland. Jiri Hosek (corresponding author, hosek@feec.vutbr.cz), Radko Krkos (krkos@phd.feec.vutbr.cz), Pavel Masek (xmasek12@phd.feec.vutbr.cz), Jakub Sedy (jakub.sedy@phd.feec.vutbr.cz), and Vit Novotny (novotnyv@feec.vutbr.cz) are with the Department of Telecommunications, Brno University of Technology, Czech Republic. Kerstin Johnsson (kerstin.johnsson@intel.com) is with the Wireless Communication Laboratory, Intel Corporation, Santa Clara, CA, USA. The research described in this paper was financed by the National Sustainability Program under grant LO1401. For the research, infrastructure of the SIX Center was used. AboutSectionsPDF ToolsRequest permissionExport citationAdd to favoritesTrack citation ShareShare Give accessShare full text accessShare full-text accessPlease review our Terms and Conditions of Use and check box below to share full-text version of article.I have read and accept the Wiley Online Library Terms and Conditions of UseShareable LinkUse the link below to share a full-text version of this article with your friends and colleagues. Learn more.Copy URL Share a linkShare onFacebookTwitterLinkedInRedditWechat Abstract This paper is a first-hand summary on our comprehensive live trial of cellular-assisted device-to-device (D2D) communications currently being ratified by the standards community for next-generation mobile broadband networks. In our test implementation, we employ a full-featured 3GPP LTE network deployment and augment it with all necessary support to provide real-time D2D connectivity over emerging Wi-Fi-Direct (WFD) technology. As a result, our LTE-assisted WFD D2D system enjoys the required flexibility while meeting the existing standards in every feasible detail. Further, this paper provides an account on the extensive measurement campaign conducted with our implementation. The resulting real-world measurements from this campaign quantify the numerical effects of D2D functionality on the resultant system performance. Consequently, they shed light on the general applicability of LTE-assisted WFD solutions and associated operational ranges. I. Introduction Device-to-device (D2D) communications technology as part of the 3rd Generation Partnership Project (3GPP) Long-Term Evolution (LTE) Release 12 specification [1] enables novel unprecedented opportunities for next-generation peer-to-peer (P2P) and location-based applications and services [2]. Most of these new opportunities emerge from discovering and exploiting user proximity, which is provided by virtue of location services in contemporary mobile broadband networks [3]. Together with improvements in network capacity, data delivery latency, individual user throughput, and energy efficiency, D2D also brings along attractive new business cases for network and service providers. With our preliminary research [4]–[5], we have identified numerous benefits that could become available if a coordinated, network-assisted D2D technology is deployed by network operators. On the other hand, introducing D2D technology within today's network infrastructure poses a number of challenges and requires updates to the current longstanding cellular architecture. Therefore, to conduct a comprehensive study and reveal the practical promises of D2D communications, we have designed a trial development and deployment program. Our trial was aimed at demonstrating how the direct connectivity paradigm could be seamlessly integrated into a real-world, operator-grade cellular network with minimal modifications and overheads, as well as within a reasonable time frame. Our secondary goal was to quantify gains that could be achieved by a fully-functional, operator-supported D2D system. To complete these challenging objectives, we have defined a deployment time constraint of one month and assigned a team of six qualified engineers for the said implementation. As a basis for our trial, we have utilized the experimental LTE network of Brno University of Technology (BUT), Czech Republic, which supports most of the functionality expected of LTE Release 10 systems (see Fig. 1). During the trial, we upgraded the LTE network of BUT with our own implementation of Proximity Services (ProSe) functionality as envisioned by the 3GPP specifications [6]–[7]. This has allowed us to perform live D2D integration trials, along with corresponding performance evaluations, and we systematically report all of our most important findings in this paper. Figure 1Open in figure viewerPowerPoint Part of experimental LTE network at BUT, Czech Republic. The rest of this paper is organized as follows. In Section II, a review of the D2D-enabling standards and technologies is conducted. Further, we outline the envisioned D2D system design and operation in Section III. Section IV, in turn, elaborates on the resultant practical implications of our vision. Finally, in Section V, we discuss our performance evaluation summary of the developed D2D system together with some insights into the realistic aspects of D2D operation and usage. II. Overview of Technologies and Standards Enabling D2D Communications The range of applications for D2D connectivity in cellular networks is truly wide [8]. Prospective D2D use cases vary from local voice and data services (offloading calls between proximate users) to context-aware applications, augmented reality, and public safety. 