future of networking, 2006 1 gregor v. bochmann, university of ottawa presentation given at the...

Gregor v. Bochmann, University of Ottawa Future of Networking, 2006 1

Presentation given at the e-Science Institute, EdinburghSeptember 14, 2006

Gregor v. Bochmann

School of Information Technology and Engineering (SITE)University of Ottawa

Canada

http://www.site.uottawa.ca/~bochmann/talks/FutureNetworking

Challenges for the Future of Networking


AbstractThe technical foundations for the Internet were developed more

than 30 years ago. Since over 10 years, it has developed into a general communication infrastructure used by people and industry for a variety of applications. While e-mail and the Web were first the most important applications, newer developments have introduced wireless communication and new applications, including multimedia, e-commerce, etc. Certain applications, e.g. in the area of e-science, have extreme requirements in terms of bandwidth or delay that cannot be provided by the current Internet. - This talk will give a personal view of the challenges that must be faced for the future of the Internet and the distributed applications using it, including managerial and technical aspects. Some of these issues are (1) the integration of wireless LANs and ad-hoc networks with the wired network, (2) fast optical switching, (3) user-empowered network management, (4) security and trust management, (5) standards for distributed applications (e.g. Service Oriented Architecture) and (6) ubiquitous computing. The talk will provide a general discussion of these issues and present certain examples of innovative applications.


Overview The current Internet and

applications Research management - Grand

Challenges Research issues in networking Optical networks (the physical

level) Issues for distributed applications Conclusions


Internet: Some Characteristics

Packet switching Buffered in each router or switch (delay)

IP : connection-less Logically simple, but requiring address look-up for each packet Connection-oriented service allows for more efficient switching,

e.g. new MPLS technology There are not enough addresses. Solutions:

use of internal addresses and address translation (NAT); however, internal addresses are not reachable

or better: use IPv6 TCP : controls flow between end-systems

Provides reliable information flow Many applications need a logical connection between processes

running in different hosts Not suitable for interactive voice or video traffic (retransmission

introduces delays) Not suitable for very large bandwidths (order of Gbps)

UDP : non-reliable alternative to TCP


Some extreme applications Large bandwidth and low delay : Video

teleconference (e.g. round-trip delay of 0.1 sec at 10 000 km)

Need for multicasting: video broadcasting (e.g. 10 Mbps to 10 000 users : 100 Gbps)

Extreme large bandwidth: e.g. 10 Gbps for e-science applications

Extremely low delays: tele-manipulation (e.g. eye surgery training); distributed music ensemble

Ad hoc networking (without fixed infrastructure)

people in local meeting Sensor networks (large number of sensors, low battery life, may fail)


Existing communications infrastructures

Terrestrial transmission infrastructures Optical fibres Wavelength division multiplexing (each wavelength : typically

10 Gbps) For transmission, data is converted (from the electrical domain)

into the optical domain (and back, by the receiver) 10 Gbps is too much for most applications, it must be

shared Bandwidth sharing for telephony (end-to-end flows of fixed

bandwidth, not packet switching) Sonet or SDH (time division multiplexing) ATM (cell switching)

Packet switching may be used for this purpose (switching in the electrical domain)

Packet switch could use 10 Gbps wavelength, or a fraction provided by SDH

Time sharing through photonic switching, e.g. burst switching Cellular networks (designed for telephony) Fixed wireless networks (WIFI)


Network management and scalability

Need for interworking between different domains (subnetworks belonging to different organizations)

Limited visibility Service level agreements (static – dynamic)

Large number of … (scalability) Domains Routers / switches Host computers Communicating devices (terminals, phones, TVs, kitchen stoves,

etc.)

