InterProcess Communication Models

1
Inter-Process Communication Models

• September 23, 2004
• Myeong-Wuk Jang
• Open Systems Laboratory

2
Outline

• Introduction
• Linda
• JavaSpace
• ActorSpace
• Directory Facilitator
• ATSpace in the extended AA
• Conclusion

3
Motivation

• Parallel and Distributed Computation
• Process Communication
• Reference-based Communication
• Channel (or Link)
• Binding of Time

message
sender
receiver
4
Motivation

• Problems
• Open Systems
• Processes (Actors or Agents) can enter and leave
  at any time
• Agent Mobility
• The location of an agent cannot be predefined
• Indirect Inter-Process Communication
• Data-oriented Communication
• Attribute-based Communication

5
The Linda Model

• David Gelernter
• Yale University
• Data-oriented communication model
• Concurrent process interaction model
• Virtual shared memory model
• A model of process creation and coordination
• Tuple Space (TS)
• Tuples

6
Tuple Space Tuples

• Tuple Space (TS)
• Virtual shared memory
• Abstract communication environment
• Pattern matching on the tuples
• Tuples
• A typed sequence of data
• Tuple Template tuple
• that has formal parameters

(a string, 15.01, 17, another string)
(a string, ? f, ? i, another string)
7
Linda Basic Primitives

• out (write)
• generate a tuple and add it to Tuple Space
• out(a string, 15.01, 17, another string)
• out(0, 1)
• in (take)
• cause a search in Tuple Space, remove a matching
  tuple from the space, and return the tuple
• in(a string, ? f, ? i, another string)
• rd (read)
• like in, but do not remove the tuple.
• rd(a string, ? f, ? i, another string)

8
Example - I

• out(a string, 15.01, 17, another string)
• rd(a string, ? f, ? i, another string)
• in(a string, ? f, ? i, another string)

TS
out(a string, 15.01, 17, another string)
(a string, 15.01, 17, another string)
9
Example - I

• out(a string, 15.01, 17, another string)
• rd(a string, ? f, ? i, another string)
• in(a string, ? f, ? i, another string)

TS
rd(a string, ? f, ? i, another string)
(a string, 15.01, 17, another string)
f 15.01 i 17
10
Example - I

• out(a string, 15.01, 17, another string)
• rd(a string, ? f, ? i, another string)
• in(a string, ? f, ? i, another string)

TS
in(a string, ? f, ? i, another string)
(a string, 15.01, 17, another string)
f 15.01 i 17
11
Linda Additional Primitives

• Extended Primitives (non-blocking operators)
• inp
• rdp
• eval
• generate active tuple process
• eval(M, i, j, compute(i, j))

12
Example - II

• eval(fac, 7, fac(7))
• rd(fac, 7, ? value)

TS
eval(fac, 7, fac(7))
(fac, 7, 4020)
(fac, 7, fac(7))
13
Example - II

• eval(fac, 7, fac(7))
• rd(fac, 7, ? value)

TS
rd(fac, 7, ? value)
(fac, 7, 4020)
value 4020
14
Example - III

• matrix_multiply(X,Y,Z)
• double Xnn, Ynn, Znn
• matrix_transpose(Y, Y)
• for (i0 i lt n i)
• for (j0 j lt n j)
• ID in j
• eval("dot", ID, dot_product(Xi, Yj))
• for (i0 i lt n i)
• for (j0 j lt n j)
• ID in j
• in("dot", ID, Zij)

15
Characteristics

• A few operators
• Easy to use
• eval operator for parallel operations
• A data-oriented communication model
• Pattern matching on tuples (tuple signature)
• A sender does not know anything about receivers
• A receiver does not know anything about the
  sender of a tuple

16
JavaSpace

• Sun Microsystems
• A reliable distributed storage system for objects
• A distributed persistence and object exchange
  mechanism
• JINI and RMI (Remote Message Invocation)