1. Current D2D Technology Candidates Presently, the most widely adopted technology candidates for direct connectivity are Bluetooth and Wi-Fi, which operate in the 2.4 GHz unlicensed spectrum. Other popular wireless networks, such as those proposed by IrDA and ZigBee alliances, are not targeted for generic D2D communications [8] by their design. As a result, Bluetooth is mostly used today for personal area networking, whereas Wi-Fi remains popular whenever high-rate user device connectivity is required. However, employing conventional Wi-Fi and Bluetooth for D2D communications is rather cumbersome and does not guarantee any quality of service (QoS) levels. Addressing the usability problems to some extent, a Wi-Fi-Direct (WFD) protocol [9] has recently been introduced. WFD enables direct communication possibilities without the need for Wi-Fi infrastructure, and does so with minimal user interaction; thus, this essentially allows a device to act as a network controller (a.k.a. "group owner"). WFD also provides a way to set up multiple networks at the same time, thus enabling multiple devices to communicate IP data directly between each other [10]. 2. Network-Assisted D2D Communications Since market penetration of WFD is already high and is only expected to grow further, this technology becomes increasingly attractive for envisioned cellular-assisted D2D communications. It is evident that for future D2D systems to benefit consistently from WFD, it is necessary to minimize the signaling overheads of neighbor discovery [8], [10]. This signaling load can be reduced with the help of an authoritative entity (for example, a mobile network operator), where several key functions of D2D connectivity would be managed by the cellular network infrastructure [4]. Consequently, current mobile broadband networks may provide proximal devices with efficient mechanisms to establish and maintain a D2D connection. This is required, first of all, to enable at least some form of fallback, which is a crucial consideration for reliable proximity-based applications. However, by utilizing the centralized control function and, in particular, user location and signal level monitoring mechanisms of LTE networks, it is also possible to coordinate a D2D network, improving the overall system efficiency. In contrast to the "loosely-controlled" D2D mode discussed above, a fully-controlled D2D operation in licensed cellular bands is currently anticipated with the advent of LTE-Direct [11], which represents a system under the control of a cellular network operator. LTE-Direct is presently under heavy development, but the real-world deployments of this technology are not expected earlier than in several years from now due to many associated technical and market challenges [11]. As expected, there are several major complications in running D2D over licensed bands, such as cumbersome interference management (which is attractively simplified in WFD), additional transceiver design and standardization efforts (in WFD, stock Wi-Fi transceivers are used), and multiple other issues. Therefore, the main focus today remains on D2D communications over unlicensed bands, which is primarily achieved with LTE-assisted WFD. 3. D2D Standardization Activities The key enabling technologies for D2D communications have been around for years, since practically any IP-ready mobile device is tentatively capable of direct connectivity. However, the development of the necessary supporting standards and user interfaces has woefully lagged behind any developments in related technologies, with the first D2D activity started by 3GPP in 2012 within the framework of ProSe. 3GPP TR 22.803 [6] is known as the initial document to specify what exactly is to be understood by the term "proximity-based services," and D2D naturally fits into this category. However, 3GPP ProSe specifications and associated work items cover, in fact, a much broader area. They address the phenomena in both society and economy that drive the need for proximity-based communications, including such examples as social networking; disaster relief and emergency service operations; advertising; and so on [8], [12]. As a result of this work, the decision was made that ProSe functionality was to become a part of future LTE releases. Consequently, the technical side of providing ProSe mechanisms demanded attention. In particular, a requirement has emerged on establishing proximity in an efficient way without revealing personal user information. Indeed, proactively requesting any content via short-range radio links discloses the type of the desired content, whereas broadcasting content advertisements discloses what is available to share. Either way, users of a proximity-based discovery service, both content providers and consumers, may prefer to remain anonymous or impose their customized policies on a set of targeted peers. With decentralized solutions, such as those offered by WFD, meeting the above requirements is clearly impossible. This, in turn, has led to research on a potential Evolved Packet Core (EPC)-level discovery procedure, where a 3GPP network would act as a trusted intermediary and implement all of the necessary policies on behalf of users. The report TR 33.833 [13] quantifies specific goals, which have been respectively targeted by 3GPP. Currently, since the overall vision of how the ProSe function is to be implemented has already taken shape, follow-up work has begun on