Future Internet: towards context information brokering

Transcription

1 Nederlandse Organisatie voor toegepast-natrwetenschappelijk onderzoek / Netherlands Organisation for Applied Scientific Research Brassersplein 2 P.O. Box GB Delft The Netherlands TNO report RA Ftre Internet: towards context information brokering T F info-ict@tno.nl Date 28 September 2009 Athor(s) Reviewer Oskar van Deventer, Pal Tilans, Mike Schenk, Eelco Cramer, Joost Adriaanse Armin van der Togt Cstomer Jan Brgmeijer, TNO ICT VP Ambient Networks and Services Projectname Architectr Ftre Internet Projectnmber All rights reserved. No part of this pblication may be reprodced and/or pblished by print, photoprint, microfilm or any other means withot the previos written consent of TNO. In case this report was drafted on instrctions, the rights and obligations of contracting parties are sbject to either the Standard Conditions for Research Instrctions given to TNO, or the relevant agreement conclded between the contracting parties. Sbmitting the report for inspection to parties who have a direct interest is permitted TNO

2 TNO report 2 / 15 This report contains the paper Ftre Internet: towards context information brokering, pblished in the book Towards the Ftre Internet", see

3 TNO report 3 / 15 Ftre Internet: towards context information brokering M. Oskar van Deventer 1, Pal Tilans 1, Mike Schenk 1, Eelco Cramer 1 and Joost Adriaanse 1 1 TNO Information & Commnication Technology P.O. Box 5050, 2600 GB Delft, The Netherlands {oskar.vandeventer, pal.tilans, mike.schenk, eelco.cramer, joost.adriaanse}@tno.nl Abstract. Ftre Internet is abot combinations of commnication, content and context services. Whereas the former two have achieved already a reasonable state of matrity, context information services are still at their infancy: at most stand-alone applications with limited on-line-or-off-line presence information. The critical sccess factor is still missing, namely context federation, which is the exchange of context information between different application, services and providers. This article investigates how context services cold be sccessflly federated by following the same pattern that many other information and commnication services have sccessflly followed in the past. First Context Information Aggregators and later Context Information Brokers play a critical role in addressing the market need for federated context information services. This article highlights challenges that have to be overcome to make this vision come tre. If Erope takes the lead in overcoming these challenges, Erope can become a florishing grond for a new context-brokering indstry. Keywords: Ftre Internet, Federation, Interconnection, Context, Presence. 1 Introdction Ftre Internet is more than fast video commnication and sharing of content. An important aspect is social networking ( Web 2.0 ) and the context of the people that are networking: where are they, what are they doing, how do they feel, what do they have to share, how do they want to be reached. Ideally, yor friends can easily obtain a fll insight in yor context, in a way that conforms yor own wishes and policies while giving yo a minimm of hassle, that respects yor privacy, that is aggregated from mltiple sorces and over mltiple service providers, and that is visalized with the best possible ser experience for yor friends. Crrently, context information is rather limited and cmbersome in se. At most, yor friends can see whether yo are logged-in to a specific service. Moreover, they first need to log in to the same service from the same service provider to get that information. Federation is a powerfl concept in Information Society Technologies. The global coverage of telephony and Internet cold only be achieved by the federation of the