17
Entry and Template

• Entry
• A typed group of objects
• Template
• An object that has some or all of its fields set
  to specified values

18
Operators

• write
• write an entry
• read
• read an entry
• take
• read an entry, and remove it from the space
• notify
• notify a specified object when an entry is written

19
Characteristics

• Rich typing
• No eval operator
• A data-oriented communication model
• pattern matching of entries
• A reliable distributed storage system
• Distributed object persistence

20
ActorSpace

• Christian J. Callsen
• Aalborg University, Denmark
• Invocation and broadcasting
• Abstract specification of a group of recipients

21
Key Terminology

• actorSpace
• A computationally passive container of actors
• A context for matching patterns
• Attribute
• An abstract external description or view of an
  actor or an actor group
• Capability
• Key for secure access control

22
Operations

• new-capability()
• create-actorSpace(capability)
• send(dest, parameter)
• broadcast(dest, parameter)
• make-visible(?, attributes_at_space, capability)
• make-invisible(?, space, capability)
• change-attribute(?, attributes_at_space, capability)

23
Example

• behavior counter
• int count
• method _at_countit()
• count
• method _at_reprotcount()
• send(system_at_print, count)
• send(system_at_print, \n)
• method _at_main()
• capability c new-capability()
• name a1 new actor(c)
• name a2 new actor(c)

name a3 new actor(c) name s new
create_actorSpace(c) init_behavior(a1,
counter, 0) init_behavior(a2, counter,
0) init_behavior(a3, counter,
0) make_visible(s, foos_at_/,
c) make_visible(a1, foo1_at_s,
c) make_visible(a2, foo2_at_s,
c) make_visible(a3, foo3_at_s,
c) send(/foos/_at_countit) broadcast(/foos/
foo12_at_reportcount)
24
Example
Actor Platform
make-visible
actor1
foo1
make-visible
send
broadcast
make-visible
actor2
actor4
foo2
foo3
make-visible
actor3
25
Characteristics

• Secure communication
• Regular expression in template
• Nested actorSpaces
• Cycle problem
• A attributed-oriented communication model
• pattern nickname of an actor
• A brokering service model

26
Attribute-oriented Communication Model

• A sender may not know anything about receivers
• The actual number of receivers and their
  locations are hidden
• A receiver may not know anything about the sender
  of a message
• Multicast (or broadcast) mechanisms are easily
  implemented

27
Directory Facilitator

• FIPA Foundation for Intelligent Physical Agents
• A yellow page directory service for agents

28
Operators

• register
• enter data about an agent and its method(s)
• deregister
• remove data
• modify
• change the contents of registered data
• search
• cause pattern matching to find service agents

29
Example
Agent Platform
DF
a1, fun
a2, sfun
register
a3, fun
register
agent1
register
agent2
agent4
agent3
30
Characteristics

• A matchmaking service model
• A sender uses information on a virtual shared
  space to find receiver agents.
• A sender knows names of receivers
• A attributed-based agent name resolution
• pattern methods (services) of agents

31
Middle Agent Services

• Enable discovery or access to previously unknown
  agents
• Examples
• Matchmaking Service (Yellow Page Service)
• cf. Directory Facilitator in FIPA
• Brokering Service
• ActorSpace

32
Problems

• Expressiveness limitation
• Search algorithm is fixed
• Parameters of a template are predefined
• Incomplete Information
• Tuples in the middle agent may not have
  sufficient information for query matching
• ATSpace Active Tuple Space
• Open Systems Laboratory

