REQUEST/REPLY COMMUNICATION

1
REQUEST/REPLY COMMUNICATION

• BHAVIN KANSARA

2
OBJECTIVE

• Request/Reply Communication
• RPC ( Remote Procedure Call )
• RPC Operation
• Issues Associated with RPC
• parameter passing
• binding
• compilation
• failure handling
• security
• Suns secure RPC
• GridRPC

3
REQUEST/REPLY COMMUNICATION
Most widely used request/reply communication
model is Remote Procedure Call (RPC)
4
WHAT IS RPC

• Process on Procedure P
• machine M1 on machine M2

Calls
M1 is suspended M2 executes P When M2 returns,
controls back to M1

• Neither message passing nor I/O at all are
  visible to the programmer

This method is known as Remote Procedure Call
(RPC)
5
RPC STUB COMPONENTS

• -- Program subsystems used by client or server
  to initiate an RPC request
• -- The stubs are responsible for managing all
  details of the remote communication between
  client and server
• Client calls client stub residing on local
  machine ,responsible for packing the call and its
  parameters into a message format
• On server side, a server stub unpacks the message
  into call request

6
RPC OPERATION
7
ISSUES ASSOCIATED WITH RPC

• Parameter passing
• Binding
• Compilation
• Failure handling
• Security

8
PARAMETER PASSING

• Functions in an application that runs in a
  single process may collaborate via parameters
  and/or global variables
• Passing parameters is typically the only way
  that RPC-based clients and servers share
  information
• Parameter-passing has following possible modes
• IN pass info from caller to callee
• OUT callee writes a value in the caller
• IN/OUT caller tells callee value of var,
  which may be updated by callee
• Parameter passing methods
• Call by value
• Call by reference
• Call by copy/restore

9
PARAMETER PASSING

• Call by Value
• IN Mode
• simple to handle
• Value passed to procedure is copied to local
  variable
• cant have an effect on caller variable
• Call by reference
• IN/OUT mode
• Much harder to handle
• Pass pointer address
• Same memory location is referred
• Call by copy/restore
• IN/OUT mode
• Call by value call by reference
• Call by value at entry of procedure and
  restricts the call by reference to
  the exit of the call
• Results are copied back to the calling procedure
  at the completion of the called procedure

10
BINDING
Manages mapping between program number and port
number
When server machine is unknown, the client will
need to locate the server machine by contacting
directory server to locate address of server
system
11
BINDING
Server starts ? registers its communication
endpoint by sending request to port mapper ?
that contains servers program, version number
and port number
12
BINDING
Client process contact port mapper to obtain
handler for accessing server with specific
program and version number. This is done by RPC
run-time library routine create.
13
BINDING
The port mapper returns the port number of the
server to the client system after verifying the
program and version number.
14
BINDING
Client system builds client handle for the client
process for subsequent use in the remote
procedure call. Binding process establish socket
connection between client and server
15
RPC COMPILATION

• The compilation of RPC requires three major
  components in the RPC package
• An interface specification file
• RPC generator which takes the interface
  specification file as input and produces the
  client and server stub procedure source code as
  output
• A run-time library for the supporting execution
  of RPC

16
RPC COMPILATION
17
RPC COMPILATION
18
RPC COMPILATION
19
RPC FAILURE HANDLING

• Failures can be of three types
• Lost message
• Server crash
• Client crash

20
RPC FAILURE HANDLING
Lost Message
Sever is allocated and request message has been
sent
Message may delay or lost
Lost message are eventually detected by time out
or no response form server
Retransmission
Request was not lost but delayed ? server will
receive two request
SOLUTION --- Request can be executed any number
of times with same effect --- using
sequence numbers
21
RPC FAILURE HANDLING
Server Crash

• Three possible assumptions for the RPC semantics
  in the presence of failure
• " Exactly once?
• Server crashes before request arrives
• server crashes before reply
• server crashes after reply, but reply dropped
• " At most once?
• If server crashes, cant know if request was
  done
• " At least once?
• Keep retrying across crashes, but may be done
  multiple times

Most desirable ? Exactly once ? difficult to
achieve SOLUCTION --- Problem lies in the loss
of cache table --- Use at least once semantics
and to log the cache table --- When the server
recovers ? it reloads cache table from log
22
RPC FAILURE HANDLING
Client Crash
Before server completes the client request, the
server has an orphan computation and its reply is
undeliverable. No easy way for the server to
detect the disappearance of clients except by
using a time-out or by waiting for the failed
client to reboot Orphan computations consume
server resources Can be eliminated by By
Client upon reboot of failed client By Server
the server occasionally tries to locate the
owners of its remote operations and abort those
whose owners can not be found By Expiration
each remote operation is given a maximum lifetime
23
SECURITY

• RPC is a form of remote execution that allows
  programs or commands to be executed on other
  systems. But it also opens doors for attacks from
  unfriendly remote users
• Mutual authentication
• --- identities of clients and server are
  verified
  --- Request Message
  Client
  server --- Reply message
  server client
  
• Message integrity, confidentiality, and
  originality
• --- request/reply messages have not been
  tampered with (integrity), their contents are not
  revealed ( confidentiality) , and the same
  message has not appeared more than once (
  originality )

Truly generated by
Intended for
24
SUNS SECURE RPC
25
GRIDRPC

• One of grid programming models
• Supports remote procedure calls (RPCs) on the
  Grid
• Standardized at GridRPC
• Standardizing only minimal set of APIs
• Higher-level features can be built on top
• Providing several reference implementations
• Usage scenarios
• Remote library calls for executing
  compute-intensive tasks
• Executing large scale task-parallel processing

26
NINF-G SYSTEM (2000..)

• Reference implementation of GridRPC
• Constructed on top of Globus
• Globus is introduced in many sites
• No more security halls are required in
  introducing Ninf-G2
• Ninfied applications can be executed on many
  sites
• Globus provides primitive functions/tools for
  grid computing
• Concentrating on the development of RPC specific
  functions
• Easy to use other tools on Globus
• Ninf-G2 hides complicated mechanisms of Globus
• Grid applications/middleware can be easily
  constructed

27
GRID APPLICATIONS USING NINF-G

• Chemical simulation
• Replica Exchange Monte Carlo
• Surveying potential energy of
  molecules
• Time Dependent Density Functional Theory
• Simulating time evolution of
  electron system in molecules
• Optimization problem
• Vehicle Routing
• Searching optimal route for delivery
• Weather simulation
• S-model
• Forecasting global weather change

28
REFERENCES

• Distributed operating system and algorithms
  Randy chow,
  Theodore johnson
• Remote procedure call - Wikipedia, the free
  encyclopedia en.wikipedia.org/wik
  i/Remote_procedure_call
• Sun RPC
  
  www.cdk4.net/additional/rmi/Ed2/SunRPC.pdf
• GridRPC A Remote Procedure Call API for Grid
  Computing www.cs.umd.edu/class/spring2
  004/cmsc818s/Readings/GridRPC-GGF-APM-jul02.pdf
• Grid technology research center
  http//projects.gtrc.aist.go.jp/en/
  http//www.ngp.org.sg/events/AIST
  -NGO_MOU/NINF.pdf