1. Common Application Components
Some frequently needed elements in applications that involve communication between separate systems

2. Common Elements of the Application Layer
Some things have to be done in all applications
Others are very common
Having a bag of tools handy can save recreating a solution every time the problem arises

3. Base set of common services
Establish contact
Remote execution
Reliable transfer of data
Coordination of distributed processing elements

4. Establishing contact Locate the processing partner
Associate semantics with messages exchanged
Understand network service level
Match processing elements
Identify initiator of the contact

5. ACSE Association Control Service Element
Two types of services offered:
Association establishment
Association termination
release
abort by the user
abort by the lower layer network services

6. A-ASSOCIATE Association establishment
parameters describe the set of desired characteristics (the context) of the association
what application service elements will be active in the cooperating systems
identification of the destination process location for communication
OSI PSAP (Presentation Service Access Point)
TCP/IP port number
Presentation context
how should the receiver interpret data that arrives
{1{PersonnelData, Encrypted}, 2{Annual Report, Compressed}, 3{AnnualReport,Basic}}
The example lists the set of abstract syntaxes and associated transfer modes to available be used in communication between these processes.
Context identifier 2 in the Presentation Context Definition List …
Optional -- user data. May be used for login purposes
Optional -- QOS specification O

7. A-RELEASE, A-ABORT
A-RELEASE: Orderly termination of an association between cooperating processes
A-ABORT: Abrupt termination, with possible loss of data

8. Closing an Association
Confirmed service
Unconfirmed service

9. Communication between layers
Peer to Peer
Service Requestor /
Service Provider
Service Requestor /
Service Provider
Peer to Peer
Service Requestor /
Service Provider
Service Requestor /
Service Provider

10. Within layers Peer to Peer Peer to Peer Service Requestor /
Service Provider
Service Requestor /
Service Provider
Peer to Peer
Service Requestor /
Service Provider
Service Requestor /
Service Provider

11. Parameters
Information shared between peers and between service requestors and providers
Each invocation of a service, or response to a service request, involves information passed both within the local stack and with peers in a cooperating stack.
The single list of parameters must be split into those that inform the service provider and those that inform the peer process.

12. ASSOCIATION ESTABLISHMENT

13. System state regarding protocols
A system is in one of a number of states at any given time.
The state depends on what conditions are in effect, what has happened so far.
The state determines what are possible responses to future events
Initial state (Idle)
limited set of events are acceptable
others are considered error conditions
For example, the arrival of data before an association is established would be an error condition

14. Protocol machine Key: Event Response The event occurs, the
protocol machine changes
state and performs
the indicated response.
Here the response is
something sent or issued.
Two views: top shows the state transition in graphical form;
bottom shows the layer interaction.

15. Association Establishment Protocol Machine
Idle
A-ASSOCIATE.req
AARQ
AARQ
A-ASSOCIATE.ind
Outgoing
association
pending
Incoming
association
pending
AARE
A-ASSOCIATE.conf
A-ASSOCIATE.resp AARE
Association
Established

16. Association Established
Once Association is Established
the context of the interaction is established
data can be exchanged
an orderly termination of the association can be done
ACSE is the OSI protocol for establishing an association.
Other stacks, like TCP/IP, do not have a named entity for this purpose, but must establish this sort of connection

17. Remote Operation
Call for the execution of an operation on a remote system
Basic questions related to remote operation
What do we know about the data to be used?
How will the initiator know when the remote operation is complete?
Is it safe to repeat a request if the remote system crashes before completing?
Stop and wait or continue working?
Do we have to send data? Will the remote system understand the data the same way the requestor does?
Completion: If a result is expected but has not arrived, it may mean that the system is still executing or that it has crashed
Idempotent operation: Bank balance examples

18. ROSE; remote procedure calls
What do we know about the data to be used?
ASN.1 and BER or alternatives such as XDR
How will the initiator know when the remote operation is complete?
Requesting some result from every operation allows the initiator to know when the operation is done
Is it safe to repeat a request if the remote system crashes before completing?
CCR or its equivalent
Stop and wait or continue working?
Synchronous or asynchronous invocation

