Transcript Distributed Systems Principles and Paradigms

Foundations: Revision
Dr. Christian Vecchiola
Postdoctoral Research Fellow
[email protected]
Cloud Computing and Distributed Systems (CLOUDS) Lab
Dept. of Computer Science and Software Engineering
The University of Melbourne
Foundations: Revision
Distributed Systems Principles and Paradigms
Outline
 Introduction
 Socket Programming
– Cross Platform/Language Communication
– .NET Socket Programming
 Thread Programming
– More on Synchronization
– Java wait() & notify()
– .NET Thread Programming
 Assignment 1
– Multithreaded Dictionary Server
Foundations: Revision
Distributed Systems Principles and Paradigms
Introduction
Distributed Systems
– Definitions
• “A system in which hardware or software components located at
networked computers communicate and coordinate their actions only by
message passing.” (Colouris, Dollimore, Kindberg)
• “A distributed system is a collection of independent computers
that appear to the users of the system as a single computer.”
(Tanenbaum and Van Steen)
– Aspect we focused
• Communication
– Technology: Sockets
• Concurrency
– Technology: Threads
Foundations
Foundations: Revision
Distributed Systems Principles and Paradigms
Introduction
Sockets
– What did we talk about
• Socket Abstraction
• TCP/IP and UDP/IP stacks
• Connection-oriented vs Connectionless
Communication
• Java & .NET Sockets
– What is missing?
• Advanced applications with Sockets
• Cross-Platform / Language Socket Programming
Today’s Focus
Foundations: Revision
Distributed Systems Principles and Paradigms
Introduction
Threads
– What did we talk about
•
•
•
•
•
Thread Abstraction
Thread/Process Relation
Multithreading (server applications)
Basic Synchronization
Java Threads & Synchronization APIs
– What is missing?
• Advanced Synchronization Problems
• .NET & Synchronization APIs
Today’s Focus
Foundations: Revision
Distributed Systems Principles and Paradigms
Socket Programming
Foundations: Revision
Distributed Systems Principles and Paradigms
Socket Programming
Socket Abstraction
– Client-server based approach
– Defined by a couple: <address, port>
– Connection-Oriented Sockets
• Based on TCP
• Reliable but more “weighty” to manage
• Stream based I/O
– Connectionless Sockets
• Based on UDP
• Unreliable but light
• Packet based I/O
Foundations: Revision
Distributed Systems Principles and Paradigms
Socket Programming
Java Sockets
– Package: java.net.*;
– Concepts:
•
•
•
•
Full implementation of the model
Stream based communication for TCP
Packet based communication for UDP
Exceptions for network errors
Foundations: Revision
Distributed Systems Principles and Paradigms
Socket Programming
Java Sockets
– Package: java.net.*;
– Classes:
• Connection Oriented Communication (TCP)
– java.net.ServerSocket (server component, always on)
– java.net.Socket (client component, connection bound lifetime)
– Stream-oriented data transfer
» java.io.DataInputStream
» java.io.DataOutputStream
– Examples: SimpleServer.java & SimpleClient.java
• Connectionless Communication (UDP)
– java.net.DatagramSocket (client and server, send & receive)
– java.net.DatagramPacket (packet abstraction, container of bytes)
– Examples: UDPServer.java & UDPClient.java
Foundations: Revision
Distributed Systems Principles and Paradigms
Socket Programming
.NET Sockets
– Namespaces:
• System.Net;
• System.Net.Sockets;
– Concepts:
•
•
•
•
•
Use of xxxClient classes to simplify communication
Where is the Socket class ?
Stream based communication for TCP
Packet based communication for UDP
Exceptions for network errors
Foundations: Revision
Distributed Systems Principles and Paradigms
Socket Programming
.NET Sockets
– Classes:
• Connection Oriented Communication (TCP)
– System.Net.Sockets.TcpListener
» server component
» always on
– System.Net.Sockets.TcpClient
» client component,
» connection bound lifetime
– Stream-oriented data transfer
» System.Net.Sockets.NetworkStream
» System.IO.StreamReader, StreamWriter
– Examples: SimpleServer.cs & SimpleClient.cs
• Connectionless Communication (UDP)
– System.Net.Sockets.UdpClient
Foundations: Revision
Distributed Systems Principles and Paradigms
Socket Programming
 Observations
– Simple abstraction for network communication
– General enough to serve different purposes
– More importantly:
Sockets are
Platform / Language Independent!