4 TNO report 4 / 15 networks of many telephony and Internet service providers. In order for context information to become as biqitosly available as telephony or Internet, context federation is needed. This article investigates how federation has developed in sccessfl mlti-operator services in the past. A pattern emerges that is sbseqently applied to context services. This pattern starts with monolithic service providers, like pre-1970 telephony service providers. Stimlated by the arrival of psedo federation providers, often identified as parasitic by the monolithic incmbents, the market need for federation is addressed. Over time, psedo federation evolves into symbiotic federation and fll federation, with a major role for Context Information Aggregators and later also Context Information Brokers. This article does not intend to provide a complete soltion for context federation. It is however or belief that federation at service level in general and context federation in particlar is a key challenge in the development of the Ftre Internet. If Erope takes the lead in overcoming these challenges, it is well placed to become a florishing grond for a new federated context-brokering indstry. Section 2 highlights the federation pattern. Section 3 applies the federation pattern to context information services. Section 4 presents the challenges associated with this vision. Section 5 concldes with a Eropean view on the context information market. 2 The federation pattern, from monoliths to fll federation Service federation is an important featre of modern telecommnications. It refers to a sitation where several service providers work closely together to provide a seamless service to their combined grop of end-sers. Prime examples of federation are international telephony and GSM roaming. In both cases, the end-ser only has a formal relationship (the sbscription) with its own operator, bt is able to access sers and resorces in the domain of other operators. Technical and bsiness agreements between all involved operators ensre that the end-sers experience a seamless service, as if it were offered by one global telecom operator. The above example is probably the most matre type of service federation possible. It has the characteristics of a political federation, as is defined below: A federation (Latin: foeds, covenant) is a nion comprising a nmber of partially self-governing states or regions nited by a central ("federal") government. In a federation, the self-governing stats of the component states is typically constittionally entrenched and may not be altered by a nilateral decision of the central government. Sorce: Wikipedia In the context of federation for ICT services, the federal government is often a standard-setting body, sch as: ITU, IETF or GSMA.

5 TNO report 5 / Types of service federation The following important types of service federation are distingished in this paper: Commnication service federation Content service federation Context service federation Commnication service federation is abot the establishment of service sessions with two or mltiple end-points, where each of these end-points is in a similar role and may have immediate reactions to the actions of other endpoints. Examples are phone/video calls and instant messaging. Content service federation is abot the delivery of content from content providers to consmers. The end-points in the established service sessions play different roles. Architectres for content federation are crrently elaborated by the EUREKA CELTIC RUBENS project [1]. Context service federation is abot the distribtion of context information between providers and consmers. Context information may be (semi) static (e.g. address information) or (highly) dynamic (e.g. location, mood, availability). The isses with context federation are standardization of the varios contexts, privacy management and the vast amont of data involved. Context federation can be sed to enrich and link content and commnication services. These types of service federation are not mtally exclsive. A single service, offered across domains, may inclde several types of service federation. The types of federation are named after the grop of services where they are most important. 2.2 The for stages of service federation Fll service federation is thoght to be the most matre stage in the development of ICT services. However, when looking at the developments in ICT services other stages can be identified, some of them offering a qite acceptable sbstitte for federation. These stages and their characteristics are described below; see also Figre. Withot aggregation Withot federation 1. Monolithic Service With federation 3a. Fll federation of s With aggregation PFP 2. Psedo Federation SFP 3b. Symbiotic Federation Figre 1: The for stages of service federation.

6 TNO report 6 / Monolithic: Service s competing for market share The monolithic stage is believed to be the first stage in the development of services. Each Service is striving to offer a niqe service and wants a large market share. Becase of the fierce competition, the providers are not concerned with working together to offer a seamless service to their combined ser base. : User : Service Figre 2: Monolithic Service competing for market share. The most obvios crrent example is Instant Messaging (IM). There are several IM providers sch as ICQ, Google Talk and Windows Live Messenger (WLM). They all want to achieve the highest market share and lre away end-sers by creating a commnity feeling. An individal end-ser that wants to se IM with a bddy that is connected to another IM provider will sally try to inflence that bddy to join the clb and sbscribe to their IM service. As a reslt, end-sers with bddies in several commnities are forced to sbscribe to several IM service providers and implement the interworking themselves (by installing several IM clients or reverting to mltiheaded clients sch as Pidgin 1, AdimX 2 or Miranda 3 ) Psedo ( parasitic ) federation: disrptive innovation The psedo federation stage is entered when a service becomes sccessfl, with several monolithic Service s offering a similar service. End-sers will grow weary of the fact that they have to implement interworking themselves and are inclined to divert to any provider that offers a soltion for their predicament. This is the time when the Psedo Federation arises. : User : Service PFP PFP: Psedo Federation Figre 3: Psedo-federation, disrptive third parties addressing a market need. Figre explains the concept of psedo ( parasitic ) federation. A Psedo Federation connects to the existing interfaces offered by each of the monolithic Service s and then provides a new interface towards the end