Security and reliability A faulty behavior of a single router should only have local

impact; idem for failures


R&D - a long path: From new idea to market place

Typical time : 20 years Example: Modeling distributed systems by state

transition diagrams 1969: Bartlett describes a communication protocol with

finite state machines (FSM) 1976: First version of SDL includes FSM notation 1977: Bochmann and Gecsei propose Extended FSMs for

modeling communication protocols 1980ies: Standardization of formal description techniques

(FDTs) by ISO and ITU, including SDL; university-based tool development

1987: Harel proposes State Charts (including certain extensions of above notations)

1990ies: Commercial development of software tools supporting these notations

1995 ?: Unified Modeling Language (UML) defined by OMG Around 2005: Integration between SDL and UML Version 2


The research planning process (A)

Funding of research and development By industry (internal or external research)

Objective: improve competitiveness Better products Better development and production methods

Only larger companies perform longer term research and planning By government organizations (industrial and university

research) Improve competitiveness of country

Competent people Improve global competitiveness of local industry Development of Intellectual Property (IP) to be used by local industry

Difficulty of prioritizing the different fields of science and technology

Give equal chances to all disciplines ? Declare certain fields as « national priority » ? Let industry buy-in for joint government-industry funding programs


The research planning process (B)

Community-based research planning Consensus building: through mailing lists,

discussions at workshops / conferences, research collaborations

Examples: The UK Grand Challenges: a perspective on long-term basic

and applied research NSF (USA) Workshop on Overcoming Barriers to Disruptive

Innovation in Networks Research program of E-NEXT (a EU - FP6 Network of

Excellence) “CoNEXT” conference in Toulouse, Oct. 2005

http://dmi.ensica.fr/conext/

Canadian research network on Agile All-Photonic Networks (AAPN, funded by NSERC and 6 industrial partners)


Grand Challenges (defined in the UK)

See http://www.ukcrc.org.uk/grand_challenges/index.cfm “Definition of a Grand Challenge

A grand challenge should be defined as to have international scope, so that contributions by a single nation to its achievement will raise our international profile.

The ambition of a grand challenge can be far greater than what can be achieved by a single research team in the span of a single research grant.

The grand challenge should be directed towards a revolutionary advance, rather than the evolutionary improvement of legacy products that is appropriate for industrial funding and support.

The topic for a grand challenge should emerge from a consensus of the general scientific community, to serve as a focus for curiosity-driven research or engineering ambition, and to support activities in which they personally wish to engage, independent of funding policy or political considerations. “ (Note: the quotes, here and in subsequent slides, indicate that the text is copied from the source documentation)

The following two slides are from Robin Milners talk “A scientific horizon for computing” at the World Congres 2004 of the International Federation for Information Processing (IFIP), held in Toulouse.


Grand Challenge Exercise


UK Grand Challenge Proposals

Note: No GC is dedicated to networking issues


Ubiquitous Computing Grand Challenge

Combination of GC 2 and GC 4 See http://www-dse.doc.ic.ac.uk/Projects/UbiNet/GC/index.html

Objective: “We propose to develop scientific theory and the design principles of Global Ubiquitous Computing together, in a tight experimental loop.”

“Engineering challenges: design devices to work from solar power, are aware of their

location and what other devices are nearby, and form cheap, efficient, secure, complex, changing groupings and interconnections with other devices;

engineer systems that are self-configuring and manage their own exceptions;

devise methods to filter and aggregate information so as to cope with large volumes of data, and to certify its provenience.

business model for ubiquitous computing, and other human-level interactions. “


Ubiquitous Computing Grand Challenge (ii)

“Scientific challenges: discover mathematical models for space and mobility, and develop

their theories; devise mathematical tools for the analysis of dynamic networks;

develop model checking, as well as techniques to analyse stochastic aspects of systems, as these are pervasive in ubiquitous computing;

devise models of trust and its dynamics; design programming languages for ubiquitous computing. “

A comment: It is not clear where – in the context of ubiquitous computing – Networking stops and Computing starts. In fact, networking involves much distributed systems management (including databases); and for the Internet applications, the application layer protocols are just as important as (if not more than) the underlying networking protocols.

Note: Milner has developed a new description formalism “Bigraphs for Mobile Processes “

( see http://www.cl.cam.ac.uk/users/rm135/ )


Research topics in “Networking”

Issues Network layer:

new wireless technologies: cellular, LAN, PAN, ad-hoc, sensor, etc. Integration with wire-line Internet Higher bandwidth

Inter-layer control and management according to application needs Physical layer: technology push

Faster electronic components, e.g. 10 Gbps Ethernet Fast optical switching Trend: IP over Dense Wavelength Division Multiplexing (DWDM);

elimination of intermediate layers of ATM, SONET; however, it may be IP over MPLS over DWDM.

