Communication
Chapter 2
Layered Protocols (1)
Physical: voltage, transmission rate,
mechanical, functional, and
electrical specs,…e.g. RS-232-C
Data link: reliable link-level
communication (framing, CRC)
e.g. HDLC
Network: routing e.g. IP, ATM VC
Transport: reliable end-to-end
communication e.g. TCP, UDP
Session: checkpointing,
synchronization
Presentation: data formats, codes
Application: FTP, HTTP, SMTP,
SNMP etc.
2-1
Layers, interfaces, and protocols in the OSI model.
Layered Protocols (2)
2-2
A typical message as it appears on the network.
Data Link Layer
2-3
Discussion between a receiver and a sender in the data link layer.
Client-Server TCP
2-4
a)
b)
Normal operation of TCP.
Transactional TCP.
Middleware Protocols
2-5
RPC, ROI, Messages, Streams
An adapted reference model for networked communication.
Conventional Procedure Call
Call: count = read(fd, buf, bytes);
a) Parameter passing in a local procedure call: the stack before the call to read
b) The stack while the called procedure is active
Issues: call-by-value/reference or copy/restore?
Client and Server Stubs for RPC
Principle of RPC between a client and server program.
Steps of a Remote Procedure Call
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
Client procedure calls client stub in normal way
Client stub builds message, calls local OS
Client's OS sends message to remote OS
Remote OS gives message to server stub
Server stub unpacks parameters, calls server
Server does work, returns result to the stub
Server stub packs it in message, calls local OS
Server's OS sends message to client's OS
Client's OS gives message to client stub
Stub unpacks result, returns to client
Passing Value Parameters (1)
2-8
Steps involved in doing remote computation through RPC
Passing Value Parameters (2)
a)
b)
c)
The little numbers in boxes indicate the address of each byte
Original message on the Pentium (little endian, left to right)
The message after receipt on the SPARC (big endian)
SPARC would read: 5.1024 (incorrect) and “JILL” (correct)
The message after being inverted.
SPARC would read: 5 (correct) and “LLIJ” (incorrect)
 More information is needed
Passing Reference Parameters
Reference parameters like pointers are more complicated to handle
Solution: Use call-by-copy/restore instead of call-by-reference
However: Only useful for simple referenced data
Not useful as a general solution like when using pointers to pointers
or pointers to complex data structures
The server stub may request the data when they are needed
Distinction between input and output parameters can raise performance
Parameter Specification and Stub Generation
RPC protocol agreements include questions like:
How many bytes for each data type? (here 4)
Big or little endian?
Data representation? (e.g. 1’s or 2’s complement, IEEE 754)
Transport service: connectionless or connection-oriented?
Stub Generation:
Using and IDL compiler
Doors
The principle of using doors as IPC mechanism.
Asynchronous RPC (1)
2-12
a)
b)
The interconnection between client and server in a
traditional RPC
The interaction using asynchronous RPC (useful when
RPC does not have a return value e.g. void functions)
Asynchronous RPC (2)
(deferred synchronous)
2-13
A client and server interacting through two asynchronous RPCs
 Another form: client does not wait for acceptance (one-way RPC)
Writing a Client and a Server in DCE
2-14
The steps in writing a client and a server in DCE RPC.
Binding a Client to a Server
(Server, Machine Address)
(/video/movie, 123.45.67.01)
…
2-15
123.45.67.01
1024
Daemon e.g.@ port 110
Client-to-server binding in DCE.
(Server, Port)
(/video/movie, 1024)
…
RPC semantics:
At-most-once operation: repetitions are impermissible (e.g. after a crash)
Idempotent operation: repetition allowed (e.g. read)
Distributed Objects
2-16
Common organization of a remote object with client-side proxy.