support infrastructure, such as billing [14]. As of 2015, most of the architectural progress on ProSe and D2D communications is summarized in the TS 23.303 document [15]. It is likely, however, that additional amendments will be made when activities on the technical side of LTE unlicensed [16] commence. Today, some of the related ideas on potential license-assisted communication (RP-140770) and on LTE-based short-range radio within licensed bands [17] have already been documented. However, aside from their in-house efforts on short-range radio, 3GPP also supports alternative non-3GPP radio technologies, including WFD, for ProSe radio communication. The integration between 3GPP and Wi-Fi solutions has been a long-standing effort, with specifications TS 23.234 [18], TS 23.327 [19], and TS 23.402 [20] outlining how LTE devices that are connected over non-3GPP access technology could still receive access to all of the 3GPP services. Following this lengthy integration effort, all current ProSe architectures support the use of IEEE 802.11 family, as a link layer for most of the ProSe functions, with the exception of public safety services [15]. In our trial implementation, we have been relying on 3GPP specifications as much as possible within the limitations of the target deployment. In the following section, we describe in detail what exactly has been developed and what obstacles have been encountered. III. LTE-Assisted WFD D2D System Implementation 1. D2D Trial: Architectural Considerations Our envisioned implementation of the generic D2D system concept offered by the standards has naturally met a number of deployment challenges that made us deviate from the reference solutions. Current architectural considerations of LTE networks preclude us from deploying the network assistance functions in the way that would have been the most "natural" from an engineering standpoint; thus, we had to adopt several clever workarounds to develop a workable system with today's technology. In what follows, we review the key decisions made in each step of the deployment process and explain our motivation behind them. The experimental cellular network setup installed at the Department of Telecommunications, BUT (see Fig. 2) is a complete commercial-grade implementation of all the crucial subsystems comprising contemporary 4G mobile networks. Figure 2Open in figure viewerPowerPoint System topology of experimental 4G cellular network deployed at BUT. Our trial deployment is configured to provide the necessary packet-switched data access services and their derivatives, such as VoIP communications over converged LTE and Wi-Fi radio access infrastructure. The EPC is dimensioned to enable high data rate services with appropriate QoS provisions, as well as to support up to 100,000 concurrently served users. For voice and video calls, the switching capabilities are implemented by employing the high capacity IP Multimedia Subsystem and its related components, for both mobile and fixed access users, as well as for connectivity to external telephone networks and teleconferencing systems. The general goal behind BUT's LTE test network deployment has been achieving the synergy of a complete and customizable experimental mobile network that allows for rapid implementation and prototyping of novel concepts and technologies, such as the D2D communications paradigm discussed at length in this paper. We thus effectively used this asset to showcase that the LTE-assisted WFD technology has matured enough for an example implementation in a commercial-grade LTE/Wi-Fi network. However, several LTE mechanisms expected by the ultimate network-assisted D2D architecture were not available at the time of our early implementation. This included the evolved Serving Mobile Location Centre (e-SMLC) server, for which an alternative interface to obtain device location information directly from a Mobile Management Entity (MME) has been developed as a substitute. In addition, the D2D server functionality has been implemented as a virtualized appliance, whereas the final commercial operator-grade implementation should preferably use a more robust technology. Further design choices are explained below. 2. Impact of Radio Access Network on D2D Communications A radio access network (RAN) is crucial for any kind of service in a mobile network. To this end, RAN is typically required to transport signaling and data. That being said, D2D communications are not particularly demanding with respect to RAN capacity, as D2D messages are only a few bytes in size, and ultimately D2D communications reduces the RAN load. The D2D signaling should, however, be prioritized to decrease service response times. Correspondingly, the impact of RAN loading on the observed service times has been measured extensively in our experiments (see Section IV). Other RAN functionality required for LTE-assisted D2D communications is positioning. Providing user location information is achieved via cooperation between eNodeBs and other EPC elements, and is described at length in Section III-6. Notably, neither special configuration nor non-standard service is needed from the RAN side for the D2D system to function. A deployed RAN is part of our test-bed installation. Generally, commercial-grade equipment is used, but the RAN itself is not a part of the radio access subsystem of a mobile network operator. Hence, we have complete control over the RAN loading conditions. Finally, as our RAN solution is deployed in the real world, external interference is present. 