A VENDOR INDEPENDENT SERVICE ARCHITECTURE FOR NEXT GENERATION MOBILE DEVICES
ABSTRACT
Mobile devices have become versatile in a very short period of time. Services extend its versatility and many anticipate this to be the next major area of growth. Yet there are several issues ranging from heterogeneous networks, differences in protocols, device limitations etc. In this paper, we identify such issues and propose a vendor independent architecture for service delivery that takes these into consideration. The architecture relies on a WLAN and Service Controller that provides access to services, using the client-directory-server model (i.e. SLP) of service delivery. Details of the architecture and interaction protocols are provided. There are several advantages of such architecture. It decouples the service from the cellular provider, allowing third parties to operate services. It is easily deployable and scalable. It provides direct connection to the directory server allowing fast service discovery. It also allows for full integration with cellular networks.
Keywords: cellular services, cellular service architecture, mobile architecture, mobile service architecture, next generation cellular services, next generation mobile phone.
INTRODUCTION
The cell phone has emerged as a versatile device in a relatively short span of time. When first introduced cell phones were bulky and barely functional while current generation phones are small and packed with features rivaling those of desktops. This has been made possible by advances in micro-electronics, battery technologies, as well as in cellular transmission technologies. In their inception, cell phones were analog based and could support only voice transmission. This quickly changed by the “2.5” generation. Use of packet switching and CDMA/TDMA technologies allowed for increased channel capacity leading to speeds of upto 100 Kbps and “SMS” (Simple Message Service) offerings. This extended to internet service in the 3rd Generation (Mangalaraj and Amaravadi 2008). While current generation phones offer SMS, ringtones, navigation, weather and internet, there is tremendous potential for extending the useability of these devices. Maps, directions, tourist information, patient monitoring (Chakravorty 2006), spontaneous language translation (Mattern et al. 2002) are only a few of the myriads of applications that are possible. It is these type of services that we refer to as services rather than internet access which would now be considered a “utility service” (Mangalaraj and Amaravadi 2008). Some applications are needed when the subscriber is travelling. In such cases the service should continue uninterrupted even when he/she moves outside a communication “cell.” Service provisioning without a formal structure/organization is an invitation to chaos but developing a universal service infrastructure is a formidable challenge. The overall problem may be characterized as superimposing or juxtaposing a service architecture on top of existing cellular infrastructure such that services could be provided in a vendor, network and location independent fashion. The architecture has to be interoperable with 2.5 G+ networks as well as with CDMA/WCDMA, GSM and the recently developed UMTS (Salkintzis et al. 2002, Park et al. 2007). Further any solution has to address issues of security (Chalmers and Almeroth 2004; Patel and McCubben 2005; Zhu et al. 2006), authentication, authorization, and accounting (AAA) in addition to interoperability and mobility management. Thus there are complex technical challenges at every stage from physical delivery of service to dealing with issues such as interruptions and billing. In this paper we discuss such issues and elaborate on a vendor independent, scalable service architecture that addresses some of the problems with existing approaches. But first we will review some relevant technologies.
RELEVANT TECHNOLOGIES
To provide a background for architectural discussion, in this section we review cellular technologies such as TDMA/CDMA, WLAN/WiFi, Mobile IP as well as service delivery technologies such as Jini, UPnP etc. Technologies in the former category are related to delivery infrastructure while those in the latter category serve as discovery/interaction methods.
TDMA /CDMA
Time Division Multiple Access (TDMA) and Code Division Multiple Access (CDMA) are technologies that allow multiple users to share the same cellular bandwidth. In TDMA, users are allocated different time slots and can re-use the same frequency whereas in CDMA, subscriber signals are encoded with a unique code that allows their signals to be interleaved with those of other subscribers during transmission. TDMA is widely used in GSM (Global System for Mobile Communication) by companies such as AT & T and TMobile. Verizon uses CDMA technology. These network technologies could be augmented by WiFi, WLANs and Manets but switching/routing among these present many problems (Balasubramaniam et al. 2010; Cavalcanti et al. 2005; Tseng et al. 2003).