– But…
• What about the examples?
• It did work, but not as expected…
• It is a problem about Java vs NET ?
Foundations: Revision
Distributed Systems Principles and Paradigms
Socket Programming
Observations
– Wait a minute…
• Sockets provide an independent abstraction……?
• Yes…, but for transferring bytes!
• As long as we limit ourselves to pure byte transfer the
communication is platform/langue independent!
– So..?
• The examples where based on UTF strings!
• String management made the difference!
• Problem with string termination.
Foundations: Revision
Distributed Systems Principles and Paradigms
Socket Programming
Solutions
– We use a byte-oriented processing!
– We do not rely on readline methods!
– We enforce define our protocol for string
termination.
Java
.NET / Mono
Foundations: Revision
Distributed Systems Principles and Paradigms
Thread Programming
Foundations: Revision
Distributed Systems Principles and Paradigms
Thread Programming
Thread Abstraction
–
–
–
–
Statically ordered sequence of instructions.
Piece of code that run concurrently with other threads.
Each process has at least one thread.
Threads
• allow to perform multiple tasks at once within the same process
• give the illusion of concurrency
– Applications
•
•
•
•
Asynchronous IO
UI rendering
Background task
Increase throughput
Foundations: Revision
Distributed Systems Principles and Paradigms
Thread Programming
Concurrency management
– Concurrent execution leads to:
•
•
•
•
Races on share data
Spurious writes and reads
Inconsistency of state
Undefined order of execution
– How can we solve this?
Foundations: Revision
Distributed Systems Principles and Paradigms
Thread Programming
Concurrency management
– Synchronization is the solution!
• There exist some APIs that allow …
– … exclusive access to shared data structures
– … protecting the state of data
– … atomic execution of a sequence of instruction
• These APIs …
– … are fundamental element of the language
– … provide a solution to the problem discussed
Foundations: Revision
Distributed Systems Principles and Paradigms
Thread Programming
Concurrency management
– Concurrency is a widely studied phenomenon
– There exist models and abstractions to avoid
concurrency issues
• Semaphores
• Monitors
• Barriers
– APIs and models changes according to the
programming language
– Mostly, the concept of Monitor is implemented
Foundations: Revision
Distributed Systems Principles and Paradigms
Thread Programming
Concurrency management
– What is a Monitor?
• It represents a guarded region where the execution of
statements is exclusive
• Classic operations are:
– <enter> : acquires the exclusive access
– <exit> : release the exclusive access
• How to do it in Java?
– synchronized(object){ } block;
– synchronized modifier keyword;
Foundations: Revision
Distributed Systems Principles and Paradigms
Thread Programming
Java Synchronization APIs
– synchronized(object obj){ }
• Defines a simple guarded region in which only one
thread is allowed to access
• All the threads that have a synchronized(obj){ … }
where obj is a reference to the same object, mutually
exclude themselves
– synchronized keyword
• Provides a guarded region that covers the entire
method to which the keyword is applied
• The synchronization instance is the instance on which
the method is called
Foundations: Revision
Distributed Systems Principles and Paradigms
Thread Programming
Java Synchronization APIs
– Observations
• synchronized provides a way to:
– acquire exclusive access to a resource
– release this access once done
• what if..
– the acquisition of a resource need is determined by a
condition?
– the condition needs to checked within a guarded region?
– the threads need to be queued, to access the resource?
– the condition is the result of multiple threads cooperating?
Foundations: Revision
Distributed Systems Principles and Paradigms
Thread Programming
Java Synchronization APIs
– Example
• Producer – Consumer
– One thread (or more) produces items
– One thread (or more) consumes these items
– The items are put in shared buffer among the threads
Producer
Consumer
Foundations: Revision
Distributed Systems Principles and Paradigms
Thread Programming
Java Synchronization APIs
– Producer – Consumer
• Conditions
– The producer thread can access the buffer only if there is
room left to store an item
– The consumer thread can access the buffer only if there is
some item to consume
• Problem
– Can we implement this model with synchronized?
– I mean….. optimally?