Application layer many new applications: importance of multimedia application will increase New protocols for organizing applications: Web Services, Grid, peer-to-peer New ways for identifying and searching services, including concern for

security and trust

Networkservice

Architectural levels of Networking Technology a narrow-waisted hourglass model:


Overcoming Barriers to Disruptive Innovation in Networks

Workshop organized by NSF (USA) “Overcoming Barriers to Disruptive Innovation in

Networking” (Jan. 2005) see http://www.arl.wustl.edu/netv/noBarriers_final_report.pdf

Starting point: “ The Internet is ossified: … Adopting a new architecture not only requires modifications to routers and host software, but given the multi-provider nature of the Internet, also requires that ISPs jointly agree on that architecture. The need for consensus is doubly damning; not only is agreement among the many providers hard to reach, it also removes any competitive advantage from architectural innovation. This discouraging combination of difficulty reaching consensus, lack of incentives for deployment, and substantial costs of upgrading the infrastructure leaves little hope for fundamental architectural change. “


NSF workshop (ii) Requirements for the new Internet:

“ Minimize trust assumptions: the Internet originally viewed network traffic as fundamentally friendly, but should view it as adversarial;

Enable user choice: the Internet was originally developed independent of any commercial considerations, but today the network architecture must take competition and economic incentives into account;

Allow for edge diversity: the Internet originally assumed host computers were connected to the edges of the network, but host-centric assumptions are not appropriate in a world with an increasing number of sensors and mobile devices;

Design for network transparency: the Internet originally did not expose information about its internal configuration, but there is value to both users and network administrators in making the network more transparent; and

Meet application requirements: the Internet originally provided only a best-effort packet delivery service, but there is value in enhancing (adding functionality to) the network to meet application requirements. “

Identified 7 areas of research (see next slides)


7 research areas: Security Economic incentives Address binding End-host assumptions User-level route choice Control and management Meeting application requirements (see next slides)


Security Problem indications

“traffic must be viewed as adversarial rather than cooperative” “To take one example, a single mistyped command at a router at

one ISP recently caused widespread, cascading disruption of Internet connectivity across many of its neighbors.”

Benefits of better security1. “ improve network robustness through protocols that work

despite misbehaving participants,2. enable security problems to be addressed quickly once identified, 3. isolate ISPs, organizations, and users from inadvertent errors or

attacks; 4. prevent epidemic-style attacks such as worms, viruses, and

distributed denial of service; 5. enable or simplify deployment of new high-value applications and

critical services that rely on Internet communication such as power grid control, on-line trading networks, or an Internet emergency communication channel; and

6. reduce lost productivity currently aimed at coping with security problems via patching holes, recovering from attacks, or identifying attackers. “


Security (ii) Interesting architectural approaches:

“prevent denial of service by allowing a receiver to control who can send packets to it “

“making firewalls a fully recognized component of the architecture instead of an add-on that is either turned off or gets in the way of deploying new applications. A clean specification for security that makes clear the balance of responsibility for routers, for operating systems and for applications can move us from the hodge-podge of security building blocks we have today to a real security architecture “

“A careful design of mechanisms for identity can balance, in an intentional way rather than by accident, the goals of privacy and accountability. Ideally, the design will permit us to apply real world consequences (e.g. legal or financial) for misbehavior. “


Economic incentives Proposition:

“A future design for an Internet should take into account that a network architecture induces an industry structure, and the economic structure of that industry. The architecture can use user choice (to impose the discipline of competition on the players), indications of value flow (to make explicit the right direction of payment flow), and careful attention to what information is revealed and what is kept hidden (to shape the nature of transactions across a competitive boundary). “


Address binding Problem with IP addresses

There are not enough – solution: IPv6 They serve as machine identity (instead of only identifying

the network attachment point, the location) this leads to difficulties for mobile devices (e.g. Mobile IP

routing is not straightforward – IP address changing dynamically)