Wireless Local Area Network (WLAN)/WiFi
WLANs are used for wireless networking for local area networks and adhere to IEEE 802.11 a/b/g/n series of standards. They permit varying data transfer rates such as 11Mbps for 802.11b and 54Mbps for 802.11g (Wang and Refai 2005). Wi-Fi/WLAN uses unlicensed bands at 2.4 and 5 GHz and relies on two different basic coding techniques: the Direct Sequence Spread Spectrum (DSSS), which 11b and 11g devices use, and Orthogonal Frequency Division Multiplexing (OFDM), which 11a and 11g devices use for using the channel (Aime et al. 2007). They can be setup using two distinct network topologies with the infrastructure mode (access point is present for centralized coordination) using a virtual star topology, and the ad hoc mode (no centralized coordination) using a mesh topology (ibid). The range for 802.11 networks varies from 0.7 KM to 3 KM which is still less than that available in cellular networks. Wireless mesh networks are claimed to enhance coverage in large areas.
Mobile IP
The Internet Engineering Task Force (IETF) specified mechanisms for mobility management of cellular devices. While roaming, a mobile node (MN) maintains two IP addresses, a home address (HA) and a care of address (CoA) that reflects its current location. The CoA is obtained through a Foreign Agent (FA) that maintains the node’s current location (de Silva and Sirisena 2001). A location register referred to as Home Agent (HA), maintains the mapping of the MN home address to the CoA. When the node is roaming, the HA tunnels packets to the CoA/FA. When the MN transmits data, it sends data using its HA as the packet source. Implementation of Mobile IP in cellular networks creates problems due to the frequent movement of the MN which results in additional delays in the handoff process, causing packet loss and delayed delivery of data due to additional hop to the FA (Campbell et al 2002).
Jini
Java Intelligent Network Interface (Jini) is a service discovery and invocation protocol that allows devices in a network such as printers and servers to find each other (Seno et al, 2007). Jini technology utilizes a central Lookup server to store directory information (Ververidis and Polyzos, 2008). Service providers detect the lookup server by multicasting (i.e. broadcasting) service announcements on the local network. Once it detects it, the service provider registers service details with the lookup server. Clients wishing services send a multi-cast request on the network. The request is responded to by the lookup server which then sends the client a Java object to interface with the service (Meshkova et al. 2008). The client then accesses the service directly, using the service interface.
Service Location Protocol (SLP)
SLP is used for service discovery in various types of networks ranging from LAN to large enterprise-sized networks (Meshkova et al. 2008). In this protocol, there are User Agents (UA), Service Agents (SA) and optional Directory Agents (DA) (Ververidis and Polyzos, 2008). When DAs are present, the SAs register services to them by sending service advertisements. The DA sends an acknowledgement back and the service is registered. User Agents discover services on behalf of clients needing services. The UA sends a unicast query to the DA which responds with the service URL. When DA is not present, the UA multicasts (i.e. broadcasts) service queries to all SAs (Meshkova et al. 2008; Ververidis and Polyzos 2008) and the SA with the service will respond. The SA then provides the service. The availability of a directory allows UAs to search for services based on various service attributes.
Universal Plug and Play (UPnP)
Universal Plug and Play (UPnP) is a set of protocols for connecting different devices in home networks such as PCs, printers, Wi-Fi access points, smart appliances and mobile devices. In this case, there is no central registry. Devices providing services advertise these to control points (potential clients) on the network. Control points wishing to utilize a service access these discovery messages. Details about the service are placed in an XML file whose URL is provided in the discovery message. UPnP technology could be implemented in all IP-based networks, irrespective of the operating system or the type of physical networking media—either wired or wireless—which adds to its appeal (UPnPForum 2009). However, the absence of a central registry will lead to chaotic conditions and the inability to search for services.
Bluetooth
Bluetooth wireless technology (IEEE 802.15.1) is widely used in wireless personal area networks. Bluetooth’s primary purpose is to replace cables for computer peripherals such as mice, keyboards, joysticks, and printers using short-range wireless media (Z´aruba et al. 2001). It uses 2.4GHz unlicensed band and provides coverage of less than 10 meters with a data rate of about 3Mbps (Lau et al. 2009). The recently introduced Bluetooth 3.0 utilizes 802.11 radio and promises to have more than 24 Mbps transfer speeds (Bit et al. 2010). The distance is still a limitation for this method of service delivery.
