ROOM

OO ADL - UML is a relative
Distributed real-time systems with reactive behavior - systems where objects need to respond to events in a timely manner.
Executable - useful for modeling
Two levels of specification - Schematic (architectural) and Detail (implementation)

ROOM basic elements

event - message received at an object (associated name and optional data object).
protocol - ordered sequence of invocation/replies between two objects - protocol classes can be defined - protocol classes can be defined
actor - concurrent object responsible for performing some task.

ROOM protocols

ROOM Modelling Structure

Concurrent Actors - act in parallel with other actors.
Actor types - defined by the set of interfaces.
Actors communicate via ports which represent an instance of a protocol class. Ports are intermediaries between the implementation and environment.
Actors can simultaneously handle multiple different protocols.
Structural Replication of ports/actors

ROOM Actors

ROOM Composition: Binding

Models a communication channel that connects two ports.
``Compatible'' protocols assumed - one must be the inverse of the other.

ROOM Composition: Layer connections

Directed relationships to model situations where there is an asymmetric dependency between two actors. (ex: client/server - the client cannot function without the server)
Often represent many different service access point (SAPSs) to service provision points (SPPs)

ROOM containment: Aggregation

grouping actors - hiding the structure.
Relay ports move messages from external to internal
Multiple containment

ROOM OO features

Actors can be subclasses
Ports can be subclasses

ROOM Dynamic Features

Optional actors - not created until needed. May have slots for imported actors
Substitutable actors - actual actor not instantited until later - must meet the constraints imposed by the interface.

ROOM behavior

Structure and behavior meets at the actor interfaces.
Behavior associated with actors - ROOMcharts based on Harel StateCharts (1984)

RoomCharts

StateCharts

Used for reactive systems - telecommunications, embedded systems, control plants
Multiple views: module, activity, behavior
Behavior based on extension of FSMs and events
- hierarchical structure (AND-/OR-decomposition)
- Variables (incorporated into the events)
Executable - code generation techniques
Dynamic analysis - evaluate scenarios, deadlock, reachability, usage of transitions

StateChart FSMs


Transitions:  alpha [C]/ beta 
   where  alpha  is the triggering event
      C is a conditions that guards the event
       beta  is the action carried out when the transition is taken
Example:
    entered(S)[in(T) and not active (C)]/
       suspend(C); X := Y + 7

Referring To	Events	Conditions	Actions
state	entered(S)	in(S)
`S`	exited(S)
	started(A)	active(A)	start(A)
activity	stopped(A)	hanging(A)	stop(A)
`A`			suspend(A)
			resume(A)
data item	read(D)	D=F	D:=exp
`D,F`	written(D)	D < F
condition	true(C)	D > F	make_true(C)
`C`	false(C)	. . .	make_false(C)
event `E`, action	timeout(E,)	schedule(A,n)
`A`, n time units

ROOM/StateChart References

Selic, B., Gullekson, G., and Ward, P., Real-Time Object-Oriented Modeling, John Wiley & Sons, New York, NY, 1994.
Selic, B., Modeling Real-Time Distributed Software Systems, Proceedings of the 4th Workshop on Parallel and Distributed Real-time Systems, 1996, pp. 11-18.
Harel, D., Lachover, H., Naamad, A., Pnueli, A., Politi, M., Sherman, R., Shtull-Trauring, A., and Takhtenbrot, M., STATEMATE: A Working Environment for the Development of Complex Reactive Systems. IEEE Transactions on Software Engineering, 16,4, April 1990, pp. 403-414.

CHAM -- CHemical Abstract Machine

Developed in theoretical CS for defining a generalized computational model
Chemical metaphor - inherently parallel
A CHAM is specified by defining molecules m , m' . . . (basic elements), solutions S , S' . . . (multisets - states of the CHAM) of molecules and transformations S --> S' dictating the way that solutions can evolve.

CHAM Software Architectures

description of the syntax by which components of the architecture (i.e. molecules) can be represented;
a solution representing the intial state of the architecture;
set of reaction rules describing how the components interact to achieve the dynamic behavior of the system.