 Object: compile-time (i.e. language-based) or run-time (e.g. built using an adaptor/wrapper) object
 Distributed object: state may be also distributed
 Remote object: only interfaces are replicated
Binding a Client to an Object
Distr_object* obj_ref;
obj_ref = …;
obj_ref do_something();
//Declare a systemwide object reference
//Initialize the reference to a distributed object
// Implicitly bind and invoke a method
(a)
Distr_object obj_ref;
Local_object* obj_ptr;
obj_ref = …;
obj_ptr = bind(obj_ref);
obj_ptr  do_something();
//Declare a systemwide object reference
//Declare a pointer to local objects
//Initialize the reference to a distributed object
//Explicitly bind and obtain a pointer to the local proxy
//Invoke a method on the local proxy
(b)
a)
b)
(a) Example with implicit binding using only global references
(b) Example with explicit binding using global and local references
Binding: a proxy is placed within the address space of the client
Object References: global and unique e.g. based on:
(1) machine address + port + object ID:
(-) port may change after crashes
(2) machine address + server ID + object ID: (+) survives crashes (-) fixed location
(3) machine address of location server + server ID + object ID: (-) not scalable
Static versus Dynamic Remote Method
Invocation
a)
Remote Method Invocation (RMI):
Similar to RPC in marshalling and parameter passing
Global object references (unlike RPC)
Potentially, a developer can design object-specific proxies (unlike RPC)
b)
Static RMI:
Using IDL or language-based (JAVA can generate stubs automatically)
Interfaces are known prior to development (hardwired in client)
Interface changes mean that client has to be recompiled
c)
Dynamic RMI:
Using IDL
Interfaces may be unknown at development time
Dynamic invocation:
invoke(object, methodName, inputParameters, outputParameters)
Interface changes do not mean a client recompilation (rather run-time errors)
Parameter Passing
2-18
The situation when passing an object by reference or by value.
L1: passed by value (local reference)
R1: passed by reference (remote reference)
The DCE Distributed-Object Model
(Distributed objects are an extension in DCE)
a)
b)
Distributed dynamic objects in DCE (client creates them - via RPC - at
run-time)
Distributed named objects (server publishes them – in a directory - for use)
Java RMI
a)
b)
c)
d)
Only remote objects (state not distributed)
Fully integrated in the Java language (stub generation)
Differences between local and remote objects in Java:
Cloning:
Remote objects cannot be cloned by clients. Only the server possessing
the object can clone it, proxies are not cloned.
Synchronization:
Only proxies can be blocked on. The remote object (in the server)
cannot be protected against simultaneous access by the mere use of
synchronized methods.
Parameter passing:
local objects: by value
remote objects: by reference
Object serialization (marshalling)
Most objects can be serialized except platform-dependent ones like file
descriptors, sockets, and so on.
Proxies are also serializable!  can be used as object references and
passed to other clients.
Message-oriented Communication
Persistence and Synchronicity in Communication (1)
2-20
General organization of a communication system in which hosts are
connected through a network
Persistence and Synchronicity in Communication (2)
Persistent communication of letters back in the days of the Pony Express.
Persistent communication: Message is stored in communication system until submission to
receiver (e.g. Email system). Sender/receiver need not be active during message delivery.
Transient communication: Message is discarded if receiver is inactive (non-persistent)
Synchronous communication: Sender blocks until message reaches receiver.
Asynchronous communication: Sender continues (immediately) after submitting its message.
Persistence and Synchronicity in Communication (3)
2-22.1
a)
b)
Persistent asynchronous communication (e.g. email)
Persistent synchronous communication
Persistence and Synchronicity in Communication (4)
2-22.2
c)
d)
Transient asynchronous communication (e.g. UDP datagrams)
Receipt-based transient synchronous communication
Persistence and Synchronicity in Communication (5)
e)
f)
Delivery-based transient synchronous communication at message delivery.
Sender waits until message is delivered to receiver. (e.g. asynchronous RPC)
Response-based transient synchronous communication. Sender waits until
the receiver already processed message and delivered the reply (e.g. C/S
interaction, RPC, RMI)
Message-oriented Transient Communication
Berkeley Sockets (1)
Primitive
Meaning
Socket
Create a new communication endpoint
Bind
Attach a local address to a socket
Listen
Announce willingness to accept connections
Accept
Block caller until a connection request arrives
Connect
Actively attempt to establish a connection
Send
Send some data over the connection
Receive
Receive some data over the connection
Close
Release the connection
Socket primitives for TCP/IP.
Berkeley Sockets (2)
Connection-oriented communication pattern using sockets.