Foundations: Revision
Distributed Systems Principles and Paradigms
Thread Programming
Java Synchronization APIs
– Producer – Consumer
public class Buffer {
private Object[] items;
private int empty;
private int available;
private boolean bEmpty;
private boolean bFull;
public Buffer(int length) {
this.items = new Object[length];
this.empty = 0;
this.available = 0;
this.bFull = false;
this.bEmpty = true;
}
public void put(Object item) throws Exception {
if (this.bFull == true) {
throw new Exception(“The buffer is full!”);
}
this.items[this.empty] = item;
this.empty = ((this.empty + 1) % this.items.length);
this.bFull = this.empty = this.available;
this.bEmpty = false;
}
Foundations: Revision
Distributed Systems Principles and Paradigms
Thread Programming
Java Synchronization APIs
– Producer – Consumer
public Object get() throws Exception {
if (this.bEmpty == true) {
throw new Exception(“The buffer is empty!”);
}
Object item = this.items[this.available];
this.available = ((this.available + 1) % this.items.length);
this.bEmpty = this.available == this.empty;
this.bFull = false;
return item;
}
public boolean isFull() {
return this.bFull;
}
public boolean isEmpty() {
return this.bEmpty;
}
}
Foundations: Revision
Distributed Systems Principles and Paradigms
Thread Programming
Java Synchronization APIs
– Producer – Consumer
public class Producer implements Runnable {
private Buffer buffer;
public Producer(Buffer buffer) {
this.buffer = buffer;
}
public void run() {
try {
while(true) {
synchronized(this.buffer) {
if (this.buffer.isFull() == false) {
Object item = this.produce();
this.buffer.put(item);
} else {
System.out.println(“Producer: no room for cheese!”);
}
}
}
} catch(Exception ex) {
System.out.println(“Producer: exception: ” + ex.getMessage());
}
}
private Object produce() {
return new Object();
}
}
Foundations: Revision
Distributed Systems Principles and Paradigms
Thread Programming
Java Synchronization APIs
Producer – Consumer
public class Consumer implements Runnable {
private Buffer buffer;
public Consumer(Buffer buffer) {
this.buffer = buffer;
}
public void run() {
try {
while(true) {
synchronized(this.buffer) {
if (this.buffer.isEmpty() == false) {
Object item = this.buffer.get();
this.consume(item);
} else {
System.out.println(“Consumer: no cheese for me!”);
}
}
}
} catch(Exception ex) {
System.out.println(“Consumer: exception: ” + ex.getMessage());
}
}
private void consume(Object obj) {
// whatever we need to do with it
}
}
Foundations: Revision
Distributed Systems Principles and Paradigms
Thread Programming
Java Synchronization APIs
– Producer – Consumer
public class ProducerConsumer {
public static void main(String[] args) {
// create the shared buffer and the producer
// and consumer classes
Buffer buffer = new Buffer(10);
Producer producer = new Producer(buffer);
Consumer consumer = new Consumer(buffer);
// initialize the two threads...
Thread p = new Thread(producer);
Thread c = new Thread(consumer);
// NOTE: let’s play with thread priorities and see
//
what is happening...
p.start();
c.start();
}
}
Foundations: Revision
Distributed Systems Principles and Paradigms
Thread Programming
Java Synchronization APIs
– Producer – Consumer
• Observations
– We have different execution traces with different thread
priorities
– In every case, we notice the following output lines:
– Producer: no room for cheese!
– Consumer: no cheese for me!
– These lines identify conditions in which one of the two
processes has acquired the resource without actually
needing it
– This is the result of a bad design!
Foundations: Revision
Distributed Systems Principles and Paradigms
Thread Programming
Java Synchronization APIs
– Producer – Consumer
• How can we implement this better?
– We need to ensure that the access to the resource is only
obtained when appropriate
– This means…
– [Producer] when there is space in the buffer…
– [Consumer] when there is some item to consume…
• Can we implement such pattern in Java?
– Yes, definitely!
– By using: object.wait(), object.notify(),
object.notifyAll()
Foundations: Revision
Distributed Systems Principles and Paradigms
Thread Programming
Java Synchronization APIs
- Object.wait(), Object.notify(),
Object.notifyAll()
• These are primitives that are used to implement more
complex patterns
• Object.wait() makes one thread wait until a
specific call to Object.notify() or
Object.notifyAll() is made (on the same
instance).
• Object.notify() signals the first waiting thread to
stop the waiting process and proceeds the execution.
