Communication in Distributed Systems
REKs adaptation of Tanenbaums Distributed Systems Chapter 2
1
Communication Paradigms
Using the Network Protocol Stack Remote Procedure Call - RPC Remote Object Invocation - Java Remote Method Invocation Message Queuing Services - Sockets Stream-oriented Services
Distributed Computing Systems
Network Stack - OSI Reference Model
Application A
Application Layer Presentation Layer Session Layer Transport Layer Network Layer Communication Network Network Layer Data Link Layer Physical Layer Network Layer Data Link Layer Physical Layer
Application B
Application Layer Presentation Layer Session Layer Transport Layer Network Layer
Data Link Layer
Physical Layer
Copyright 2000 The McGraw Hill Companies
Data Link Layer
Physical Layer
Electrical and/or Optical Signals
Leon-Garcia & Widjaja: Communication Networks Distributed Computing Systems Figure 2.6
Copyright 2000 The McGraw Hill Companies
OSI Reference Model
Leon-Garcia & Widjaja: Communication Networks
Application A
Application Layer Presentation Layer Session Layer Transport Layer Network Layer
data
Application B ah ph sh th nh dh
Application Layer Presentation Layer Session Layer Transport Layer Network Layer
data
data
data data data dt
Data Link Layer
Physical Layer
data
bits
Data Link Layer
Physical Layer
Distributed Computing Systems
TCP Connection Overhead
Normal operation of TCP
Transactional TCP
5
Distributed Computing Systems
Remote Procedure Calls
RPC
Conventional Procedure Call
Count = read(fd, buf,bytes)
Note call-by-value and call-by-reference parameters on the stack.
a)
b)
Parameter passing in a local procedure call: the stack before the call to read. The stack while the called procedure is active.
Distributed Computing Systems
Remote Procedure Call
RPC concept :: to make a remote procedure call appear like a local procedure call. The goal is to hide the details of the network communication (namely, the sending and receiving of messages). The calling procedure should not be aware that the called procedure is executing on a different machine.
Distributed Computing Systems
Remote Procedure Call
When making a RPC:
The calling environment is suspended. Procedure parameters are transferred across the network to the environment where the procedure is to execute. The procedure is executed there. When the procedure finishes, the results are transferred back to the calling environment. Execution resumes as if returning from a regular procedure call.
Distributed Computing Systems
RPC differs from OSI
User does not open connection, read, write, then close connection client may not even know they are using the network. RPC may omit protocol layers for efficiency. (e.g. diskless Sun workstations will use RPC for every file access.) RPC is well-suited for client-server interaction where the flow of control alternates.
Distributed Computing Systems
10
RPC between Client and Server
Distributed Computing Systems
11
RPC Steps
1.
2.
3.
4.
5.
The client procedure calls a client stub passing parameters in the normal way. The client stub marshals the parameters, builds the message, and calls the local OS. The client's OS sends the message (using the transport layer) to the remote OS. The server remote OS gives transport layer message to a server stub. The server stub demarshals the parameters and calls the desired server routine.
Distributed Computing Systems
12
RPC Steps
6. The server routine does work and returns result to the server stub via normal procedures. 7. The server stub marshals the return values into the message and calls local OS. 8. The server OS (using the transport layer) sends the message to the client's OS. 9. The client's OS gives the message to the client stub 10. The client stub demarshals the result, and execution returns to the client.
Distributed Computing Systems
13
RPC Steps
Distributed Computing Systems
14
Passing Value Parameters
a) b) c)
Original message on the Pentium The message after receipt on the SPARC The message after being inverted. The little numbers in boxes indicate the address of each byte
Distributed Computing Systems
15
Marshaling Parameters
Parameters must be marshaled into a standard representation. Parameters consist of simple types (e.g. integers) and compound types (e.g., C structures). The type of each parameter must be known to the modules doing the conversion into standard representation.
Distributed Computing Systems
16
Marshaling Parameters
Call-by-reference is not possible in parameter passing. It can be simulated by copy-restore. A copy of the referenced data structure is sent to the server, and upon return to the client stub the clients copy of the structure is replaced with the structure modified by the server.
Distributed Computing Systems
17
Marshaling Parameters
However, in general marshaling cannot handle the case of a pointer to an arbitrary data structure such as a complex graph.
Distributed Computing Systems
18
Parameter Specification and Stub Generation
The caller and the callee must agree on the format of the message they exchange, and they must follow the same steps when it comes to passing complex data structures.
A procedure
The corresponding message
Distributed Computing Systems
19
RPC Details
An Interface Definition Language (IDL) is used to specific the interface that can be called by a client and implemented by the server. All RPC-based middleware systems offer an IDL to support application development.
20
Distributed Computing Systems
Doors
The principle of using doors as IPC mechanism.
Distributed Computing Systems
21
Doors
A Door is a generic name for a procedure in the address space of a server process that can be called by processes colocated with the server. Doors require local OS support.
Distributed Computing Systems
22
Asynchronous RPC
2-12
a)
b)
The interconnection between client and server in a traditional RPC The interaction using asynchronous RPC
Distributed Computing Systems
23
Asynchronous RPCs
A client and server interacting through two asynchronous RPCs
Distributed Computing Systems
24
