[concurrency-interest] j.u.c/backport performance on Windows

Gregg Wonderly gregg at cytetech.com
Tue Apr 3 12:41:10 EDT 2007



Peter Kovacs wrote:
> On 4/4/07, Gregg Wonderly <gregg at cytetech.com> wrote:
>> Peter Kovacs wrote:
>> > Yes, this is basically what I am doing. Only I use a "custom" class
>> > for holding the result
>> > 
>> (http://www.chemaxon.com/shared/pkovacs/concurrent/processors/ScheduledWorkUnitData.java) 
>>
>> > instead of implementing the stock Future interface.
>>
>> So it would seem that there probably are some OS related 'priority' and
>> 'scheduling' thing that might be changing the perceived performance.  
>> Do you
>> have a profiler to run the application with?  This might tell you a 
>> lot more
>> about where the 'latency' is at.
> 
> 
> Yes, I think the hour is slowly aproaching where I cannot get around
> using a profiler. My company has a JProfiler license. Do you think it
> will do the job? Do you know a better one? (I've been thinking about
> looking at the NetBeans profiler [I do not know how it is called],
> because it is free...)

Well, I've become fond of the YourKit Java profiler because it can be started in 
the JVM, and turned on remotely for casual observation from time to time.  When 
you take a snapshot, it is transfered over the network to your computer, but 
also stored on the remote machine for later access.  So, if you request the 
snapshot and the transfer fails, you can go to that machine and FTP the snapshot 
off, for example.

It really has a number of nice features related to remote, periodic performance 
monitoring.

Gregg Wonderly


More information about the Concurrency-interest mailing list