IP address (as machine identifier) also used for security Proposed solution approaches

Host Identity Protocol It provides secure host identification Routing is based on IP addresses that are treated only as

ephemeral locators “… end-points (as equated with physical machines or

operating systems) need not have any globally known identity at all. Instead, application level entities have shared identities … , and higher level name spaces such as a redesigned DNS are used to give global names to services, so that they can be found. “


End host assumptions Issues with sensor networks

sensors may be intermittently connected routing may be based on data values

Solution approaches: Overlay networks Overlay for realizing special routing

functions, e.g. diffusion routing Overlay for delay-tolerant routing (e.g. for e-

mail; also allowing “access in a variety of impoverished and poorly connected regions “)


User-level route choice Objectives: increase the user’s choice

and introduce more competition “ Instead of applying a "one-size-fits-all" policy to

their traffic, ISPs could perform routing and traffic engineering based upon the user traffic preferences … offer unique policies such as keeping all traffic within the continental United States for security reasons. “

“ This selection creates a more complex economic environment; it offers potential rewards in user choice and competition, but requires solutions to issues of accounting, pricing, billing, and inter-ISP contracts. “


Control and management Statement: Management of the Internet is very

complex (for all parties involved) Solutions: not clear (there are references to

ongoing work) One problem: limited visibility of internal

parameters from outside the network (opaqueness) A network should “support communication of operationally

relevant information to each other. Such information could be aggregated and analyzed, thereby facilitating load balancing, fault diagnosis, anomaly detection, application optimization, and other traffic engineering and network management functions.”

One needs a compromise between information hiding and visibility for management.


Meeting application requirements

Protocol layer architecture is a narrow-waisted hourglass model

Additional requirements: “QoS control, multicast, anycast,

policy-based routing, data caching …”

Possible solutions: Add more functions to IP layer Use overlay networks to provide additional

functions

IP Networkservice


Some personal commentsOverlay networks

Principle: A certain number of servers connected to the Internet play the role of « virtual routers » in the overlay network. Note: This is the way MBone implements multicasting over the current IP Internet service.

The NSF workshop stresses the use of overlay networks for experimentation with new approaches

Could such architectures present the final solution ? NO, overlay technology, such as peer-to-peer computing, may be

useful for certain applications, but cannot be a solution for building a network

Existing well-known applications Napster and BitTorrent media distribution, and other peer-to-

peer applications Multicasting of multimedia presentations, possibly including

different quality variants A Testbed: US-based Planetlab http://planet-lab.org/; see also

http://www.arl.wustl.edu/netv/main.html


Some personal comments (2)

Lightpaths - “Underlay Networks“ ? Experimental research networks provide high-bandwidth

“lightpaths“ between different sites for e-science and other applications that require guaranteed high-bandwidth connections.

For an overview of current applications, see http://www.internet2.edu/presentations/fall05/20050920-lambdas-sauver.htm

User-Controlled Lightpath Provisioning (UCLP) allows the e-science users to establish lightpaths dynamically through a graphic user interface.

Note: UCLP has been initiated in Canada with partial funding from Canarie (the Canadian research network), see for instance http://www.uclp.ca

These networks make use of user-owned fibers and condominium facilities for long-haul transmission and switching

This is not an overlay, but also provides a new networking service, independently from the existing Internet. The Internet can be built on top of it.



Packets vs. (virtual) connections The old debate between packet switching and circuit

switching (from the 1970ies) is not dead !! Distinction: In packet switching, the header of the

packet/frame/cell/burst contains the destination address; in circuit switching, it contains a number (label) identifying the circuit (in TDM, this number is the timing position).

MPLS (label switching) provides packet switching over dynamically established paths (virtual connections)

Optical lightpaths are connection-oriented. It is expected that existing ROADM (Reconfigurable optical add/drop multiplexers) technology will be widely deployed within a few years; see for instance http://lw.pennnet.com/Articles/Article_Display.cfm?

Section=ARTCL&ARTICLE_ID=203231&VERSION_NUM=1 An optical lightpath at a given wavelength is very large,

typically 10 Gbps. Sub-multiplexing of a lightpath in the time domain is proposed by many research projects;

Sharing between packets or virtual connections ??