Near Field Communication (NFC)
NFC is a short range wireless communication technology that is gaining popularity. NFC operates at a low frequency (13.56 MHz) and short range (10 cm) for data transfer rates of 424 Kbps (Ortiz 2006). NFC devices can be self-energized (i.e., active) or passive (Fischer 2009). When one device gets near another NFC device a magnetic coupling is established that enables transfer of energy/data. This distinguishes NFC from Bluetooth (Ortiz 2006). NFC based devices are expected to replace many contact based payment cards (Leavitt 2010) but may not be a viable solution for service delivery.
CRITICAL ISSUES IN SERVICE PROVISIONING
In this section, we review critical issues for designing an architecture to provision services. When a mobile device connects with a service entity, the device attempts to find from the entity what services are offered (if any). The entity will respond with a listing and the subscriber chooses a service. Once a service is selected, the service entity gives the device a service handle and delivery of can then begin. Thus important issues revolve around the service life cycle including selecting, discovering, offering and provisioning services as discussed below.
SERVICE SELECTION
Service selection is concerned with how services are selected, whether automatic or manual (Zhu et al. 2005). It is expected that in certain environments such as corporate campuses and metropolitan areas there will be dozens if not hundreds of services. Global service offerings could easily escalate this number up by a factor of a thousand. Due to resource constraints, manual methods are not feasible when there are offerings of this magnitude. Researchers have suggested different automated approaches based on optimization and attribute or user preference matching (Adomavicius and Tuzhilin 2005, Nishkam et al. 2005). Results are promising for small number of services, but may not scale up for large volumes. Kwon et al. (2010) in an experiment report a selection time of 0.02 sec for selecting 3 out of 20 services, using an algorithm that takes into account past experience. In large sets of services, selection times may be longer and issues similar to those in document retrieval such as relevance and recall are likely to occur. One solution is to list services under a classification scheme so that the number of services under each category is reduced.
SERVICE DISCOVERY
Service discovery refers to how mobile clients find services when they are needed. Service discovery protocols originated from various network environments including LAN (Jini, UPnP, SLP), WAN (Secure Service Discovery Service (SSDS), GloServ) and Ad hoc (Bluetooth, SDP, Konark, Service rings) networks (Marin-Perianu et al, 2005). They are classified as directory-based, directory-less, or hybrid architecture based protocols (Ververidi and Polyzos, 2008). The basic difference among these is whether or not services are found through a directory or by a multi-cast (i.e. broadcast) query to all potential service providers. As pointed out earlier, directory-less environments do not support searching for services and further deprive service users of privacy.
Essential components in service discovery include Client, Directory and Server. The names of these components may differ based on a particular service discovery protocol. For instance, Sun’s Jini calls the directory as the look-up service and Service Location Protocol (SLP) calls the directory as the directory agent. The components in the service discovery architecture can act based on a client-server model (e.g., Bluetooth) or peer-to-peer model (e.g., Universal Plug and Play - UPnP). Adhoc protocols such as Bluetooth can take a long time for service discovery (22.5 sec) and may not be suitable for the “go-go” dynamic environment that we envision (Nishkam et al. 2005).
SERVICE OFFERINGS (SERVICE CLASSIFICATION)
The method and type of service provisioning often depends on the type of service being provided; a service classification facilitates delivery. Researchers have often classified services as: location specific vs location independent (Linwa and Pierre 2006) and alternatively as “local” vs “global” (Raverdy et al. 2006); free vs paid (Giaglis et al. 2003); profit vs non profit (Nishkam et al. 2005) and domain based (Amaravadi et al. 2009) (see figure 1). Location specific or Location Based Services are those services where there is a location specificity or dependence such as information on wait times at nearby attractions. Obviously, such information would not be of use to anyone outside the vicinity. Location independent services are those where the service is not bound to a particular location. Stock market quotes or streaming broadcast of a selected game are examples of the latter because these will be the same regardless of location. Domain based services include ‘educational,’ ‘government,’ ‘medical’ etc. services as well as ‘utility’ services (Mangalaraj and Amaravadi 2008). The last category comprises routine services such as traffic, printing, storage etc.