Concurrent Processing in ClientServer Software

1
Lecture 3

• Concurrent Processing in Client/Server Software

2
Overview

• Introduction
• Concurrency in Networks
• Concurrency in Servers
• Terminology and Concepts
• Process Concept
• Processes vs. Programs
• Procedure Calls
• Concurrent Process Creation
• Time-slicing
• Process Divergence
• Context Switching
• Concurrency and Asynchronous I/O
• Summary

3
Introduction

• Client/Server paradigm as it relates to
  concurrency
• Provides much of the power between client/server
  interactions
• Makes the software difficult to design and build
• Facilities provided by the operating system to
  support concurrent process execution

4
Concurrency in Networks

• The term concurrency refers to real or apparent
  simultaneous computing
• Time-sharing
• Multiprocessing
• Concurrent processing is fundamental to
  distributed computing
• Concurrency among multiple process on separate
  machines
• Concurrency among a single computer system
• Concurrency among a set of clients on a single
  machine

5
Client Concurrency

• Client software does not require any special
  effort to make it concurrent
• The client program is designed without regard to
  concurrency
• Concurrency among client programs occurs
  automatically because the operating system
  supports simultaneous program execution

6
Concurrency Among Client Programs
C1
C4
Internet
C2
C3
7
Concurrency in Servers

• Unlike concurrent client software, concurrent
  server software requires significant effort
• Why is server concurrency important
• Concurrent client access
• Throughput
• Response Time

8
Concurrent Server
C2
server
C1
Internet
C3
C4
9
Process Concept

• Process abstraction defines the fundamental unit
  of computation
• Address space and at least one thread of
  execution
• Essential process information
• Instruction pointer
• Process Id
• Executing program
• Program text and data areas (memory locations)

10
Local vs. Global Variables

• A single threaded program doesnt need to worry
  about concurrent access to variables.
• Multiple threaded programs need their own copy of
  the variables

11
Procedure Calls
Thread 2

• Concurrent programming systems provide separation
  between procedure calls in executing threads
• When multiple threads execute a piece of code
  concurrently, each has its own run-time stack

Thread 1
Procedure A
Procedure B
Procedure C
12
Concurrent Process Creation
A Sequential C Example include
ltstdio.hgt include ltstdlib.hgt int sum int main()
int I sum 0 for (i 1 i lt 5
i) printf(The value of i is d \n,
i) fflush(stdout) sum i
printf(The sum is d \n, sum) exit(0)
13
Concurrent Process Creation
A Concurrent Version include ltstdio.hgt int
sum int main() int i sum 0
fork( ) / Create a new process / for (i
1 i lt 5 i) printf(The value of i
is d \n, i) fflush(stdout) sum i
printf(The sum is d \n, sum)
exit(0)
14
Timeslicing

• Timeslicing attempts to allocate the CPU equally
  among all threads
• If there are N threads executing, each thread
  receives 1/N of the CPU
• All threads appear to proceed at an equal rate
• Many threads executing the rate is low with few
  the rate is high

15
Context Switching Protocol Software Design

• Concurrent processing makes programs more
  powerful but it adds computational overhead
• A context switch occurs when the operating system
  switches from one thread to another
• Context switching is pure overhead needed to
  support concurrent processing
• Protocol software should be designed to minimize
  context switching
• Benefits of concurrency needs to outweigh the
  costs of switching contexts

16
Concurrency Asynchronous I/O

• Some operating systems allow an application to
  control concurrent input and output operations
• The select system call provides a fundamental
  operation to manage concurrent I/O

17
Summary

• Concurrency is fundamental to TCP/IP applications
• Concurrency in client applications results in
  multiple clients executing at the same time
• Concurrency in server applications is more
  difficult because concurrency has to be handled
  explicitly.
• In Unix/Linux systems additional processes are
  created by using the fork system call
• Concurrency generates system resource overhead
• The select system call allows a process to manage
  concurrent I/O