7 TNO report 7 / 15 sers. End-sers that find interworking important will start obtaining the service from the Psedo Federation, while other sers will stay with the existing Service. As a reslt, the Psedo Federation is in direct competition with the Service s and is regarded as parasite by these monoliths. In the context of Instant Messaging (IM), Palringo 4, Fring 5, Nimbzz 6 and many others can be regarded as the first Psedo Federation s. However, these have not yet gained mch momentm, as apparently not enogh end-sers are bothered by the lack of federation in IM. Psedo-federation is difficlt to block for the monoliths becase the reqired technology is rather simple. They can try to make life difficlt for the Psedo Federation, bt they cannot go very far in this. Becase the Psedo Federation ses the same interface as the reglar end-sers, barriers implemented here will also drive away the reglar sers Symbiotic federation: matring of the market The next stage is symbiotic federation. From a technical point of view, it is similar to psedo federation. However, the monolithic Service s have started to accept the fact that there are Psedo Federation s and they also start to see the vale of those parties. Especially for niche markets they will refer their end-sers towards the Symbiotic Federation. SFP : User : Service SFP: Symbiotic Federation Figre 4: Symbiotic Federation s becoming accepted players. Paypal 7 is an example of a Symbiotic Federation. A web shop that wants to provide its sers with the ability to pay with credit cards can either obtain the services from Paypal or from each of the individal credit card companies. However, with Paypal the web shop owner reqires only one connection and corresponding contract instead of hooking p with each of the credit card companies. Likewise, the credit card companies are not that interested to directly deal with each individal small web shop and will refer a small starting web shop to Paypal Fll federation, towards global service coverage Fll federation is federation as we know it from telephony. It does not matter from which service provider the end-ser obtains the service; it can interwork with all of

8 TNO report 8 / 15 the service providers reslting in global coverage for the service. There is no intermediate federation provider. Important in fll federation is standardization; each service provider offers a similar interface to both end-sers as well as the other service providers. : User : Service Figre 5 Fll federation; integrating the Service and Federation roles. 3 Applying the pattern to context services: towards brokering As described, the federation pattern is market-driven. It is the ser that ltimately reqires federation. Reglators often play a key role in this process. This section applies the federation pattern to context information services. Figre sketches the exchange of context information at a conceptal level. Following the IETF terminology [2], the presentity is the identity to which the presence or context information relates. In its most basic form, a presentity wold be the identity of a device or application, e.g. a mobile phone or PC being on-line or off-line. In a more advanced form, the presentity cold refer to a person with mltiple devices and accessing mltiple applications simltaneosly. A watcher is the identity that reqests and consmes presence or context information. Watcher reqests context information abot Presentity Figre 6: Conceptal representation of a context information service. Figre applies the for stages of the federation pattern to context services. The following bsiness roles are distingished. : Context Information. This is the Service for whom at least some form of context information (e.g. simple on-line-off-line presence information) is part of the service offering. CIA: Context Information Aggregator. This is the Psedo Federation whose core bsiness is the aggregation of context information aggregation. CIB: Context Information Broker. This is the Symbiotic Federation whose core bsiness is the brokering of context information aggregation. The following sbsections explain the model of Figre in more detail.

