Introduction to MIDAS (Middleware for Composable and Dynamically Adaptable Services)

1
Introduction to MIDAS(Middleware for Composable
and Dynamically Adaptable Services)

• Santosh Shrivastava
• Newcastle University

2

• MIDAS is an IST Accompanying Measure
• One year duration (starting June 02)
• Objectives
• (i) to create a research roadmap for the
  development of the next generation middleware
  capable of supporting services that are
  composable and adaptable within the context of
  large scale systems where quality of service
  issues ranging from fault-tolerance, timeliness
  to security and survivability are of paramount
  importance
• (ii) to build Europe wide partnerships for
  creating an integrated project (IP) in the
  Framework Six Programme (FP6)

3

• MIDAS members
• Santosh Shrivastava (University of Newcastle,
  UK), coordinator
• André Schiper (Ecole Polytechnique Fédérale de
  Lausanne)
• Roy Friedman (Technion - Israel Institute of
  Technology)
• Danny Dolev (The Hebrew University of Jerusalem)
• Dahlia Malkhi (The Hebrew University of
  Jerusalem)
• Roberto Baldoni (The University of Rome "La
  Sapienza")
• Steve Battle (Hewlett Packard Laboratories,
  Bristol)
• Anthony Wiley (Hewlett Packard Laboratories,
  Bristol)
• Jean-Bernard Stefani (INRIA Rhône-Alpes,
  Grenoble)
• Gérard Vandôme (INRIA Rhône-Alpes, Grenoble)
• Executive Director, ObjectWeb open source
  consortium

4

• We assume a ubiquitous computing environment
  where processors will be everywhere and will be
  interconnected by a diverse array of networks,
  from ad-hoc to the global Internet.
• Constructing the software infrastructure -the
  middleware - for ubiquitous computing environment
  poses a number of scientific and technical
  challenges.
• scalability,
• quality of service,
• manageability, and
• programmability.

5

• Scalability
• scaling in machine forms - from smart labels to
  server farms to metacomputing network overlays
  scaling in numbers - objects, machines, users,
  locations scaling in logical and organizational
  structures - from ad-hoc collaborations networks
  to federations of multi-domain enterprises.
• Quality of service (QoS)
• ability to obtain correctness and service-level
  guarantees such as timeliness, availability,
  fault-tolerance, survivability for applications
  executing in large scale environments.
• Manageability
• the ability monitor and control the operation and
  evolution of large scale, long-lived distributed
  applications and services, avoiding manual
  intervention and centralisation
• Programmability
• the ability to compose new applications from
  existing applications and services, deploy and
  maintain them in highly dynamic and heterogeneous
  computing environments

6

• MIDAS workplan
• To consult the RD community
• organise an invited workshop on research
  directions
• Prepare a research roadmap for the scientific
  community
• Based on the above, work towards the formulation
  of an IP
• Within the IP, we expect to address core research
  topics such as
• ? Fault-tolerance and survivability
• ? Performance evaluation
• ? Challenges of supporting mobility
• ? Web services composition through automated
  negotiation
• Operating systems and network support for
  emerging applications
• Trust and security
• ----

7

• Progress so far
• Involved in two EOI (expression of interests)
• EUCOSM European Ubiquitous Component-based Open
  Source Middleware
• UBICOM Ubiquitous computing middleware
• (lack of time prevented integration of the two in
  one)
• Workshop on "Middleware for Composable and
  Adaptable Services", Montreux, 14-15 November
  2002.
• contained sessions on future trends and
  requirements in a number of application areas
  that include Telecommunications, Air Traffic
  Control. Research problems in components, service
  provisioning, networking, Security and trust,
  Dynamic Networks
• Preliminary Research roadmap available at
• http//www.newcastle.research.ec.org/midas/

8
Roadmap

• We have identified research challenges in
• system structure software engineering dynamic
  reconfiguration, programming models, analysis and
  SE tools, system management, manageability and
  composability
• quality of service models, algorithms and tools
  for reliability, fault tolerance, survivability,
  security trust, real-time
• scalable dynamic networks P2P, ad-hoc
  networking, mobility and emerging applications.
• Revised version to be available by March/April 03
• We briefly examine the research challenges in the
  above three areas