19. Remote operation services
In the OSI protocols,
ROSE - The Remote Operation Service Element
Other,
RPC - Remote Procedure Call
RMI - Java’s Remote Method Invocation

20. ROSE service RO-INVOKE RO-RETURN-RESULT RO-RETURN-ERROR
call a method located in another system
have it run in the other system
RO-RETURN-RESULT
send a result to the invoking method
RO-RETURN-ERROR
send an error notification to the invoking method
RO-REJECT-U/RO-REJECT-P
reject the invocation request

21. ROSE characteristics Minimal state information
All services unconfirmed
All responsibility for correctness is on the designer of the application

22. Synchronous or Asynchronous
blocking
Asynchronous
non-blocking
report results and errors
report errors only
report results only
report nothing
Not all operations have results that need to be reported. Example, request to delete a file or reset a printer or display a message -- these do not require that a result be sent back to the initiator.
Good practice: get a result so you know what happened.
Synchronous operation: RO-RESULT or RO-ERROR must follow and RO-REQUEST

23. ASN.1 definition of a remote operation

24. Using the definition RO-INVOKE (…,0, “PS”…)
other required parameters identify remote system to be contacted, etc.
0 is the identifier of getcount; PS is the name of the queue to examine.
RO-RETURN-RESULT (“PS”,250)
Queue PS has 250 entries
RO-RETURN-ERROR (1, “PS”)
Queue PS is not available

25. Remote Operation Summary
Protocols cover only the communication between the caller and the called method
how method is identified
what information is provided
what is understood between them
Separate issue
write the method to carry out the desired action

26. Reliable transfer of information
What is needed beyond a dependable transport layer?
Protect against errors that occur after the data is delivered to the machine
example: Jammed printer or a disk failure prevents completion of the desired action, but the transport layer is satisfied

27. RTSE: Reliable Transport Service Element
Invokes services of the session layer, uses ACSE
Services provided by RTSE:
RT-OPEN
RT-TRANSFER
RT-TURN-PLEASE
RT-TURN-GIVE
RT-CLOSE
RT-U-ABORT/RT-P-ABORT

28. RT-TRANSFER, RT-TURN confirmed service
parameters specify data to be sent, maximum time allowed for completion
If two-way data transfer, RT-TURN-PLEASE, an unconfirmed service, requests the opportunity to send data; RT-TURN-GIVE, unconfirmed service to yield turn

29. RTSE protocol Use A-ASSOCIATE to establish connection with peer
Invoke turn-taking facilities of the Session layer
Invoke activity management facilities of the Session layer to establish checkpoints between data segments
used by X.400 mail system and available to ROSE when a lot of data must be moved

30. Commitment, Concurrency, and Recovery
assurance that the server process will carry out a request regardless of difficulties that might arise
Concurrency:
protection from the intrusion of interleaved operations that interfere with correct completion of a requested task
Recovery:
establishment of procedures to overcome failures by one or more participating process during execution of a distributed application

31. The Lost Transaction Problem
Process 1
Credit Limit
Process 2
5,000
Read credit
5000
Read credit
5000
Subtract new purchase
(850) update credit
4,150
Subtract new purchase
(54.50) update credit
4,945.50
Time

32. CCR Services Bracket atomic action C-BEGIN, C-PREPARE
SERVER --> CLIENT ready
C-READY
SERVER --> CLIENT no
C-REFUSE
CLIENT, SERVER agree to go on
C-COMMIT
CLIENT --> SERVER no
C-ROLLBACK
CLIENT/SERVER go back
C-RESTART

34. Multiple servers participate

35. Server refuses request

36. Some TCP/IP handy tools
Not exactly the same kind of thing, but useful nonetheless
ping
sends a message and looks for a response to determine that the other machine is alive and active
traceroute
displays the path taken by messages between two cooperating systems
on sun cluster: /usr/sbin/ping or traceroute
NOTE-- overuse of these is considered very unfriendly behavior. Be gentle, considerate.

37. Overall summary
Net-centered computing has more elements than computing on a single machine.
Many characteristics are the same from one application to another
Understand what is the same and what varies between applications