9 TNO report 9 / 15 CIA!? CIB Figre 7: Applying the pattern to context services. 3.1 Monolith: Context Information s have no interest in federation Crrently, presence is not a stand-alone service. Context information is sally delivered as part of a larger service, like instant messaging and voice/video commnication services. Examples of sch commnication services are Skype 8, Windows Live Messenger 9 and Google Talk 10. Users can see whether other sers of the same service are on-line or not, and that is abot it. The context information is available for sers to see whether other sers are available for commnication throgh the particlar service. These services are typically paid by advertisements [3]. The main driver for players in this market is to attract more cstomers, getting a larger market share and obtain more advertisement revenes, see Figre. Notice that context information is not the core bsiness of parties acting as Context Information. Even thogh it wold be more practical for sers to se a single service to commnicate with any other sers, the advertisement revenes of the service providers provides no incentives to offer sch commnication interconnection, let alone the sharing of context information. : Presentity : Watcher : Context Information Figre 8: Context Information s main driver is market share

10 TNO report 10 / Context Information Aggregators fill in an pcoming the market demand Context information aggregation is a first step to address the market demand for context information services. Two types of presence aggregation can be distingished: watcher aggregation and presentity aggregation: Watcher aggregation is the aggregation of presence information from mltiple sorces offered/tailored to watchers. Watchers can attribte different sorces of presence information to a single presentity, and watch the aggregated presence information throgh a single application. Presentity aggregation is the aggregation of presence information from mltiple sorces abot a single presentity, which is typically a person in the case of presentity aggregation. The presence information cold be abot availability to commnicate throgh different channels (fixed phone, mobile phone, instant messaging, SMS), abot the location/speed of the presentity or abot the activities of the presentity. It depends on the configred policies what information specific watchers may see and at what level of aggregation and detail. Presentity federation is complex, as it reqires identity federation [4]. As monolithic Context Information s have no interest in federation,this offers a bsiness opportnity to external Context Information Aggregators to deliver context information aggregation services, see Figre. Initially, sch a service wold focs on watcher aggregation, as it the easier to implement by a third party. The Jabber network 11 can be seen as an example of a Context Information Aggregator. Context Information s perceive Context Information Aggregators as parasitic as they se the Context Information s capabilities withot respecting the Context Information s bsiness model. CIA : Presentity : Watcher : Context Information CIA: Context Information Aggregator Figre 9: Market demand for aggregation addressed by third parties, Context Information Aggregators. 3.3 Symbiotic federation: a role for Context Information Brokers Symbiotic federation is a next step to address the market demand for context information services. This will indce changes in the bsiness model for context information. Crrently this information is for free, paid for throgh advertising following the classical Internet bsiness model. Over time, other bsiness models will 11

11 TNO report 11 / 15 arise. Watchers and/or presentities may recognize the vale of good context information and become willing to pay for the pblication and/or reception of this information. Alternatively, the costs related to context information may be sbsidized by the paid commnication services that they indce in a similar way that free SS7 signaling is being paid for by phone calls. In this alternative, presence has become tomorrow s dial tone [5][6]. With the changing bsiness models, the relationship between parasitic Context Information Aggregators and the incmbent Context Information s will evolve into a more symbiotic one. Upcoming Context Information Brokers wold collect and aggregate context information on behalf of presentities, enabling sers to provide a self-controlled presentity-aggregated view to watchers. Context Information Brokers wold play a complementary role to the Context Information s in a rich context-information ecosystem, with the Context Information s focsing on commnication services and providing context information to Context Information Brokers with context information, see Figre. : Presentity CIB : Watcher : Context Information CIB: Context Information Broker Figre 10: Context Information Brokers aggregating and brokering context information in association with Context Information s. 3.4 Fll interconnection-based context federation. Fll interconnection-based context federation wold in theory be the final step in the evoltion of presence into context services. This step assmes the arrival of pre Context Information s to whom context information services is core bsiness. Context Information s se direct interconnection to share context information with other Context Information s, see Figre. We consider sch a scenario nlikely, as presence and context services are typically spplementary to commnication services. So, even if the commnication services themselves are interconnected, the brokering of context information may remain the field of specialized Context Information Brokers.? : Presentity : Watcher : Context Information Figre 11: Fll federation for Context Information