• Object.notifyAll() does the same but wakes up all
the threads waiting and a resource contention process
is activated.
Foundations: Revision
Distributed Systems Principles and Paradigms
Thread Programming
Java Synchronization APIs
– Producer – Consumer
• Can we use Object.wait() & Object.notify() ?
– Yes, definitely!
– Producer:
– Checks buffer.isFull() and waits if returned value is
true.
– Signals the buffer at the end of the insertion
– Consumer:
– Checks buffer.isEmpty() and waits if returned value is
true.
– Signal the buffer at the end of the extraction.
Foundations: Revision
Distributed Systems Principles and Paradigms
Thread Programming
Java Synchronization APIs
– Producer – Consumer
public class Producer implements Runnable {
private Buffer buffer;
public Producer(Buffer buffer) {
this.buffer = buffer;
}
public void run() {
try {
while(true) {
synchronized(this.buffer) {
if (this.buffer.isFull() == true) {
this.buffer.wait();
}
Object item = this.produce();
this.buffer.put(item);
this.buffer.notify();
}
}
} catch(Exception ex) {
System.out.println(“Producer: exception: ” + ex.getMessage());
}
}
private Object produce() {
return new Object();
}
}
Foundations: Revision
Distributed Systems Principles and Paradigms
Thread Programming
Java Synchronization APIs
– Producer – Consumer
public class Consumer implements Runnable {
private Buffer buffer;
public Consumer(Buffer buffer) {
this.buffer = buffer;
}
public void run() {
try {
while(true) {
synchronized(this.buffer) {
if (this.buffer.isEmpty() == true) {
this.buffer.wait();
}
Object item = this.buffer.get();
this.consume(item);
this.buffer.notify();
}
}
} catch(Exception ex) {
System.out.println(“Consumer: exception: ” + ex.getMessage());
}
}
private void consume(Object item) {
// whatever we need to do with it
}
}
Foundations: Revision
Distributed Systems Principles and Paradigms
Thread Programming
Java Synchronization APIs
– Producer – Consumer
• Observations
– This implementation is better…
– But still, we are using a single lock instance to control two
different conditions!
– Buffer empty
– Buffer full
– This does not completely avoids the fact that the resource is
acquired or tested without any need…
– There still room for optimization….
– …by using two different conditions.
Foundations: Revision
Distributed Systems Principles and Paradigms
Thread Programming
Java Synchronization APIs
– Producer – Consumer
public class Buffer {
private Object[] items;
private int empty, available;
private boolean bEmpty, bFull;
private Object emptyHandle, fullHandle;
public Buffer(int length) {
this.items = new Object[length];
this.empty = 0;
this.available = 0;
this.bFull = false;
this.bEmpty = true;
this.emptyHandle = new Object();
this.fullHandle = new Object();
}
public Object getEmptyHandle() {
return this.emptyHandle;
}
public Object getFullHandle() {
return this.fullHandle;
}
………
Foundations: Revision
Distributed Systems Principles and Paradigms
Thread Programming
Java Synchronization APIs
– Producer – Consumer
public boolean isFull() {
return this.bFull;
}
public boolean isEmpty() {
return this.bEmpty;
}
public void put(Object item) throws Exception {
synchsronize(this.fullHandle) {
if (this.bFull == true) {
throw new Exception(“The buffer is full!”);
}
}
this.items[this.empty] = item;
this.empty = ((this.empty + 1) % this.items.length);
synchronized(this.emptyHandle) {
this.bFull = this.empty == this.available;
this.bEmpty = false;
}
}
….