Appearently contradictory approaches IP : packet-oriented switching The concept of virtual connections are natural for

providing QoS guarantees. The lower layers of broadband wireline networks

appear to use connection-oriented technologies. The overlay networks would like to obtain more

visibility about the performance aspects of the underlying IP service.

Suggestion: Maybe there should be more visibility at the IP service level about the underlying virtual and physical circuits that exist within the network and their performance parameters; and the application should have some choice about the routing of its data.


Optical networks Currently deployed:

optical transmission with DWDM Some optical switching

Note: most “optical switches“ convert the optical signal into the electrical domain and perform the switching in the electrical domain.

Expected to be deployed: ROADM used for transparent optical

switching in the millisecond speed range; good for protection switching and bandwidth on demand.


Burst switching Question: Can one do packet switching in the

optical domain (without oeo conversion)? At a switching speed of 1 μs, one could switch

bursts of 10 μs length (typically containing many packets)

Traditional packet switching involves packet buffering in the switching nodes. Should one introduce optical buffers in the form of delay lines?

The term “burst switching“ originally meant “no buffering”: in case of conflict for an output port, one of the incoming bursts would be dropped.

Note: Burst switching allows to share the large optical bandwidth among several virtual connections.


AAPN

An NSERC Research Network

The Agile All-Photonic Network

Project leader: David Plant, McGill University

Theme 1: Network architecturesGregor v. Bochmann, University of Ottawa

Theme 2: Device technologies for transmission and switching


AAPN Professors (Theme 1 in red)

McGill: Lawrence Chen, Mark Coats, Andrew Kirk, Lorne Mason, David Plant (Theme #2 Lead), and Richard Vickers

U. of Ottawa: Xiaoyi Bao, Gregor Bochmann (Theme #1 Lead), Trevor Hall, and Oliver Yang

U. of Toronto: Stewart Aitchison and Ted Sargent McMaster: Wei-Ping Huang Queens: John Cartledge (Theme #3 Lead)

Note: Theme 2 deals with device technologies for transmission and switching

For further information see: http://www.aapn.mcgill.ca/


The AAPN research network Our vision: Connectivity “at the end of the

street” to a dynamically reconfigurable photonic network that supports high bandwidth telecommunication services.

Technical approach: Simplified network architecture (overlaid stars) Specific version of burst switching

Fixed burst size, coordinated switching at core node for all input ports (this requires precise synchronization between edge nodes and the core)

See for instance http://beethoven.site.uottawa.ca/dsrg/PublicDocuments/Publications/Hall05a.pdf

Burst switching with reservation per flow (virtual connection), either fixed or dynamically varying

See for instance http://beethoven.site.uottawa.ca/dsrg/PublicDocuments/Publications/Agus05a.pdf

Future of Networking, Lausanne, 2005 37

Edge node with slotted transmission (e.g. 10 Gb/s capacity per wavelength)

Opto-electronic interface

Fast photonic core switch (one space switch per wavelength)

- Provisions sub-multiples of a wavelength

- Large number of edge nodes

Agile All-Photonic Network

Overlaid stars architecture


Starting Assumptions Avoid difficult technologies such as

Wavelength conversion Optical memory Optical packet header recognition and

replacement

Current state of the art for data rates, channel spacing, and optical bandwidth

Simplified topology based on overlaid stars

Edge based control in small/medium size edge nodes


Starting Assumptions (ii) No distinction between long-haul and

metro networks Fast optical space switching (<1 sec) Slotted Time Division Multiplexing

(TDM) or slotted burst switching Need for fast compensation of

transmission impairments (<1 sec)


Bandwidth allocation schemes

For flows between edge nodes Optical wavelength: Whole wavelength

(for large bandwidth flows) – like the PetaWeb explored by Nortel Networks

Optical circuit: One or several time slots within each TDM frame

Burst switching: individual bursts (with or without reservation)

Coordination by controller at core node Signaling protocol between edge and core

node (suitable for metro and long-haul networks)


Integration higher layer (MPLS and IP)

MPLS flows passing through the AAPN

With N edge nodes, there are N x N links in the AAPN (scalability problem for IP routing protocol)

“Virtual router” star architecture OSPF sub-areas How to find optimal inter-area route

(work sponsored by Telus)


Deployment aspects - Questions

Long-haul or Metro ? connectivity “at the end of the street”; to a server farm AANP as a backbone network ?