12 TNO report 12 / 15 4 Challenges This section presents the challenges that need to be tackled to trn the vision of the last section into reality. These topics shold be addressed in a research program focsing on context federation in the Ftre Internet. 4.1 Crrent developments on service federation The traditional telecom operators have applied a closed and central federation model for many years. With the newly evolved VoIP services, we see new and open federation models being applied once VoIP islands need to be interconnected, mainly becase of the different bsiness models applied (see for example the IETF EERMINT initiative [7]). As the reqirements of specific commnication services differ, there is also a need for different federation models. Althogh still nclear, it is expected that different commnication federation models will coexist [8]. A crrently important sbject for context service federation is performance. Althogh crrent protocols standards are sfficient to spport small-scale presence and messaging environments, spport of large scale and federated environments is lacking. Even in new IETF drafts on the optimization of SIP as presence and messaging protocols, we see dobt on applicability of the crrent SIP protocol in large mlti domain implementations [9]. Possible soltions are proposed where the client ses a different protocol than the central servers [10] or modifications are made to SIP [11]. Frthermore we see that federation of the identity management context service, implemented over different organisations, is becoming more and more important [12]. 4.2 Bsiness role and model challenges Bsiness roles and changing bsiness models are a major challenge for context information brokering. The emergence of lcrative context information bsiness will depend on sccessfl bsiness models, either evoltionary or disrptive. As is stated in [13] federation is the next best thing as seen from a ser s perspective bt there is often no incentive for the bsiness parties involved to give p their direct link with the end-ser. While the referred paper focses on identity federation, it is or belief that similar isses will arise with any kind of federation in the Ftre Internet. Crrent services in the Internet are provided by monoliths with bsiness revenes based on the nmber of end-sers they serve. They are nlikely to give p their direct ser contact withot a fight. Therefore, it is likely that federation will be slowly introdced and an analysis shold be made of the characteristics of the bsiness roles in the different market evoltion stages. It shold be stdied to what extent the different forms of presence aggregation (presentity aggregation verss watcher aggregation) relate to similar or distinct bsiness roles. It shold also be stdied how bsiness roles cold gradally evolve and which bsiness roles can only emerge in a disrptive way, see Figre.

13 TNO report 13 / 15 CIA1 CIB1 CIA2 CIB2 Figre 12: Elaboration of bsiness roles and associated technical interfaces Also bsiness models shold be frther elaborated, starting with a vale analysis. In an advertisement-based bsiness model the vale is in the watched advertisements, and the focs is on stimlating the se of the service. In a watcher aggregation model, the vale is in the context information itself and the aggregation of this information, which implies that the model shold fel on revene streams related to watchers. In a presentity aggregation model, it is the presentity that ses context information to pblish his reachability for commnication services, which implies that the model may be feled from commnication services revenes ( dial-tone of the ftre ). Finally, migration and evoltion of bsiness models shold be stdied. For instance, commercial television channels sed to be paid from advertisement revenes and broadcast networks wold share advertisement revenes with television stations. 4.3 Architectre and traffic management challenges Defining technical soltions for aggregation services and context information brokering implies the design of one or more architectres. Traffic analyses will form an important inpt to the architectre design. An analysis shold be made on the flow of context information, inclding realistic estimates on amonts of context information traffic, aggregated and not-aggregated and traffic matrices. Different traffic scenarios may be worked ot depending on the se of the context information. For example, a semi-continos stream of context information may case serios congestion in mobile networks. Based on traffic analyses, recommendations can be made on a high-level architectre. For instance, a client-server-based architectre wold aggregate context information in servers, which has the advantage of central processing, limited context information traffic and simpler (hence cheaper) end terminals. On the other hand, a peer-to-peer-based architectre wold limit the centralized operation to a minimm, most likely only grop list management, and have the context information be