Foundations: Revision
Distributed Systems Principles and Paradigms
Thread Programming
Java Synchronization APIs
– Producer – Consumer
public Object get() throws Exception {
synchronized(this.emptyHandle) {
if (this.bEmpty == true) {
throw new Exception(“The buffer is empty!”);
}
}
Object item = this.items[this.available];
this.available = ((this.available + 1) % this.items.length);
synnchronized(this.fullHandle) {
this.bEmpty = this.available == this.empty;
this.bFull = false;
}
return item;
}
}
Foundations: Revision
Distributed Systems Principles and Paradigms
Thread Programming
Java Synchronization APIs
– Producer – Consumer
public class Producer implements Runnable {
private Buffer buffer;
private Object full;
public Consumer(Buffer buffer) {
this.buffer = buffer;
this.full = this.buffer.getFullHandle();
}
public void run() {
try {
while(true) {
synchronized(this.full) {
if (this.buffer.isEmpty() == true) {
this.full.wait();
}
}
Object item = this.produce();
this.buffer.put(item);
}
} catch(Exception ex) {
System.out.println(“Consumer: exception: ” + ex.getMessage());
}
}
…
}
Foundations: Revision
Distributed Systems Principles and Paradigms
Thread Programming
Java Synchronization APIs
– Producer – Consumer
public class Consumer implements Runnable {
private Buffer buffer;
private Object empty;
public Consumer(Buffer buffer) {
this.buffer = buffer;
this.empty = this.buffer.getEmptyHandle();
}
public void run() {
try {
while(true) {
synchronized(this.empty) {
if (this.buffer.isEmpty() == true) {
this.empty.wait();
}
}
Object item = this.buffer.get();
this.consume(item);
}
} catch(Exception ex) {
System.out.println(“Consumer: exception: ” + ex.getMessage());
}
}
…
}
Foundations: Revision
Distributed Systems Principles and Paradigms
Thread Programming
Java Synchronization APIs
– Producer – Consumer
public class ProducerConsumer {
public static void main(String[] args) {
// create the shared buffer and the producer
// and consumer classes
Buffer buffer = new Buffer(10);
Object full = new Object();
Object empty = new Object();
Producer producer = new Producer(buffer);
Consumer consumer = new Consumer(buffer);
// initialize the two threads...
Thread p = new Thread(producer);
Thread c = new Thread(consumer);
// NOTE: let’s play with thread priorities and see
//
what is happening...
p.start();
c.start();
}
}
Foundations: Revision
Distributed Systems Principles and Paradigms
Thread Programming
Java Synchronization APIs
– Producer – Consumer
• Observations
– By separating the two conditions…
– The producer process will only check the buffer ONCE and
will be notified only if there is “room for cheese”
– The consumer process will only check the buffer ONCE and
will be notified only if there is “cheese”
– The consumer will signal the producer
– The producer will signal the consumer
– This implementations solves the following problem:
– The consumer signal itself … (useless)
– The producer signal itself … (useless)
This is possible because
of wait & notify!
Foundations: Revision
Distributed Systems Principles and Paradigms
Thread Programming
Java Synchronization APIs
– Is there more?
• More complex synchronization problems
• Package java.util.concurrent.*:
– Introduced since Java SE 5.0
– A set of utility classes useful in concurrent programming.
– Contains a set of small, standardized, extensible frameworks:
» Executors
» Queues
» Timing
» Concurrent collections
– More on:
http://download.oracle.com/javase/6/docs/api/java/util/concurrent/package-summary.html
Foundations: Revision
Distributed Systems Principles and Paradigms
Thread Programming
Java Synchronization APIs
– For you to read/explore:
• Java Concurrency in Practice, B. Goez, T. Peierls, J.
Bloch, J. Bowbeer, D. Holmes, and D. Lea, Addison
Wesley, 2006.
• Effective Java, 2nd Ed., J. Bloch, Prentice Hall, 2008.
(Chap. 10)
Foundations: Revision
Distributed Systems Principles and Paradigms
Thread Programming
 .NET Synchronization APIs
– Main APIs:
– Language:
– lock(object obj) {…} statement (same as synchronized)
– Namespace :
– System.Threading
– Classes:
– Basic thread management:
– Thread, ThreadPriority, ThreadStart,
ThreadPool
– Synchronization:
– ManualResetEvent, AutoResetEvent
– Interlocked, Monitor, Semaphore, Mutex
– ReaderWriteLock
– Timing:
– Timer
Foundations: Revision
Distributed Systems Principles and Paradigms
Thread Programming
 .NET Synchronization APIs
– lock(object obj) { … }
– Provides a synchronization context in which only one
thread is allowed to execute.
– All the threads that open a lock statement on the same
obj reference are synchronized and mutually excluded.
– There is no equivalent for the synchronized method
modifier in Java.
Foundations: Revision
Distributed Systems Principles and Paradigms
Thread Programming
 .NET Synchronization APIs
– System.Threading.Thread
– Represents a Thread of execution.
– The class is sealed (final in Java) and cannot be
extended.
