JINI Spontaneous Networking in Java

1
Java User Group Torino
JINISpontaneous Networking in Java
Second Part JINI Concepts
JUG_at_Webbit, Padova 9-10-11 Maggio 2003
2
Topics

• The state of the art
• JINI ideas
• Possible scenarios
• JINI concepts and terminology

3
Objectives

• Understand the value of Ubiquitous Computing
• Understand the fallacies of classical distributed
  computing
• Understand the key JINI concepts

4
About the JUG

• The JUG was born in February 2002 by the idea of
  some Java fans in the Torino area
• The aim of the group is to promote the knowledge
  exchange between members and to create an
  important community in the landscape of Italian
  Java programmers
• Participating is free and voluntary
• You can participate at the meeting, write
  articles, expose your doubts, ask for support etc
  etc .... .... ....

5
Who am I?

• Renzo Borgatti is a Java developer since 2000
• I was involved in several projects during my
  activity mainly focused on server side Java and
  J2EE
• I spend the rest of my time (too short!) playing
  piano, running and, of course, partecipating at
  the JUG activities!
• Feel free to contact me at renzo.borgatti_at_jugtorin
  o.it

6
Distribuited ApplicationsThe State Of The Art
7
WW-What?
The State Of The Art

• The first distributed applications built, used to
  live inside a very constrained network space
• There were few subnetworks in the world that
  couldn't talk to each other
• No browsers or world wide TCP/IP connections were
  available to access services from home

8
No micro-devices
The State Of The Art

• Computers were huge and unmovable
• Few years later the Home Computer appears
  smaller but not enough
• No other devices with computational power (such
  as phones, PDAs or laptops)

9
Some years later...
The State Of The Art

• Laptops
• World Wide Web
• Widely adopted communication protocol (TCP/IP)
• HTTP and browsers
• Electronic wearable devices
• ... but still the same old-fashioned distributed
  programming model.

10
Old-Fashioned?
The State Of The Art

• Http programming
• Text oriented vs Object Oriented
• No services on the browser side
• Well known addresses
• Distributed applications
• 7 fallacies

11
The Seven Fallacies of Distribuited System
The State Of The Art

• The network is reliable
• Latency is zero
• Bandwith is infinite
• The network is secure
• Topology doesnt change
• There is one administrator
• Transport cost is zero
• By Peter Deutsch

12
The JINI idea
13
The vision
The JINI idea

• Every processing capable hardware is potentially
  a platform to run a JVM
• Java interfaces expose remote services in the JVM
• Sometimes the platform stands still sometimes it
  moves network changes but applications don't
  break
• The infrastructure should be built using existing
  technologies and connections
• Networked applications should be reliable and
  self-repairing

14
Why not using CORBA-RMI?
The JINI idea

• They can't run on a PDA the next network they
  reach, where's the registry? And on what port the
  ORB is listening to?
• With CORBA no code mobility
• Remember the seven fallacies?
• To be reliable an application should use many
  tricks heartbeats, callbacks, multicasting

15
Home automation scenarios
The JINI idea

• Appliances that maintain themselves through
  remote diagnostics and repair procedures
• Water heaters that adjust temperature and water
  pressure to compensate for someone turning on the
  dishwasher while you're taking a shower
• Use the TV remote as a control for your home
  security system, or to turn off the ringer on the
  telephone.

16
In the car scenario
The JINI idea

• You enter a car (not necessarily yours)
• When a call arrives, the phone knows you are
  driving, turn down the radio volume and display
  caller information on the GPS screen
• The phone use the car speakers as an hands-free
  system
• If an SMS arrives, the phone use TFT screen of
  the GPS system to display it
• If the call is important, you might decide to
  register it on your PDA as an MP3 file for later
  use

17
At the airport scenario
The JINI idea

• Imagine arriving at an airport, suitcase in one
  hand, Palm PDA in the other.
• The airport has a radio-based proximity-specific
  communications network, which your Palm PDA, with
  the aid of Jini technology, spontaneously
  discovers and joins.
• You are instantly notified of your flight's
  departure gate, and your name is added to the
  list of checked-in passengers--no waiting in line.

18
Other scenarios
The JINI idea

• Clustering adding computing power without
  changes to the existing configuration at run-time
• JMX manageable applications become available at
  startup

19
JINI Concept
20
Services
JINI CONCEPT

• Services participate in a JINI community exposing
  functionalities
• Their initial target is to register themselves to
  be publicly available
• They could be remote services (RMI or whatever)
  or they can use clients CPU

21
Clients
JINI CONCEPT

• Clients participate in a JINI community searching
  and using services
• Their initial target is to find a lookup service

22
Lookup Service
JINI CONCEPT

• A lookup service is a kind of RMI registry with
  more powerful capabilities
• It partecipates in discovery to find services and
  other lookup services
• Is itself a JINI service (without the need to
  register itself)

23
Discovery
JINI CONCEPT

• Discovery is the early part of the JINI protocol
• It defines how clients, services and lookup
  services behave to find each other
• Addresses and ports
• Structure of bytes to be sent
• Network protocols
• Interfaces involved

24
Join
JINI CONCEPT

• Join follows the discovery phase and defines
• How a client can find services using a lookup
  service
• How a service can register to a lookup service

25
JINI Groups
JINI CONCEPT

• JINI Community the group of services registered
  with a lookup services at a given time
• JINI Federation a group of interconnected JINI
  community obtained by registering a lookup
  service as a normal service into another lookup
  service

26
Lookup by Interfaces
JINI CONCEPT

• Clients lookup services using Java interfaces,
  ServiceID and Attributes as opposed to String in
  RMI
• Strings can be easily changed in different
  deploys
• Interface uniqueness is guarantee at compile
  time. No doubled-registration
• Powerful queries via Attributes

27
Leasing
JINI CONCEPT

• Clients and services that want to use JINI
  resources must explicitly declare how many time
  they intend to use them
• When the time expires, all resources are freed
• To continue using the resource, the lease must be
  renewed before it expires

28
Remote events
JINI CONCEPT

• JINI gives the community an asynchronous message
  notification system
• Components may register to receive remote events
• No temporal dependencies between parties
• Events participate in the leasing model of using
  resources

29
Transaction
JINI CONCEPT

• JINI supports the two phase commit transaction
  model
• There are helper and libraries to coordinate
  transaction participants which could be
  potentially remote

30
JINI Services standardization

• JINI services, to be effective, must agree on a
  well defined operational interaction. Similar
  services should always
• Have the same interface (and Java interfaces
  guarantee it)
• Have the same exceptions meaning, pre and
  post-conditions rules ecc
• Efforts have been made for banking, insourance
  and automotive industries
• Different vendors must agree on interfaces and
  business behaviors producing JINI services
  accordingly

31
JINI and heuristics

• What if you dont know exactly the service you
  are looking for? You dont have the interface but
  you know the description of the service
• You should use some kind of heuristic to find a
  best fit
• The client tries to find the best service

32
Service Interface Registry 1/2

• A JINI service whose responsibility is to find
  the best interface for a service given some
  parameters
• The client downloads a general class, with a
  single invoke(String methodName) method
• The client invoke a method using an
  approximation something starting with Y,
  containing the word X or whatever. If a match
  is found then that method is called. Otherwise
  there is a collision (or not methods at all that
  match the criteria)

33
Service Interface Registry 2/2

• The principle here is a kind of human reasoning
  approximation
• The algorithm search for a best fit given the
  name of an interface and then the name of a
  method to call on that interface
• The result is a standard Field wrapping

34
Resources