• Who is Mark Kraus?
• What is Concurrent Programming?
• Examples of concurrent programming in PowerShell
• What is the Producer-Consumer Pattern?
• Demo

• Senior Systems Engineer @ LinkedIn
• PowerShell Core Project Collaborator
• Web Cmdlets Contributor
• Author of Get-PowerShellBlog
• Co-Author of The PowerShell Conference Book

Concurrent computing is a form of computing in which several computations are executed during overlapping time periods—concurrently—instead of sequentially (one completing before the next starts). (Wikipedia)

• Core
  • CPU
• Thread
  • "Virtual CPU" that runs on a Core
• Task
  • Code that runs in a Thread

• Tasks complete one after the other
• Tasks never run at the same time
• Uses a single thread
• Slowest

Task 1: +
Task 2: =

Core 1: +++++++++++++++++++++++++++===========================

• Uses one core per task
• Not possible on single core system
• Faster but more expensive

Task 1: +
Task 2: =

Core 1:   +++++++++++++++++++++++++++
Thread 1: +++++++++++++++++++++++++++

Core 2:   ===========================
Thread 1: ===========================

• Uses multiple threads
• Can run on single core
• Slower but cheaper

Task 1: +
Task 2: =

Core 1:   +++===++++++======+++===++++++======+++++++++=========
Thread 1: +++++++++++++++++++++++++++
Thread 2: ===========================

A quick recap of the available concurrent programming methods in PowerShell.

• PowerShell Jobs
• PoshRSJob
• PSThreadJob
• Runspaces

• Producer creates (produces) items
• Consumer uses (consumes) items from the Producer
• The PowerShell Pipeline is a Producer-Consumer

Get-Job | Wait-Job | Receive-Job

• Get-Job produces a list of Jobs
• Wait-Job consumes Get-Jobs results then also produces a list of jobs
• Receive-Job consumes Wait-Job results

• Multiple Producers of the same item
• Multiple Consumers of those items
• Producers and Consumers Threads
• Threads are ScriptBlocks
• Increase and decrease Producers and consumers as needed

The Widget Factory turns monads into widgets.

Receiving:

• Multiple suppliers deliver monads at various times
• Sometimes Multiple suppliers deliver at once
• Receiving has multiple delivery bays
• All deliveries feed to a single monad line
• monads travel to Manufacturing

Manufacturing:

• Line workers take monads and build widgets
• Multiple line workers
• When their current widget is done they grab the next available monad
• Sometimes need Monads at the same time
• Widgets go out to Shipping

Shipping:

• Multiple pickups will be made by multiple distributors
• Sometimes multiple distributors arrive at the same time
• All shipments must have X number of widgets
• Shipping bundles the widgets and puts them on distributor trucks

• Sometimes we need to deal with more than one source of data
• Sometimes the amount of data in is to great to process serially
• Sometimes there are multiple consumers of our processed data
• Most often, this work is being broken up into batches.
• When one batch completes the next starts.
• Leads to under-utilized threads

We want all or workers 100% utilized at all times unless there is no work to be done!

• Hybrid exchange with On-prem and On-cloud mailboxes
• Must enumerated all mailboxes in both
• Getting Inbox rules is a slow and expensive Task
• Multiple service accounts needed to get rules
• Service accounts cannot constantly open and close PowerShell sessions
• Throttling per-user considerations
• Rules then need to be processed and compiled into a report
• Logging and error detection

Gist

• Thread-Safe Collections
• BlockingCollection<PSObject> Link
• ConcurrentQueue<PSObject> Link
• ConcurrentStack<Int> link
• RunspacePool Link
• PowerShell Link

using namespace System.Collections.Concurrent
$Queue = [BlockingCollection[PSObject]]::new(
    [ConcurrentQueue[PSObject]]::new()
)
$RunspacePool = [runspacefactory]::CreateRunspacePool(1,4)
$RunspacePool.Open()
$Runspace = [PowerShell]::Create()
$Runspace.RunspacePool = $RunspacePool
$Runspace.AddScript($ScriptBlock).AddArgument($Queue)
$Runspace.BeginInvoke()

• BlockingCollection.GetConsumingEnumerable()
• Blocks the thread until it can take an item
• BlockingCollection.CompleteAdding()
• Marks the collection complete
• GetConsumingEnumerable() will then exit the loop

Thread 1:

 foreach ($LogEntry in $LogQueue.GetConsumingEnumerable()) {
     # do stuff
 }

Thread 2:

$LogQueue.Add($Message1)
$LogQueue.Add($Message2)
$LogQueue.Add($Message3)
$LogQueue.CompleteAdding()

• ConcurrentStack<Int>
• Used to track how peer threads
• last thread completes the output queue

 $FileConsumerStack.Add(
    [System.Threading.Thread]::CurrentThread.ManagedThreadId
)
# do stuff
# Remove a thread from the stack
$null = $FileConsumerStack.Take()
# Close LogQueue if this is the last thread
if($FileConsumerStack.Count -lt 1) {
    $FileConsumerStack.CompleteAdding()
    $LogQueue.CompleteAdding()
}

• Enumerate files in multiple folders
• Take the Filenames and reverse them
• Log new names

Main Demo

• Threads are started before any folders are supplied
• Toggling the number of File Producers and File Consumers
• Can have more File producers than Folders
• Single log consumer to prevent file locks

• https://github.com/markekraus
• Material: https://github.com/markekraus/ConcurrentPowerShellProducerConsumer
• https://get-powershellblog.blogspot.com/
• @markekraus