– In order to create a thread it is necessary to provide a
delegate (pointer to a method) ,which represents the
method that will be run inside the thread.
– .NET threads offer the basic APIs:
– Start, Stop, Suspend, Resume, Join, Sleep,
SpinWait, Interrupt
– IsBackGround, IsAlive, Priority, Name, …
Foundations: Revision
Distributed Systems Principles and Paradigms
Thread Programming
 .NET Synchronization APIs
– System.Threading.ThreadStart
– Provides a simple way to configure a method that will
be run in a thread.
– Wraps a delegate: void methodname()
– System.Threading.ParametrizedThreadStart
– The same as the previous one but accepts a method
that has a variable number of parameters
– Wraps a delegate: void methodname(param obj[])
Foundations: Revision
Distributed Systems Principles and Paradigms
Thread Programming
 .NET Synchronization APIs
– System.Threading.ThreadPriority
– Provides an enumeration of values that allow to set a
qualitative value for the priority of threads
– Values:
–
–
–
–
–
Highest
AboveNormal
Normal
BelowNormal
Lowest
– Operating systems are not requested to honor the
value set for the priority
Foundations: Revision
Distributed Systems Principles and Paradigms
Thread Programming
 .NET Synchronization APIs
– System.Threading.ThreadPool
– Provides a simple implementation of a pool of threads.
– It can be used to post work items that will be executed
asynchronously.
– Main features:
– MinThreads, MaxThreads
– QueueUserWorkItem(WaitCallback)
– QueueUserWorkItem(WaitCallback, object)
– Where:
– delegate: void WaitCallback(object stateInfo);
Foundations: Revision
Distributed Systems Principles and Paradigms
Thread Programming
 .NET Synchronization APIs
– ManualResetEvent & AutoResetEvent
– They provide the corresponding features implemented
in:
– Object.wait(),
– Object.notify(), and Object.notifyAll()
– A xxxResetEvent provides the following:
– Encapsulates a binary state (signaled, not signaled).
– Provides a handle for threads to wait until the state is
signaled.
– Provides other threads with a method to signal the state.
– Can be initialized in any state (signaled, not signaled).
Foundations: Revision
Distributed Systems Principles and Paradigms
Thread Programming
 .NET Synchronization APIs
– ManualResetEvent & AutoResetEvent
– ManualResetEvent provides the wait handle that
needs to be explicitly reset in order to signaled again.
– AutoResetEvent returns to its original state once
signaled.
– Operations:
– ManualResetEvent.WaitOne: waits for a signal
– ManualResetEvent.Set: signals a thread that is waiting
– ManualResetEvent.WaitAny, WaitAll: more complex
waiting patterns.
– ManualResetEvent.Reset: resets the status of the object.
Foundations: Revision
Distributed Systems Principles and Paradigms
Thread Programming
 .NET Synchronization APIs
– Interlocked
– Provides a synchronization context in which it is
possible to increment or decrement an integer value.
– It is a more practical alternative to lock(…) { … }.
– Operations:
– Interlock.Increment(ref int intValue)
– Interlock.Decrement(ref int intValue )
– The use of this structure does not require the caller to
setup any guarded region.
Foundations: Revision
Distributed Systems Principles and Paradigms
Thread Programming
 .NET Synchronization APIs
– Monitor
– Provides an implementation of the monitor concept.
– It is a static class that operated on object references.
– Operations:
–
–
–
–
Monitor.Enter(object handle)
Monitor.Wait(object handle)
Monitor.Pulse, PulseAll(object handle)
Monitor.Exit(object handle)
– Observations:
– Enter and Exit are used to acquire/reacquire the lock on handle.
– Wait is used to temporarily release the lock and block the current thread.
– Pulse/PulseAll, to signal one/all threads that are blocked on a wait
call.
Foundations: Revision
Distributed Systems Principles and Paradigms
Thread Programming
 .NET Synchronization APIs
– Semaphore
– Provides an implementation of the integer semaphore
concept.
– Limits the number of threads that can access a
resource concurrently.
– Can be used to synchronize both Threads and
Processes.
– Operations:
– Creation of an integer OS/local semaphore
– Wait / Release operations.
Foundations: Revision
Distributed Systems Principles and Paradigms
Thread Programming
 .NET Synchronization APIs
– Mutex
– Provides an implementation of a binary semaphore.