9
Research challenges in

• System structure software engineering
• A prime requirement that has emerged in the last
  decade in small and large scale distributed
  systems is the need to build adaptable
  (dynamically reconfigurable) systems to deal with
  a variety of changes (changes in the environment
  of the system, changes in application
  requirements, upgrade and maintenance changes,
  etc). This is a crucial requirement for future
  global computing environment where systems will
  be deeply embedded and potentially long-lived.
• Research challenges in Dynamic reconfiguration
• efficiency issues how to design and build
  dynamically reconfigurable systems without
  imposing stiff run-time performance penalties?
• system integrity how to characterize and enforce
  system integrity constraints so as to ensure safe
  dynamic reconfiguration?
• dealing with failures and scale how to design
  and build large scale, dynamically reconfigurable
  systems in presence of failures?
• The roadmap goes on to discuss further research
  challenges in the area of System structure
  software engineering

10
Research challenges in

• Quality of Services (QoS)
• First and foremost, in order to provide QoS,
  there must be a good definition of the desired
  qualities. The challenge is to provide a good
  specification in a formal way, that is also
  measurable and verifiable in practice. An open
  direction is to devise models capable of
  capturing the probabilistic nature of distributed
  processing.
• Some of the key goals are
• Achieve systems in which QoS guarantees can be
  specified and enforced during unstable periods in
  which failures and recoveries take place
• Provide means for verifying compliance with a
  specification.
• new application areas present new challenges for
  old problems.
• One such domain is that of peer-to-peer (p2p)
  resource sharing applications that exhibit high
  decentralization, scalability and dynamic
  behaviour.
• new technologies and platforms present the
  application builder with system design problems
  that differ from the past ones.
• mobile ad-hoc networks (MANETS), there is no
  fixed infrastructure and all information and
  service must be carried on the fly among existing
  participants

11
Research challenges in

• Scalable Dynamic Networks P2P, ad-hoc
  networking, mobility and emerging applications.
• Recent advances Overlay networks, peer-to-peer
  Networks, Mobile Ad-hoc Networks (MANET), Highly
  Dynamic Environments, Sensor networks, Grid
  systems
• Challenges
• Approach Completely decentralized vs. core based
  solutions vs. hybrid solutions
• not clear which is the right approach to
  developing middleware systems for such dynamic
  environments.
• Large scalable lookup/discovery services
• Geographical and proximity based mapping
• Existing lookup and discovery methods ignore
  issues of geography and proximity.
• Relationship between routing/networking and
  middleware level

12
Research challenges

• MIDAS members have contributed to the workshop on
  future directions in distributed computing held
  in Bertinoro, June 02, organised by Ozalp
  Babaoglu (Univ. of Bologna), Ken Birman (Cornell
  Univ.) and Keith Marzullo (UC, San Diego).
• Proceedings based on the talks to be published
  soon by Springer, LNCS.

13
formulation of an IP

• We expect the IP to contain
• Research tier long term research challenges
• Technology tier Advanced development activities
  integration of software development around the
  ObjectWeb code base and ObjectWeb consortium
• Application tier integration of software
  development around application specific platforms

14

• Example Research and technology tiers work on
  Trust and security
• Trust is a vital component of every business
  interaction
• Trust Firm belief in the reliability or truth or
  strength of an entity
• Customers must trust that sellers will provide
  the services they advertise
• Sellers must trust that the buyer is able to pay
  for goods or services
• if such direct trust does not exist then you need
  intermediaries trusted by both (trusted third
  parties) through which the interaction can take
  place

15

• Practical example Open Buying on the Internet
  (OBI) protocol the specification defines the
  following trust model (excerpts)
• The validity of digital certificates is
  fundamental to this trust model since
  certificates establish the identities of
  requisitioners and machines during
  authentication. The trust between two trading
  partners is another fundamental aspect of this
  model. Highlights of the trust model are outlined
  below
• Trading partners trust each others certificates
  and Certificate Authorities
• Certificates are issued by trusted CAs who
  operate secure environments. Both trading
  partners trust the CAs to issue certificates
  securely.
• The certificate (more specifically, the
  associated private key) used to identify an
  individual is stored securely within the browser
  software on the individuals desktop computer and
  is protected from access by others.
• The Selling Organization trusts the Buying
  Organization to
• insure that OBI orders are only submitted on
  behalf of authorized purchasers. For example, a
  former employee may still have access to a
  previously issued digital certificate that allows
  access to selling organization catalogs.

16

• There is a need for high-level way of specifying
  and managing trust and translating these into
  right set of trust services
• identity service
• Authorisation service
• Trust rating and recommendation services
• Auditable receipt generation
• Timestamp generation
• Guaranteed message delivery