DIDA, a Distributed Discovery Architecture for Grid Environments

1
DIDA, a Distributed Discovery Architecture for
Grid Environments

• Carlos de Alfonso
• Miguel Caballer
• Vicente Hernández
• Universidad Politécnica de Valencia

2
Index

• Index
• Background
• eGovernment framework
• Administration features
• The citizen in the administration
• Grid Technology
• gCitizen project
• Discovery Architecture
• Service Naming
• Service Discovery
• Service Discovery in GT4
• Design of a Service Discovery System
• DIDA
• DIDA Architecture
• Distributed Discovery Services
• FADA Topology

3
Background

• Administration
• Organized in different units and levels
• State
• Region
• Local
• Different specialised unit in each level.
• Despite its heterogeneity the citizen should have
  a unique vision.
• Should not have to know the internal operation
  complexity
• Simple and effective interaction

4
Background

• The Citizen in the Administration
• Citizen is the aggregation of information
  belonging to the administrative units.
• This data dispersion produces some problems

• Not consolidated data
• Different names for semantically equal data
• etc.

5
Background

• Features of the information in the Administration
• Geographically distributed
• Different units located in different places.
• Security
• Only some people is allowed to read or modify the
  data.
• Privacy
• The information must be protected on access and
  transmission.
• Ubiquitous access
• Despite the geographical distribution the access
  must be granted.
• Interoperability
• Among systems, for enabling the collaboration.
• Scalability
• The system must enable the addition of new
  components efficiently.

6
Grid Technology

• Grid Technology Provides some of the required
  features
• Security, Transparency, Ubiquitous access,
  Scalability,
• GT4 implementation of the WSRF standard selected
• Not all the needed functionality is provided
• The service oriented programming model enables to
  develop some services over the middleware.
• Log System
• Service Discovery System
• Distributed Authorisation

7
gCitizen Project

• Creation of a grid middleware for eGovernment.
• Flexible and scalable prototype of a grid
  citizen

• System integration project using grid technology.
• Not invasive for the current systems.

8
Discovery Architecture

• Main issue Discover services
• System visibility
• Any service should be searchable
• Connectivity failure detection
• Components
• Service Naming
• Identify the services in the system
• Discovery System
• Find the services using the naming defined

9
Service Naming

• WSRF uses URI as service identifier
• Location dependent.
• Redundancy
• Not transparent
• Changing the location of service implies a
  identifier change.
• Not enable Mobility
• Needed Features
• Location independent
• Show the hierarchy of the administration
• Levels
• Departments

10
Service Naming

• Design of a Service Naming system based on DNs
• DNs are used to identify users and host in grid
  environments
• Location independent
• Service Mobility
• A service can change its location maintaining the
  DN
• Transparent Redundancy
• Different physical services can have the same DN
• Represented as a Resource Property
• The hierarchy of the DN will place the service
  inside the context of the gCitizen system.

11
Service Naming

• Schema proposed
• Country Name (C)
• Region Name (A)
• State or Province Name (ST)
• Locality (L)
• Organization Name (O)
• Organizational Unit Name (OU)
• Common Name (CN)
• Example
• /CES/AC. Valenciana/STValencia/LValencia/OAyt
  o. Valencia/OUInformática/CNPadrón

12
Service Discovery in GT4

• Based on the use of Index Services
• Services deployed that enable to index other
  services
• Search by means of Resource Properties
• Published in XML
• Using XPath
• Hierarchical structure

• All the information only accessible on top level
  IS.
• Not ubiquitous access
• Central point of failure

13
Design of a Service Discovery System

• Objectives
• Completely Distributed
• Not use a central point of information
• Ubiquitous access
• Enable to access the information from different
  points.
• Transparency
• All the possible complexity of the system must be
  hided to the user.
• Scalability
• Enable to add components to the system easily
• Fault tolerance
• The failure of one component must not affect the
  whole system.
• Detection of the failure of one component

14
Distributed Discovery Architecture

• Built up on the Globus Toolkit 4 Index Services
• Services are indexed in the ISs
• Every entity will have an IS
• Use XPath as query language
• Link the ISs to create a vision of a unique
  Virtual Server
• Avoiding cascade node organizations

15
DIDA Architecture

• Use FADA to create the view of a virtual server.
• Distributed directory of services
• The ISs are registered into a set of FADA nodes.
• FADA nodes distributed across the system.
• Each FADA node know about its neighbours.
• Fault tolerance

16
FADA Network Topology

• The FADA topology is Scale-Free network
• Few highly connected hubs
• Affect the whole system
• Many nodes with few connections
• Not affect whole system

• Proposed a 2-Regular topology
• 2 kind of links
• Strong create the 2-Regular graph.
• FADA node know 2 neighbours
• Weak some additional links.
• If one node fails the connectivity of the system
  is assured.
• The fault of one component is detected

17
Distributed Discovery Services

• Entry points to the discovery system
• Every entity has to deploy the DDS.
• Ubiquitous access
• All the users of the entity must know the
  location.
• Implemented as WSRF services
• Uses XPath as query language
• Knows some FADA nodes
• Transparency
• The client query the service and get the results.

18
Distributed Discovery Services

• UML Interaction Diagram of DDS

19
Conclusions

• eGovernment
• New field for grid technology.
• Grid as integration information system.
• Used GT4 as WSRF implementation
• Provides some interesting features, but have some
  lacks.
• New service naming based on DN
• Location independent
• Discovery Service
• Based on IS of GT4 Discovery Service
• Linked using FADA network to create a virtual
  server
• Totally distributed without a central point
• Use a 2-Regular topology
• More topologies must be studied
• Cache of neighbour services