Social Carrier Recommendation for Selecting Services in Electronic Telecommunication Markets

The proliferation of telecommunication services and the need to manage quality of service on an end to end basis require an approach for automatically selecting services that provide su cient quality of service at minimal cost An agent based approach is appropriate for such a purpose For this reason this paper presents a social carrier recommendation method which is an essential component of application level end to end quality of service management as well as a way to make the nal step towards electronic telecommunication markets For electronic telecommunication markets the proposed approach provides a consumer based evaluation of services as well as rational user agents that select services and carriers based on needs o ered prices and ratings Therefore this approach complements existing market mechanisms that either provide means to buy services or intend to improve sales and customer service of carriers The Evolution of Distributed Information Networks Distributed Information Networks DINs are developed to exploit the diversity of o ered telecommunication services by utilizing multiple networks to minimize costs circumventing technology limitations e g lack of security and reliability and insu cient quality of service QoS as well as extending reachability and connectivity The need of an agent based approach for automatically selecting services raises from the proliferation of services and the need to manage QoS on an end to end basis These two reasons are discussed in Section and Section respectively The agent system using a social carrier recommendation SCR method is explained in more detail in Section of this paper It minimizes costs and ensures su cient QoS when choosing an appropriate service A user is assisted in SCR by a so called User Agent that selects services according to the directives and ob served preferences of the user The presentation of SCR focuses on the selection process since agents acting as buyers in electronic markets exist already KH This leads in a natural way to electronic telecom munication markets in which telecommunication services are o ered traded and bought electronically by agents Such markets require also agents on the service provider side Realizing services by means of mobile agents within end terminals enables service subscription for single usages Mag which is very attractive for an electronic telecommunication market In WBT Bus a customer is connected in the network of a provider to a customer agent that negotiates the service requests of the customers and pro actively o ers services to the customers Before o ering pro actively services a customer agent determines whether for instance the current price is attractive for a customer Especially this will become important if pricing is dynamic and based on current load as it is the case in WBT Bus Such customer agents improve the service o ered to the customer by a single service provider A true electronic market requires the herein proposed user agent which would evaluate pro active o ers e g to nd a time when a service is cheap and interact with the corresponding customer agents of all carriers in the market In addition to this the community of user agents would rate the delivered services and recommend services based on this rating Hence this community represents an electronic consumer organization Proliferation of Services In the information age telecommunication is used for communication between people to distribute informa tion and electronic services and to coordinate processes These applications pose very di erent requirements on telecommunication services For instance researchers employ the Internet to locate and disseminate infor mation whereas residential users might be mostly interested in communication between people KSM The most successful telecommunication services are the ones that meet best the requirements of their target market which results in a proliferation of services Even the introduction of universal protocols e g ATM and IPv and the standardization of network application interfaces e g Integrated Service Digital Network ISDN and TCP UDP do not prevent the proliferation of services because of the competition in liberalized and global telecommunication markets requires market di erentiation as well as focusing on particular markets di erent service classes e g best e ort real time and so on requires usage based pricing as the Internet community is nding out and multiple networks with di erent capabilities exist due to the evolution of networking technologies and more importantly the variety of available transmission media e g wire line and wireless This trend is not reversed by the Internet since it consists of multiple networks which cover partly the same geographical area Since the start of the privatisation of the Internet individual networks of the Internet are administrated by distinct service providers Therefore it becomes much harder within a multiple adminis trative domain to meet the speci c QoS requirements of service classes In particular the development of performance management protocols and tools become di cult Considering the proliferation of the services from the user s view some other problems have also to be solved The customers have to choose appropriate services and service packages for their applications in such a free market This is di cult for the average consumer and also not its primary task For instance the long distance Public Switched Telephone Network PSTN service providers in the U S make frequently time limited special o ers to increase their market share But some customers are becoming tired of these special o ers Furthermore a free market functions only if there is su cient consumer feedback A special o er might lure a large number of customers but results in a degradation of the QoS For instance the availability of the local call PSTN is reduced in some areas of the U S due to the long connection time of Internet users End to End QoS Management Today s QoS management is insu cient and inappropriate for DINs because QoS is an end to end property experienced by the application and no service provider has global control about all networks In particular this limited control is visible in Global Area Networks for which Albanese et al propose application level network management ABE Clark propagates application level management for high speed protocols because the application knows best how to handle packet loss delay and so on CT Last but not least globally coordinated QoS management is potentially very expensive Therefore an agent based approach managing the QoS on application level might reduce theses expenses The decentralized network model e g embodied by the Internet with its often powerful end terminals i e computers is well suited for application level network management Such end terminals empower applications to judge the QoS themselves and to choose carriers Note that a nearby management server could control a whole range of end terminals e g mobile devices and network computers which have only a limited computing power A hierarchical QoS management approach aggregated QoS information is employed at the higher level would not be very useful for global networks Such an abstraction implies a loss of precision and is di cult to determine In a competitive environment it is unlikely that the service providers exchange extensive QoS data which enables one to evaluate networks of competitors Such an exchange of information about the network status e g load delays congestion and so on is required for QoS routing for example The perceived QoS depends on the user of an application the situation as well as the application itself Therefore it is necessary to enable the user to specify his needs in particular situations These requirements have to be mapped into network service QoS requirements AS Since the routes of information ows have to be determined the selection of a service provider is of particular interest in this paper For this purpose source routing is required to enforce the choice IPv DEF for instance provides the necessary support which is also required for QoS routing Source routing for provider selection is di erent from policy routing Cla which enables network providers to route tra c based on policies Nevertheless many issues are the same for both approaches The implementation of source routing for service provider selection is not further discussed in this paper Social Carrier Recommendation The basic idea of SCR is to adapt social information ltering techniques SM for recommending services i e enable users with similar applications and with similar communication needs to recommend services and service providers respectively The SCR provides assistance to a user for selecting appropriate services and carriers or service providers for the applications the user utilizes This Section focuses on the selection process because agents acting as buyers on behalf of users exist already KH The SCR User Agent and its QoS Manager Function A SCR User Agent keeps track of the user s expectations i e user level QoS which are recorded in a User Pro le the applications used and their requirements Application Pro le and service agreements and the measured delivered QoS A SCR User Agent see Fig has to handle multiple devices some of which use the same communication services Therefore a user agent is itself a distributed and mobile agent system in general This does not impact the task of a user agent so that this paper assumes without loss of generality for explanatory purposes that a user operates only a single device The User Pro le de nes what applications a user how uses i e the time of the day the frequency and the user s expected QoS An Application Pro le de nes the generic application speci c requirements For instance best e ort services might specify a minimal required bandwidth and maximal consumable band width A user pro le might specify a higher minimal required bandwidth for a le transfer service in order to take into consideration the wish of the particular user to complete le transfers in a short time These expec tations depend on the time of day a user might demand shorter le transfer times at the end of his working user’s expectations (User Profile) the application and their requirements (Application Profile) service agreement and the measured delivered QoS Service 2 keeps track of SCR User-Agent selects services according to directives Application User