3. Providing Unrestricted IP Connectivity One of the key requirements for any direct user connectivity is the ability to communicate between devices without any intermediary hosts operating at the transport layer or above. Typically, in cellular networks, the associated devices acquire their IP addresses from private ranges (for example, 192.168.0.0/16). As long as a D2D link is set up within a single operator's network, this does not impose any constraints. However, the effective firewall policies deployed in the core network, which deny direct access between user devices to enhance security, actually cause difficulties. The original purpose of the aforementioned firewall policies could have been prevention of undesired incoming connections and P2P data transfer between different mobile network users. As a result, they deny any and all P2P connectivity over cellular networks [21]. For the purposes of our network-assisted D2D trial, we needed to circumvent the firewall by making the D2D server open direct communication paths for selected connections whenever necessary. In our case, implementing such functionality has been fairly straightforward, since the D2D server can reconfigure the firewall on a per-connection basis. In our implementation, the firewall is located inside the Unified Gateway (UGW) entity; logically composed of a Serving Gateway (SGW) and Packet Data Network Gateway. 4. Communication between D2D Server and Users By design, D2D network assistance relies on a network's ability to communicate with those user devices that are engaged in direct connectivity with the network. This inherent ability must be augmented by an efficient means of initiating such communications. For example, the said connectivity could be straightforwardly enabled with a session initiation protocol, but this would require an active radio bearer. Having an active LTE radio bearer for only several packets is naturally not conducive to system efficiency; hence, this is to be avoided. In LTE, there are multiple ways to transfer short messages between a network and those individual users of the network that do not require a dedicated bearer setup, such as non-access stratum (NAS) signaling, which is typically used to set up the bearers themselves. As a result, the practical deployment of the D2D functionality requires an implementation where the D2D server would be positioned outside of the system core as a conventional IP service, but with a capability to access certain core network functions (see Fig. 3). While this may not be the optimal solution in final commercial deployments, it enabled us to move forward promptly with regard to D2D system implementation. Although the proposed location of the D2D server does not follow the 3GPP guidelines exactly, we believe that our modification does not produce any negative impact with respect to latency, as the connection between the SMLC and the D2D server is implemented via a tunnel over a fiber channel. Figure 3Open in figure viewerPowerPoint Comparison between (a) hypothetical/standardized and (b) our deployed implementations of D2D system in 4G mobile networks. 5. Integrating Location Services For the network-assisted direct connectivity to operate efficiently, the D2D server needs regular updates on the current locations of users. In LTE, such information is conventionally aggregated by the e-SMLC entity to be then made available for the user devices via Secure User Plane Location (SUPL) bearers. A copy of the location information in question is typically stored inside the MME for its internal usage. Whereas the exact means of how this information is obtained may vary, the general procedure is such that a phone's GPS location will be used if available. Alternatively, the positioning reference symbols within the LTE frame will be used to pinpoint the location of user equipment (UE) through triangulation. Either way, the coordinates are obtained in line with the standard techniques outlined by 3GPP and are enabled in most modern equipment (post Release 9). As a result, enabling location-based services is not a major challenge in contemporary LTE networks, and most of the time such functionality is already provided by the operators for the mobile devices to use. In the considered proximal scenario, the D2D server accesses the location information on behalf of the UE, and then draws conclusions on whether other UEs are sufficiently close to initiate direct communication. An example of such decision logic is presented in Fig. 4. With the help from location services, the UE can thus power on its radio only when the intended contact is in proximity, hence saving battery and network resources. The specific signaling used in the trial is further discussed in Section IV. Naturally, one would need to select the thresholds triggering various decisions, but those are largely hardware-specific and have not been the core subject of this study. Figure 4Open in figure viewerPowerPoint Switching logic between LTE and proximal D2D connections, as mobile UE is moving along its route and meeting static D2D peer. 6. Linking D2D Server and Core Network Components Recall that the location information is typically made available in LTE only for MME and the UEs themselves. Similarly, the