The Compressing Proxy

merges gzip utility with HTTP server to improve performance over slow networks
Pseudo-filter is an adaptor for mismatch between Unix i/o and CERN HTTP filters

Problem

Behavior
1. Adaptor passes data to gzip as it receives it.
2. When all data read, Adaptor reads results from gzip
More subtle mismatch:
1. gzip tries to write intermediate results, blocking until it can be received.
2. Adaptor tries to write new information to gzip blocking until it can be received.
3. Deadlock

CHAM Molecules


 M    ::=    P | C | E | M ^  M 
 P    ::=    CFu   |   CFd   |   AD  |   GZ          Components
 C    ::=    i(N) | o(N)     Communication channels
 N    ::=    1 | 2 | 3 | 4 
 E    ::=    eof-i   |   eof-o     Control signals

CHAM Solutions

Initial Solution:


 S0    =    CFu   ^   o(3) ,
            CFd   ^   i(4) ,
            i(1) ^   eof-i   ^ o(2) ^   eof-o   ^   GZ, 
            i(3) ^ o(1) ^   eof-o   ^   AD

Example:

AD is initially in the state of consuming data on channel 3. Then it can produce data on channel 1, generate the appropriate eof and be done.
GZ will initiall get input from channel one until it receive eof. Then it will output on channel 2 and generate eof.

CHAM Laws

General laws
- Reaction Law. An instance of the right-hand side of a rule can replace the corresponding instance on its left-hand side. Given M1,M2,. . .,Mk--> M'1,M'2,. . .,M'l , if m1,m2,. . .,mk and m'1,m'2,. . .,m'l are instances of M{1..k} and M'{1..l} then we can apply the rule to the current solution.
- Chemical Law. Reactions can be performed freely within any solution:
  S --> S' ==> S U S'' --> S' U S''
Architecture Specific Laws
- [ T1 ] == i(x) ^ m1, o(x) ^ m2 --> m1 ^ i(x), m2 ^ o(x)
- [ T2 ] == e ^ m ^ c --> c ^ e ^ m
- [ T3 ] == eof-o ^ m1 ^ o(x), eof-i ^ m2 ^ i(x) --> m1 ^ o(x) ^ eof-o , m2 ^ i(x) ^ eof-i
- [ T4 ] == eof-i ^ m --> m ^ eof-i
- [ T5 ] == GZ ^ m --> m ^ GZ
- [ T6 ] == f ^ c --> c ^ f
- [ T7 ] == AD ^ i(3) ^ m --> i(2) ^ eof-i ^ o(4) ^ AD
- [ T8 ] == AD ^ i(2) ^ m --> i(3) ^ o(1) ^ eof-o ^ AD
where m,m1,m2 in M, x in N, c in C, e in E,f in F

Applying rules


 S0    =    CFu   ^   o(3) ,
            CFd   ^   i(4) ,
            i(1) ^   eof-i   ^ o(2) ^   eof-o   ^   GZ, 
            i(3) ^ o(1) ^   eof-o   ^   AD

 S0 -->{T6} S1 , where
 S1    =     o(3) ^   CFu ,
            CFd   ^ i(4), 
            i(3) ^ o(1) ^    eof-o   ^   AD,
            i(1) ^    eof-i   ^ o(2) ^   eof-o   ^   GZ

 S1 -->{T1} S2 , where
 S2    =    CFu   ^ o(3) ,
            CFd   ^ i(4), 
            o(1) ^    eof-o   ^   AD  ^ i(3) 
            i(1) ^    eof-i   ^ o(2) ^   eof-o   ^   GZ

 S2 -->{T1} S3 , where
 S3    =    CFu   ^ o(3) ,
            CFd   ^ i(4), 
            eof-o   ^   AD  ^ i(3) ^ o(1) 
            eof-i   ^ o(2) ^   eof-o   ^   GZ  ^ i(1)

At this point, T2 , T3 or T4 can occur nondeterministically.