– Can be used to synchronize both Threads and
Processes.
– Only one thread/process at time owns the mutex.
– Operations:
– Creation of an integer OS/local mutex
– WaitOne / ReleaseMutex operations.
Foundations: Revision
Distributed Systems Principles and Paradigms
Thread Programming
 .NET Synchronization APIs
– ReaderWriterLock
– Defines a lock that supports single writers and
multiple readers.
– The ReaderWriterLock object is the shared instance
on which the locks are acquired and released.
– Operations:
–
–
–
–
–
–
IsReaderLockHeld
IsWriterLockHeld
AcquireReaderLock
AcquireWriterLock
ReleaseReaderLock
ReleaseWriterLock
Foundations: Revision
Distributed Systems Principles and Paradigms
Thread Programming
 .NET Synchronization APIs
– Timer
– Provides a way to executed a method at a repeated
constant interval.
– Operations:
– Timer creation:
– Timer(TimerCallback)
– Timer(TimerCallback, object obj,
int start, int interval)
– …
– Timer management:
– Timer.Chnage(int start, int interval)
– Timer.Dispose()
Foundations: Revision
Distributed Systems Principles and Paradigms
Thread Programming
.NET Synchronization APIs
– .NET Parallel Extensions
• Implemented since .NET 4.0
• Provides advanced features for parallel programming
–
–
–
–
–
–
Task Parallel library
Parallel LINQ (PLINQ)
Data Structures for Parallel Programming
Parallel Diagnostic Tools
Task Factories
Task Schedulers
• Reference:
– http://msdn.microsoft.com/en-us/library/dd460693.aspx
Foundations: Revision
Distributed Systems Principles and Paradigms
Thread Programming
.NET Parallel Extensions
Foundations: Revision
Distributed Systems Principles and Paradigms
Thread Programming
.NET Synchronization APIs
– Examples
• ThreadPriorities
• Producer Consumer
– By using lock
– By using xxxResetEvents
Foundations: Revision
Distributed Systems Principles and Paradigms
Assignment 1
Foundations: Revision
Distributed Systems Principles and Paradigms
Assignment 1
 Multithreaded Dictionary Server
– Concept
• Build a client-server system
• The server maintains a dictionary
• The clients can query the dictionary for word meaning
– Demonstrates the use of
• Sockets (communication among nodes)
• Threads (provide concurrent requests management
and better performance)
Problem: ******** Using a client-server architecture, design and implement a multi-threaded server that returns the meaning of a word as stored in a dictionary. Belo
Foundations: Revision
Distributed Systems Principles and Paradigms
Assignment 1
Scenario
Process Request
Process Request
Client Process
Client Process
Client Process
Server
Process
Process Request
Foundations: Revision
Distributed Systems Principles and Paradigms
Assignment 1
Guidelines
– Client application:
• Implement a method to query the dictionary:
– Input:
» string (word to look for)
– Output:
» status code (found, not found, error)
» string (meaning(s) of the word) eventually array
• Use sockets:
– TCP and UDP are ok
– Provide a reliable communication
Foundations: Revision
Distributed Systems Principles and Paradigms
Assignment 1
Guidelines
– Server application:
• Implement an server (always on)
• Use any of
– Thread per request
– Thread per connection
• Dictionary
– Maintain the dictionary in a file
– Maintain the index of the dictionary in a separate file
– Use indexing for speeding up search
Foundations: Revision
Distributed Systems Principles and Paradigms
Assignment 1
Guidelines
– General considerations
• Provide error handling for all the functions
– IO and input from user
– Network communication
• Fully customizable from console
– Port & Address
– Word to look for
– Dictionary and index file to use
Foundations: Revision
Distributed Systems Principles and Paradigms
Assignment 1
Marking, Schedule, and Venue
– Total marks assigned: 10
– Deadline: Monday 31 August
– What to deliver:
• Source code (listing of all the files, zipped)
• Documentation (report, class documentation)
• Mail to: [email protected]
– Demonstration:
• Will be scheduled in Lab 217
• More on LMS about date and time.
Foundations: Revision
Distributed Systems Principles and Paradigms
Summary
Brief review of
– Socket programming
• Focus on inter platform/language communication
– Thread programming
• Focus on concurrency
• More to come on the LMS website
– Assignment 1
• Description
• Deadline and details