Multithreading

1
Multithreading Networking

• PJ Dillon
• CS401

2
Introduction

• Recall one of the first concepts we discussed
• A program or method is
• a sequence of instructions
• One executes after another
• Execute on the processor
• Operates on a set of variables
• Operating Systems can run multiple programs at a
  time
• Multitasking
• Boils down to multiple sequences of instructions
  that need executed
• Processor switches between each

3
What is a Thread?

• We term a sequence of instructions a Thread
• Every program has at least one thread
• Starts executing main()
• The Thread class provides means to create other
  threads
• When started, executes run() method
• run() method defined in Thread is empty
• Two ways of structuring a thread
• Create a subclass of Thread
• Implement Runnable interface
• Override run() method in each case
• Executing a thread
• In another thread (possibly main())
• First create an instance of a thread
• Pass it a runnable object
• or subclass
• call start() method
• Actually creates separate execution space
• Starts executing appropriate run()

4
Thread states

• Just as programs start, execute, and terminate,
  threads do the same
• Occurs within the view of other threads
• Thread.getState() method returns state
• New
• Initial state of a newly created thread
• Has not been started yet
• No associated execution space
• Runnable
• Actively executing instructions
• Contending with other threads for processor
• Blocked, Waiting, Timed-Waiting
• Waiting for something an event to complete
• IO operation needs to complete
• Sleep interval to expire
• Called wait()
• Terminated
• Executed run() to completion

5
Thread Scheduling

• Most PCs have only a single processor
• Can only execute one thread at a time
• Scheduling is the process of deciding which
  thread gets the CPU
• A topic of most OS courses
• In Java, each thread is given a priority
• Integer between MIN_PRIORITY and MAX_PRIORITY
  (1-10)
• setPriority(), getPriority() methods
• Higher integers have higher priority
• Of all the threads in the Runnable state
• Thread with the highest priority gets the
  processor
• executes until completion
• All other threads starve until its done
• If multiple threads of equal priority exist, one
  is arbitrarily chosen

6
Sharing Memory

• See ex48.java
• Whats the minimum value sum could take?
• The statement, sum, is compiled into multiple
  instructions
• Value copied to processor register
• Addition is performed
• Value is copied back to memory
• Several threads may copy the same value to a
  register
• Copy stale value register
• Write over newer value

7
Producer/Consumer Relationship

• Large programs often have threads associated with
  different tasks
• One task produces data needs by the other
• Could be producing data faster than it can be
  processed
• Ex A thread is often associated with reading
  data and interpreting from network sockets
• Could read a request
• Passes the request to some manager responsible
  for handling it
• Returns to listening for more socket data
• ex49.java

8
Synchronizing Tasks

• We need some tools to be able to do this
• synchronized keyword
• Added to method declarations
• Stand-alone synchronized blocks
• States that only one thread can execute the
  associated method or code at a time
• Other threads wait until executing one is done
• Underlying lock associated with the object
• wait() method
• Blocks the calling thread
• Like an indefinite sleep()
• waits for corresponding notify() or notifyAll()
  call
• Must be called in a synchronized block
• notify()/notifyAll()
• Wakes up one or all other threads, respectively,
  that previously called wait on the object
• Lets do ex49b.java

9
Locks and Conditions

• Synchronized, wait(), and notify() have
  limitations
• Can be used to implement more complex
  synchronization
• Java 5.0 adds more advanced API
• Lock
• Synchronization object
• A thread must acquire it before executing code
• Calls lock() method
• Only one thread can acquire it at a time
• All others wait until owner calls
• Condition
• Manages threads that are forced to wait for some
  condition to become true
• Consumer has no data available to consume
• Producer has no space to put new data
• Calling thread blocks itself calls await()
• Signaled by another thread to continue
• Condition must know what lock is protecting it

10
Concurrency Structures

• The java.util.concurrent. package contains
  predefined classes for
• More complex synchronization
• Collections supporting simultaneous access/update
• Managing threads
• See The Java Tutorial and API for more info
• Take CS1550 to learn more about this

11
Networking
12
Basics

• The Internet
• Data is transferred in groups of bytes
• Called a packet
• Each connected machine is identified by a unique
  IP address
• 32-bit number
• Packets are tagged with IP address of the
  destination machine (and source)
• Transferred to machine through various networks
• Ultimately arrive at the NIC card of the
  destination
• Packet data is passed up from the NIC and OS to a
  waiting application
• Each machine can be running multiple applications
  that expect packets
• Each application associates itself with a port
• A unique 16-bit number
• Different from any other application on the
  machine
• Packets are tagged with proper port number

13
Communication Protocols

• The Internet is not perfect
• Packets may get dropped, never delivered
• may arrive out of order
• May get duplicates
• Two common protocols to deal with this
• TCP Transmission Control Protocol
• Connection oriented
• Orders data prior to passing it up to application
• Automatically retransmits lost packets
• Ignores duplicates
• Application sees data in the same order it was
  sent
• UDP User Datagram Protocol
• Connectionless protocol
• Provides no guarantees with transmission
• Packets are independent of each other
• Well focus on TCP

14
Client/Server Model

• Connections are formed between two nodes
• One is designated the server
• Other the client
• Establishing a Connection
• Server application starts
• Associates itself with a port on the NIC
• Sits, idle, waiting for clients to connect
• Client application starts
• Creates TCP communication channel
• Sends Initial connection request packet to server
  IP and Port
• Server accepts connection
• Opens new communication channel
• One end of a communication channel is called a
  Socket
• Composed of an InputStream and OutputStream

15
Client/Server in Java

• The java.net. package defines the ServerSocket
  and Socket classes for establishing connections
• Server
• Create ServerSocket object
• Can be bound to a port at creation time
• ServerSocket s new ServerSocket(3459)
• Call accept() method
• Waits for and accepts a new connection from a
  client
• Returns a Socket object represented the
  connection
• Read/Write from associated streams
• Client
• Create Socket
• Can connect to server at creation time or
• Call connect() method later
• Read/Write from associated streams
• MyServer.java, MyClient.java

16
Handling Multiple Connections

• Lets see MyServer2.java
• loops forever to continuously accept new
  connections
• Theres still a problem with this code
• Server only accepts one connection at a time
• Only call accept() again after client closes
  connection
• Code needs to get back to accept() as soon as
  possible
• Solution is multithreading
• One thread accepts all connections
• Spawns others to process the information on the
  socket
• MyServer3.java, MyClients.java

17
Application Protocol

• To do anything useful, the server and client have
  to agree on a protocol
• A set of common messages is defined
• Text or byte messages
• Fields laid out in a defined order
• Marker between messages
• All messages are the same byte size
• Special byte or character marks end of a message
• When one message is sent, another can be expected
  in return
• E.g. clients sends username in password to server
  
• Server responds with login success or failure
• MyServer4.java, MyClients2.java

18
Going Beyond

• Large servers with thousands of clients cant
  create a separate thread for each one
• Most of the time, the thread is blocked waiting
  for input from the user anyway
• Each client thread performs the same task
• An alternative lets a thread monitor a number of
  sockets
• Wait for data on any one of them
• Returns a group of sockets that have data
  available
• Code can respond as needed
• See java.nio. and java.nio.channels. API for
  more