Some javascript/style in this post has been disabled (
why?)
Hello,
For those interested, here is the patch for the below mentioned items.
NLog-1.0\src\NLog\Targets\Wrappers\AsyncTargetWrapper.cs
208c208
< } while (_flushAll);
---
> } while (_flushAll || RequestQueue.RequestCount > BatchSize);
NLog-1.0\src\NLog\Win32\Targets\MSMQ.cs
189a190,224
> /// Writes the specified array of logging events to a queue specified in the Queue
> /// parameter.
> /// </summary>
> /// <param name="logEvents">An array of <see cref="LogEventInfo "/> objects.</param>
> /// <remarks>
> /// This function override increases performance of batches by not having to
> /// check the queues existence and connect to it for each logEvent.
> /// </remarks>
> protected internal override void Write(LogEventInfo[] logEvents)
>
{
> if (_queue == null)
> return;
>
> if (!MessageQueue.Exists(Queue))
> {
> if (CreateQueueIfNotExists)
> MessageQueue.Create(Queue);
> else
>
return;
> }
>
> using (MessageQueue mq = new MessageQueue(Queue))
> {
> foreach (LogEventInfo logEvent in logEvents)
> {
> Message msg = PrepareMessage(logEvent);
> if (msg !=
null)
> {
> mq.Send(msg);
> }
> }
> }
> }
>
> /// <summary>
Thanks,
Jake Milnes
----- Original Message ----
From: Jake Milnes
<
[hidden email]>
To:
[hidden email]Sent: Tuesday, June 3, 2008 10:24:25 AM
Subject: Performance Enhancements
Hello,
For a new project of ours, we looked at several different logging frameworks and decided on using NLog. Thanks for making NLog great!
We made a couple of minor performance enhancements to speed up our particular usage of NLog. I would like to find out if others are interested in these performance enhancements or if anyone sees any problems with these:
1. Added Write(LogEvent[]) override to the MSMQ class. This removes the overhead of checking for the queue's existence, connecting and disconnecting for every insert if you are batching the inserts (e.g. through a buffered wrapper or async wrapper).
2. Updated the Async wrapper class so that it doesn't
go back to sleep after processing a callback if there are still enough items
on the queue for a full batch.
We did some decent performance testing benchmarks before and after, and saw significant gains in the applicable test cases (by roughly a factor of 5). I can provide more detail on the performance test if someone is truly interested (the test results currently are not written up for release).
Thanks,
Jake Milnes
-------------------------------------------------------------------------
Check out the new SourceForge.net Marketplace.
It's the best place to buy or sell services for
just about anything Open Source.
http://sourceforge.net/services/buy/index.php_______________________________________________
Nlog-list mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/nlog-list