33
Brokering Service
Tuple Space
Get models of offered computers?
UAN2, Champaign, computer, 950
UAN3, Urbana, computer, 650
UAN2, Urbana, printer, 120
UAN4, Austin, computer, 1290
UAN1buyer
UAN3seller
UAN4seller
UAN2seller
34
Matchmaking Service
Tuple Space
Get models of offered computers?
UAN2, Champaign, computer, 950
UAN3, Urbana, computer, 650
UAN2, Urbana, printer, 120
UAN4, Austin, computer, 1290
UAN1buyer
UAN3seller
UAN4seller
UAN2seller
35
A Challenging Query
Tuple Space
Get models offered by best two sellers within 20
miles from my city?
UAN2, Champaign, computer, 950
UAN3, Urbana, computer, 650
UAN2, Urbana, printer, 120
UAN4, Austin, computer, 1290
UAN1buyer
UAN3seller
UAN4seller
UAN2seller
36
A Challenging Query
Tuple Space
Get models offered by best two sellers within 20
miles from my city?
UAN2, Champaign, computer, 950
UAN3, Urbana, computer, 650
Translate into an easy query
UAN2, Urbana, printer, 120
Filter the result
Continue as before
UAN4, Austin, computer, 1290

• Inefficient ad-hoc solution
• The translated query results in a large amount
  of tuples sent back
• Can we do better?

UAN1buyer
UAN3seller
UAN4seller
UAN2seller
37
The ATSpace Approach
ATSpace
Get models offered by best two sellers within 20
miles from my city?
UAN2, Champaign, computer, 950
UAN3, Urbana, computer, 650
UAN2, Urbana, printer, 120
UAN4, Austin, computer, 1290
UAN1buyer
UAN3seller
UAN4seller
UAN2seller
38
Architecture of ATSpace
ATSpace
Tuple Space
message queue
tuple
msg
msg
msg
tuple
return names (matchmaking)
tuple
Tuple Space Manager
tuple
ask service agents (brokering)
39
ATSpace

• Common Shared Space
• write, read, take
• writeAll, readAll, takeAll
• Middle Agent Services
• matchmaking services search, searchAll
• brokering services deliver, deliverAll
• Active services mobile objects transmitted for
  remote search
• Service Call Message Generation Services
• Service call messages may be different for
  matched agents and generated by the supplied
  object

40
Security Issues

• Data Integrity
• Prevent modification of the original data
• Deep copy of data
• Denial of Service
• Consumption of all processor cycles
• User-level thread scheduling
• Illegal Access
• Unauthorized access
• Access keys

41
Extended Architecture of ATSpace
42
Characteristics

• Common Shared Space
• Middle Agent
• Brokering Matchmaking Services
• Active Brokering Matchmaking Services
• Execution Environment for Mobile Objects

43
Summary

• Process / Actor / Agent Communication
• Reference-based Communication
• Attributed-based Communication
• Attribute-oriented Communication Model
• Linda
• JavaSpace
• ActorSpace
• Directory Facilitator
• ATSpace
• Middle Agent / Actor Services
• Matchmaking Service
• Brokering Service

44
The End

• Thanks for your attention.

45
References

• D. Gelernter, Generative Communication in Linda,
  ACM Transactions on Programming Language and
  Systems, 7(1) 80-112, 1985.
• N. Carreiro, and D. Gelernter, Linda in
  context, Communications of the ACM, Vol. 32, No.
  4, pp. 444-458, 1989.
• Sun Microsystems, JavaSpacesTM Service
  Specification, version 2.0, June 2003.
  http//java.sun.com/products/jini/specs
• C. Callsen and G. Agha, Open Heterogeneous
  Computing in ActorSpace, Journal of Parallel and
  Distributed Computing, Vol. 21, No. 3, pp.
  289-300, 1994.
• Foundation for Intelligent Physical Agents,
  SC00023J FIPA Agent Management Specification,
  December 2002. http//www.fipa.org/specs/fipa00023
  /
• M. Jang, A. Abdel Momen, and G. Agha, "ATSpace A
  Middle Agent to Support Application-Oriented
  Matchmaking and Brokering Services," IEEE/WIC/ACM
  IAT(Intelligent Agent Technology)-2004, September
  20-24, Beijing, China, 2004.