High capacity (many wavelengths) or low capacity (single or few wavelengths) ?

Multiple core nodes ? For reliability For load sharing

Transmission infrastructure ? Using dedicated fibers Using wavelength channels provided by ROADM

network


Issues for

Distributed applications Multimedia Ubiquitous computing and location-

awareness Service-oriented architecture and Grid

computing Making it easy for the end-user Scalability – peer-to-peer computing Related technologies

Security Trust management Software development technology


Distributed multimedia applications

The basics are relatively well understood Video requires high bandwidth Conversational applications require short

transmission delays In many cases, multicasting is required (possibly

provided through the overlay approach) Aspects to be further explored

Shared virtual environments, e.g. for collaborative work or games

Tactile applications; tele-haptics require very short delays

Quality of service management for multiple receivers; media transcoding


TT

Alice

Bob

Bob’sVisiting Domain

12

3

Internet

4

Directory

HDA

5

Bob Home Domain

HDA

FDA

T

TT

T

TT

T

Example: Locating suitable transcoding servers (El-Khatib)See http://beethoven.site.uottawa.ca/dsrg/PublicDocuments/Publications/

ElKh04c.pdf


Ubiquitous computing and location-awareness

See Grand Challenge Example: Some issues encountered in

our project on teleconferencing for mobile users

Problem: In ad-hoc environment (e.g. on a trip) find out what devices may be useful to the user to establish a video-conference with a friend in another country.

Consider quality of service (QoS) negotiation to find most suitable devices according to the user’s preferences and the remote site.

Assumption: User has a PDA that can detect through short-range wireless communication (e.g. Bluetooth) which devices are available in the environment.

Approach: We use a Home Directory to store the preferences of the user; it must be down-loaded into the PDA for processing (it may be a rented PDA). See http://beethoven.site.uottawa.ca/dsrg/PublicDocuments/Publications/ElKh04a.pdf


Example: Device selection in an ad-hoc environment

Internet

Bob’s HDA

Alice’s HDA

Alice

PA(PDA)

12 3

4

447 5

56

5

4

4

4

5

55

7


Example: Session mobility and QoS adaptation

QoSNegotiation

and SelectionAgent User

ContextAgent

Personal Agent

User Profile

ServiceRegistry

Communic-ation Agent

ServiceDiscovery

Agent


Service-oriented architecture and Grid applications

Concepts RPC for accessing services Directory service

Realizations: CORBA, Jini (Java environment)

WS and SOA: use similar concepts Use HTTP and SOAP (based on XML) Workflow specifications (BPEL, etc.) Advantages:

use of HTTP (firewalls) programming language independent (like

CORBA)


Notes on XML text-oriented encoding of data structures

(based on SGML, like HTML) used for storage and/or transmission Data structure (type) definition in the form

of DTD or XML Schema Developed by WWW Consortium

http://www.w3.org/ Used for a multitude of applications, see

for instance list of resources at http://www.extensinet.com/


WS: Example applications E-commerce:

Historical: First e-commerce: Electronic Data Interchange (EDI)

Standards about data elements required in purchase order, invoice, shipping documents, etc.

