M2G Posted May 31, 2012 Report Posted May 31, 2012 (edited) Authored by: Dr. Joe Hummel Duration: 3h 44mLevel: IntermediateReleased: 5/29/2012Discusses the design of asynchronous and parallel applications using the new Task-based model available in .NET 4 and Silverlight 5.Contents: Understanding the Dangers of Concurrency 01:22:56Introductions and Agenda The pitfalls of concurrency Correctness guarantees: safety, liveness Terminology: race conditions, critical sections Race conditions involving shared resources Demo 1: parallelizing a Logfile Search app Demo 1: source code layout Demo 1: from sequential to parallelDemo 1: summary of why it's broken Solving race conditions --- an overview Solution 1: locking Demo 2: solution using locking Solution 2: interlocking Demo 3: solution using interlocking Solution 3: lock-free Demo 4: lock-free solution The danger of shared objects Thread-safety and solutions to shared objects Demo 5: using thread-safe objects Synchronization primitives The concurrent data structures: thread-safe collections Performance lessons... Demo 6: improving performance Summary and References Execution Model and Types of Parallelism 01:04:06 Introductions and Agenda Tasks: developer's role vs. .NEt's role Execution model: high-level view Task granularity: just how lightweight? Execution model: tasks vs. threads vs. cores Custom task scheduling Additional observations regarding task scheduler Work-stealing in detail Two important assumptions: short-lived, random order Fairness option to execute tasks in order LongRunning option for tasks > 1-2 seconds Demo 1: executing 100 long-running tasks Demo 1: solution attempt with default behavior Demo 1: solution attempt with long-running option Solution for 100 long-running tasks Demo 2: proper solution in action Parallel.For = more concise solution Types of Parallelism Data parallelism Task parallelism Dataflow parallelism Embarrassingly parallel TPL support for parallelism types Parallel.For, .Foreach, .Invoke Structured (fork-join) parallelism Demo 3: Mandelbrot app overview Demo 3: sequential version Demo 3: Parallel.For version Data partitioning Custom data partitioning Exception handling with Parallel class Breaking out of a Parallel loop Cancelling a Parallel loop Demo 4: Mandelbrot app with cancellation Summary and References Designs and Patterns for Parallel Programming 01:17:20Introductions and Agenda Design Problem 1: 100+ CPU-intensive operations Design Problem 2: download 20+ web pages Design Problem 3: application logging task Parallel Patterns --- master list Pipeline pattern Dataflow pattern Increasing parallelism in pipeline and dataflow Concurrent Data Structures --- master list ConcurrentQueue T Demo 1: overview of Netflix data mining app Demo 1: overview of parallel solution (v1) Demo 1: implementation with ConcurrentDictionary T Demo 1: results Producer-Consumer pattern Implementation of Producer-Consumer with BlockingCollection T Demo 2: Netflix solution using Producer-Consumer (v2) Demo 2: results MapReduce pattern Implementing MapReduce Demo 3: Netflix using MapReduce Demo 3: results Parallel LINQ (PLINQ) Demo 4: Netflix with PLINQ Speculative Execution pattern APM: Asynchronous Programming Model pattern APM example of async file I/O Demo 5: Stock History with Speculative Execution and APM Parallel I/O design challenge Summary and ReferencesDownload: http://ge.tt/6LO6rTI/v/0(Asteptati pana il termin de urcat, probleme cu site-ul de filesharing)Cele de aici: Async and Parallel Programming: Application Design - Online Training Course for .NET Developers Edited May 31, 2012 by M2G Quote
phreak Posted May 31, 2012 Report Posted May 31, 2012 The pitfalls of concurrency : loss of sanity. Quote