The Message-Passing Interface (MPI)
Primitive
Meaning
MPI_bsend
Append outgoing message to a local send buffer (asynchronous)
MPI_send
Send a message and wait until copied to local or remote buffer (delivery/receiptbased)
MPI_ssend
Send a message and wait until receipt starts (delivery-based)
MPI_sendrecv
Send a message and wait for reply (response-based)
MPI_isend
Pass reference to outgoing message, and continue
MPI_issend
Pass reference to outgoing message, and wait until receipt starts
MPI_recv
Receive a message; block if there are none
MPI_irecv
Check if there is an incoming message, but do not block
Some of the most intuitive message-passing primitives of MPI.
 MPI is designed for multicomputers (parallel systems)
 Supports group communication
Message-Queuing Systems (1)
(also called MOM: message-oriented middleware)
2-26
Four combinations for loosely-coupled communications using queues.
(persistent asynchronous communication)
Message-Queuing Model (2)
Primitive
Meaning
Put
Append a message to a specified queue
Get
Block until the specified queue is nonempty, and remove the first message
Poll
Check a specified queue for messages, and remove the first. Never block.
Notify
Install a handler to be called when a message is put into the specified
queue.
Basic interface to a queue in a message-queuing system.
Traditional example: email
However: MOM does more:
- is targeted to applications and users (not only users)
- is more generic (cf. filtering emails)
- promotes application goals such as fault-tolerance, load balancing, guaranteed
delivery, priorities etc.
- In fact email systems could make use of MOM but MOM supports database,
workflow mgmt, groupware, and batch processing applications as well
General Architecture of a Message-Queuing System (1)
e.g. 127.23.45.01
The relationship between queue-level addressing and network-level
addressing.
General Architecture of a Message-Queuing System (2)
2-29
The general organization of a message-queuing system with routers.
(routers know the network topology)
Message Brokers
2-30
The general organization of a message broker in a message-queuing
system.
Broker: conversion between different message formats (interoperability)
Example: IBM MQSeries
2-31
General organization of IBM's MQSeries message-queuing system.
(used with financial mainframes e.g. banks)
MCA: message channel agent
Channel: a (TCP) connection between 2 MCAs
Channels in IBM MQSeries
Attribute
Description
Transport type
Determines the transport protocol to be used
FIFO delivery
Indicates that messages are to be delivered in the order they are sent
Message length
Maximum length of a single message
Setup retry
count
Specifies maximum number of retries to start up the remote MCA
Delivery retries
Maximum times MCA will try to put received message into queue
Some attributes associated with message channel agents.
Message Transfer (1)
IBM MQSeries
The general organization of an MQSeries queuing network using routing tables and
aliases.
(QM: Queue Manager, SQ: Send Queue, LA: Local Alias)
Message Transfer (2)
IBM MQSeries
Primitive
Description
MQopen
Open a (possibly remote) queue
MQclose
Close a queue
MQput
Put a message into an opened queue
MQget
Get a message from a (local) queue
Basic Primitives available in an IBM MQSeries MQI
Data Stream (1)
Setting up a stream between two processes across a network.
Data Stream (2)
2-35.2
Setting up a stream directly between two devices.
Data Stream (3)
An example of multicasting a stream to several receivers.
Transmission:
– Asynchronous: data units are temporally independent (e.g. text)
– Synchronous: there is a max. end-to-end delay for each data unit
– Isochronous: min. and max. delays = jitter (e.g. continuous media)
Specifying QoS (1)
Characteristics of the Input
•maximum data unit size (bytes)
•Token bucket rate (bytes/sec)
•Toke bucket size (bytes)
•Maximum transmission rate
(bytes/sec)
Service Required
•Loss sensitivity (bytes)
•Loss interval (sec)
•Burst loss sensitivity (data units)
•Minimum delay noticed (sec)
•Maximum delay variation (sec)
•Quality of guarantee
A flow specification.
Specifying QoS (2)
The principle of a token bucket algorithm.
Setting Up a Stream
The basic organization of RSVP for resource reservation in a distributed system.
Data link layer: (frame) priorities
ATM: QoS parameters (min./max. rate, max. jitter, ..)
Synchronization Mechanisms (1)
The principle of explicit synchronization on the level data units.
Synchronization Mechanisms (2)
2-41
The principle of synchronization as supported by high-level interfaces.