Standard coding format Message transmission over telephone or leased lines

Transition to the use of the Internet: Development of SOAP (new coding standard based on XML)

Nowadays: many new applications and developments See “Electronic Business using XML” http://www.ebxml.org/ OASIS http://www.oasis-open.org/

Resource sharing E-science projects - Grid computing Network management, e.g. UCLP (see above)

Need for common understanding of information (semantics)

Work by the W3C on the “Semantic Web” http://www.w3.org/2001/sw/


Making it easy for the end-user

“Everyday use” (for our normal day activities) Content creation by the end-user See “It's A Whole New Web” (Businessweek)

http://www.businessweek.com/magazine/content/05_39/b3952401.htm


Peer-to-peer computing Scalability to the millions and more

Load is shared on a peer-to-peer basis Individual servers may come and go Robustness of the overall system

Example of service: distributed storage and search facility

Not only applicable to file sharing

Note: this is an overlay system


Related technologies Security Trust management Software development technology


Security Services

Privacy of message exchanges Integrity of messages Authentication of users and devices Signature with non-repudiation

Cryptographic technologies Secret key encryption Public key encryption (RCA, elliptic, etc.) Hash functions, etc.

Secure private and public networks Integration of security into application layer

protocols New types of applications

Electronic cash


Trust management trust is the outcome of observations

leading to the belief that the actions of another may be relied upon, without explicit guarantee, to achieve a goal in a risky situation

-- Greg ElofsonKey elements

Observations (experience, interaction) Belief (assumption) Goal (expectation) Without guarantee (risk) Subjective


Trust: An example scenario Alice visits her friend Bob who lives since a year in a

foreign country. She wants to invite Bob and some of his friends for supper. She does not know which restaurant to choose, since she wants tasty food, a nice atmosphere and good service.

In her own city, she has experienced many restaurants and she knows the restaurants she would choose depending on how important food, atmosphere and service is for the occasion. She trusts these restaurants, based on her past experience.

Now she asks Bob for his experience in order to select an appropriate restaurant. She trusts Bob for telling her the truth and for evaluating restaurants based on similar criteria as herself.

Then she selects a restaurant with good food, because the friends find food more important than service. (Note: food is the utility to be optimized)


Some observations Trust is used for decision making Trust means a prediction of the outcome of a service invocation

E.g. based on the experience, we predict that the chosen restaurant will provide tasty food.

Our trust model based on statistics and Bayesian estimation http://beethoven.site.uottawa.ca/dsrg/PublicDocuments/Publications/Shi04a.pdf

The space of possible outcomes usually depends on the context in which the trust model is used

Trust is the estimation of a probability distribution over the possible outcomes of experiences

Our own experience is more reliable than the experience of peers, however, peers may have more experiences than we.

Question: can we trust the recommendations of others ? Our recommendation evaluation algorithm

http://beethoven.site.uottawa.ca/dsrg/PublicDocuments/Publications/Shi05a.pdf Weight each recommendation according to the trust in the recommender The trust in the recommender will decrease if a given recommendation is “unfair” How can one determine the “fairness” of a recommendation ??

How detailed should the trust model be ? Should one distinguish different dimensions, e.g. food, atmosphere and

service, or simply have one evaluation category, e.g. the restaurant being either excellent, good, bad or very bad ?

Is it possible to determine the expected error of predictions?


Transactions based on trustExisting access control model for mobile users:

“Autonomic Distributed Authorization Middleware”


Systematic development of distributed applications

UK Grand Challenge ”Dependable Systems Evolution”

use of assertions for defining component requirements “verifying compiler” as a goal

Personal comment: Is this the right approach ??

UML - formalizing its semantics Work in Ottawa:

Defining requirements by scenarios (see http://beethoven.site.uottawa.ca/dsrg/PublicDocuments/Publications/Sand05a.pdf )

Using notations of Activity Diagrams or Use Case Maps (UCMs) (see http://www.site.uottawa.ca/~damyot/pub/index.shtml )

Define semantics of these languages based on Coloured Petri nets

Consideration of performance parameters (see http://www.sce.carleton.ca/rads/puma/ )

Relationship to workflow modeling, transaction processing, BPEL


Conclusions Networking implies different system layers

physical transmission network services and their management distributed applications

There is technology push (higher bandwidth, wireless transmission, computing power) and application pull (after e-mail and WWW: IP telephony and conferencing, VOD, e-commerce, e-society)

There are many interesting topics of research relevant to the future of networking

future of networking, 2006 1 gregor v. bochmann, university of ottawa presentation given at the...

Documents

new applications

certain applications

important applications

variety of applications

escience applications

future of networking

university of ottawa

networking optical networks