interfaces to control the firewall policies at the UGW are also core-local. In our implementation, we had to create a secure connection to the core to allow the D2D server to communicate with the MME and UGW, as to extract the location information and configure firewall policies via the maintenance interfaces, which is, of course, not the preferred final solution to go with. In reality, one would require enabling the D2D server to connect to the MME/SMLC and UGW via an efficient and secure application programming interface (API), without exposing the entire management console. Due to the time limitations imposed by the purposes of this trial and to the vendor-specific nature of the firmware, we have left the development of such API for future work. 7. D2D Connection Control Efficient control of D2D connectivity between UEs deserves a separate dedicated research, which can be generally split into signaling and executive components. The signaling may be handled by a service that is running as a service on the mobile device, whereas the executive part is essentially integrated into the kernel drivers of the operating system (OS) of the UE. While implementation of the signaling part as an application is rather trivial for most of today's mobile phones, actually forcing control over Wi-Fi and cellular connections (together with the associated routing table) requires significant modifications to the permission levels of the UE. In practice, for most platforms (Android, the majority of Linux-based systems, and so on), this means employing custom-built firmware for the phone, or obtaining the administrative privileges by virtue of hacks and exploits. For closed platforms, such as iOS and Windows Phone, these solutions are nearly impossible without cooperation from the platform vendors. IV. Performance Evaluation of Implemented D2D System The primary goals of our performance evaluation are as follows: ■ Indicating bottlenecks that could potentially hinder the adoption of D2D connectivity in future wireless technologies. ■ Establishing appropriate performance bounds and limitations for D2D technology, as well as outlining what services could be most suited for direct communications in contemporary and near-future markets. To achieve these diverse goals, we follow the measurement procedures as outlined in the remainder of this section. 1. Measurement Methodology It is important to note that in this work (by contrast to numerous past publications) we are not interested in the performance of the D2D link itself, since that would largely depend on the current channel and user contention levels. In this study, we only concentrate on assessing signaling performance and network assistance logic, as the latter can be reliably measured in our controlled trial environment. On the other hand, such metrics as D2D throughput are extremely difficult to assess conclusively in practice due to the high variability in wireless environments. Based on the above reasoning, the single most important parameter of D2D signaling is the connection setup time. This latency is crucial, as lengthy connection setup times may delay the transfer of the flow to an alternative radio link, thus affecting other parameters such as energy efficiency and user experience. Due to small message sizes, in the order of a few tens of bytes, other QoS requirements on the D2D signaling are trivial and can be provided by any access network; therefore, we are not considering these as key performance indicators for this technology. Other aspects of D2D link performance, which are not directly related to mobile network infrastructure, do not affect service setup response times and, as such, are out of the scope of this work. The connection setup time may in turn be decomposed into several components, as described below. First, by looking at the proposed D2D protocol signaling illustrated in Fig. 5, we learn that before any WFD connection is actually set up, there are several important actions to be taken from the network side to establish user proximity. Therefore, we would like to decompose the signaling procedure in question into several distinct stages. In stage 1, the responsiveness of the system is not of particular interest, since the procedures taking effect there are constrained primarily by the human user input. Figure 5Open in figure viewerPowerPoint Considered D2D protocol signaling including all five stages required for proximity discovery, D2D link establishment, and termination. However, once stage 1 is completed and the proximity is established, the actual D2D signaling is triggered. Ultimately, we would like for stages 2 (D2D link negotiation), 3 (link setup), and 4 (actual flow switching) to occur as quickly as possible to maximize the effective time during which we can take advantage of the D2D link, especially if the communicating peers are highly mobile. These stages also constitute natural measurement checkpoints, where the resultant system agility would straightforwardly depend on how long it takes to transit from one checkpoint to another. Therefore, we further commit to measure an aggregate latency at each of the above state transitions. To provide the best available accuracy, we have additionally decomposed the considered D2D protocol into individual messages and performed our measurements on a statistically large sample set. The latency has been meas
Referência(s)