14 TNO report 14 / 15 exchanged directly between end devices withot any direct interference of the context information broker. One or more fnctional architectres shold be worked ot in terms of services spported, elementary fnctions, fnctional elements, data model, reference points/interfaces, protocol flows and protocols sed. Part of the analysis is the qestion which interfaces reqire standardization, either from a mlti-vendor perspective of the technical interfaces between different bsiness roles. Rese of existing standards, possible modifications and profiling, and making contribtions to standardization bodies are part of this challenge. 4.4 Privacy protection and policy management challenges Privacy protection and policy management are essential aspects of any context information ecosystem. An end ser shold be able to rely that his precios context information is treated confidently and only be disclosed to selected watchers. A first step is a thorogh privacy threat analysis. An evalation shold be made which different types of context information there are and what levels of sensitivity shold be distingished. For example, simple on-line-or-off-line information wold in most cases be mch less sensitive than accrate GPS location information, let alone health sensor information. One analysis wold be primarily from the presentity point of view: who is entitled to watch which context information, which person is crrently actively watching. Another analysis wold be from the service provider perspective, which does not want its ser information be data-mined by potential competitors. Finally, an analysis from a context information broker perspective is applicable, focsing establishing trst as key vale, similar to the role of e.g. a bank. Directly related to privacy protection is policy management. A ser wants to control which other sers can watch what types of information. For example, direct colleages may access most information in one s electronic agenda, whereas cstomers wold be restricted to aggregated in-office-or-ot-of-office information. Sch policies are typically coded in XML [15]. An analysis shold be made on the types of policies reqired from the end-ser perspective. Following the 80/20 rle, a limited nmber of defalt policies shold be defined. 5 Conclsion: Erope shold become a florishing grond for a new context-brokering indstry This article explains the for-stage federation pattern associated with the matring of commnication, content and context services. By applying the pattern to Context Information services, it is shown that the market for context information is still at its infancy, and that there are major bsiness opportnities for Context Information Aggregators and Context Information Brokers. In the past, Erope has shown its force in the federation of mobile commnication services. Stimlated by Eropean collaborative projects and Eropean-scale

15 TNO report 15 / 15 standardization, technologies like GSM and UMTS have reached matrity. Federation has been the critical sccess factor here. The market, architectres, technology and standards for context information services are at a similar infant stage as GSM and UMTS respectively in the early 1980 s and 1990 s. Erope has the opportnity to take the lead in overcoming both technical and bsiness model challenges. If Erope takes this lead, then Erope can become a florishing grond for a new context-brokering indstry. References [1] EUREKA CELTIC RUBENS, Rethinking the Use of Broadband access for Experienceoptimized Networks and Services [2] IETF RFC 2778, A Model for Presence and Instant Messaging; Day, et al, Febrary [3] Bsiness models on the web; Professor Michael Rappa. - Advertising [4] Liberty Alliance [5] Presence in AIM: Evoltion of the Online Dial Tone; Kevin Farnham, November [6] Presence: The Dial Tone for Internet Commnications; Peter Saint-Andre, Director of Standards Jabber, March [7] IETF EERMINT draft-ietf-speermint-architectre-07; R. Penno, D. Malas, S. Khan, A. Uzelac, M. Hammer, [8] Analysis of VoIP interconnection evoltion; Mika Lahti, Helsinki University of Technology, Master s Thesis, Febrary 14, [9] IETF SIP/SIMPLE draft draft-hori-simple-interdomain-scaling-optimizations-00.txt; A. Hori, V. Singh, H. Schlzrinne, S. Parameswar, E. Aoki, [10] Presence Interdomain Scaling Analysis for SIP/SIMPLE; A. Hori et al, October [11] IETF SIP/SIMPLE draft draft-ietf-simple-view-sharing-02; J. Rosenberg, S. Donovan, K. McMrry, [12] Identities federation and Identity s; Miska Laakkonen, Helsinky University of Technology, [13] Identifying Patterns of Federation Adoption; Heather Hinton and Mark Vandenwaver in ISSE 2006 Secring Electronic Bsiness Processes, Vieweg [14] Interdomain Presence Scaling Analysis for the Extensible Messaging and Presence Protocol; P. Saint-Andre, Janary [15] Presence is as presence does (sheet no. 17); Robert Sparks.