You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@cassandra.apache.org by Jason Fager <jf...@gmail.com> on 2010/08/06 22:25:33 UTC
interest in minor contention issues?
Hi all,
I'm doing some work with the Java Debugger Interface, and I'm using
Cassandra as a test bed. I've found a few minor contention spots
(example: MessageDigest.getInstance() being called for every
invocation of FBUtilities.hash()), but before I start possibly
spamming your jira with them, I wanted to gauge the general interest
in knowing these kinds of minor issues. I'm happy to report them as
they come up while I'm doing my work, but if you consider them too
minor to worry about until they actually show up as impacting
performance during profiling, I don't want to annoy anyone. Let me
know.
- Jason
Re: interest in minor contention issues?
Posted by Jason Fager <jf...@gmail.com>.
Hi Jonathan,
No problem, thanks for all the work you and everyone on the dev team
do on Cassandra.
I have a small bit of wrapper code for scripting the Java Debugger
Interface, which is up at https://github.com/jfager/jdiscript. I also
have a short writeup on it at
http://jasonfager.com/888-introducing-jdiscript/. In a nutshell, I
just ran Cassandra with the debugger set to fire events on JDI's
MonitorContendedEnterEvent, and handle them by printing out
stacktraces.
If I get some time in the next day or so, I'll take another shot at
the 0.7 codebase.
Best,
Jason
On Tue, Dec 28, 2010 at 3:14 PM, Jonathan Ellis <jb...@gmail.com> wrote:
> Hi Jason,
>
> Thanks for looking into this!
>
> What tool are you using? I belatedly noticed your tickets (we should
> have them resolved for 0.6.9) and it might be useful to do another
> check against the 0.7 code.
>
> On Fri, Aug 6, 2010 at 3:25 PM, Jason Fager <jf...@gmail.com> wrote:
>> Hi all,
>>
>> I'm doing some work with the Java Debugger Interface, and I'm using
>> Cassandra as a test bed. I've found a few minor contention spots
>> (example: MessageDigest.getInstance() being called for every
>> invocation of FBUtilities.hash()), but before I start possibly
>> spamming your jira with them, I wanted to gauge the general interest
>> in knowing these kinds of minor issues. I'm happy to report them as
>> they come up while I'm doing my work, but if you consider them too
>> minor to worry about until they actually show up as impacting
>> performance during profiling, I don't want to annoy anyone. Let me
>> know.
>>
>> - Jason
>>
>
>
>
> --
> Jonathan Ellis
> Project Chair, Apache Cassandra
> co-founder of Riptano, the source for professional Cassandra support
> http://riptano.com
>
Re: interest in minor contention issues?
Posted by Jonathan Ellis <jb...@gmail.com>.
Hi Jason,
Thanks for looking into this!
What tool are you using? I belatedly noticed your tickets (we should
have them resolved for 0.6.9) and it might be useful to do another
check against the 0.7 code.
On Fri, Aug 6, 2010 at 3:25 PM, Jason Fager <jf...@gmail.com> wrote:
> Hi all,
>
> I'm doing some work with the Java Debugger Interface, and I'm using
> Cassandra as a test bed. I've found a few minor contention spots
> (example: MessageDigest.getInstance() being called for every
> invocation of FBUtilities.hash()), but before I start possibly
> spamming your jira with them, I wanted to gauge the general interest
> in knowing these kinds of minor issues. I'm happy to report them as
> they come up while I'm doing my work, but if you consider them too
> minor to worry about until they actually show up as impacting
> performance during profiling, I don't want to annoy anyone. Let me
> know.
>
> - Jason
>
--
Jonathan Ellis
Project Chair, Apache Cassandra
co-founder of Riptano, the source for professional Cassandra support
http://riptano.com
Re: interest in minor contention issues?
Posted by Jason <jf...@gmail.com>.
Right now there are just a couple. It sounds like there's enough interest, so I'll just add them as they come up, and if it starts to be alot, I'll start coalescing them.
Thanks,
Jason
On Aug 6, 2010, at 5:19 PM, Jeremy Hanna <je...@gmail.com> wrote:
> Are you talking about a handful of issues or a large number of issues?
>
> If it's the former, then a Jira issue for each isn't unreasonable. If it's the latter, maybe all of your results could go into one Jira issue with steps to reproduce your results. Then someone could do some more investigation from the Cassandra community.
>
> Either way, thank you for reaching out.
>
> On Aug 6, 2010, at 4:01 PM, Jason Fager wrote:
>
>> My main concern is that I'm not primarily interested in Cassandra at
>> this point, I'm interested in my JDI project, and formal benchmarking
>> of Cassandra on each item I find is probably more work than I want to
>> put in at this time. If there's an easy benchmark that's already
>> generally agreed upon, I'll be happy to run it first, but otherwise,
>> I'm basically asking if the dev group is interested in reports that
>> won't reach the threshold of having a demonstrated performance impact,
>> or if those kinds of reports are generally just regarded as annoying.
>>
>>
>>
>> On Fri, Aug 6, 2010 at 4:37 PM, Gary Dusbabek <gd...@gmail.com> wrote:
>>> I think if there is a demonstrable increase in performance or other
>>> obvious wins, then you should go ahead and submit jiras (and patches).
>>>
>>> Gary.
>>>
>>> On Fri, Aug 6, 2010 at 15:25, Jason Fager <jf...@gmail.com> wrote:
>>>> Hi all,
>>>>
>>>> I'm doing some work with the Java Debugger Interface, and I'm using
>>>> Cassandra as a test bed. I've found a few minor contention spots
>>>> (example: MessageDigest.getInstance() being called for every
>>>> invocation of FBUtilities.hash()), but before I start possibly
>>>> spamming your jira with them, I wanted to gauge the general interest
>>>> in knowing these kinds of minor issues. I'm happy to report them as
>>>> they come up while I'm doing my work, but if you consider them too
>>>> minor to worry about until they actually show up as impacting
>>>> performance during profiling, I don't want to annoy anyone. Let me
>>>> know.
>>>>
>>>> - Jason
>>>>
>>>
>
Re: interest in minor contention issues?
Posted by Jeremy Hanna <je...@gmail.com>.
Are you talking about a handful of issues or a large number of issues?
If it's the former, then a Jira issue for each isn't unreasonable. If it's the latter, maybe all of your results could go into one Jira issue with steps to reproduce your results. Then someone could do some more investigation from the Cassandra community.
Either way, thank you for reaching out.
On Aug 6, 2010, at 4:01 PM, Jason Fager wrote:
> My main concern is that I'm not primarily interested in Cassandra at
> this point, I'm interested in my JDI project, and formal benchmarking
> of Cassandra on each item I find is probably more work than I want to
> put in at this time. If there's an easy benchmark that's already
> generally agreed upon, I'll be happy to run it first, but otherwise,
> I'm basically asking if the dev group is interested in reports that
> won't reach the threshold of having a demonstrated performance impact,
> or if those kinds of reports are generally just regarded as annoying.
>
>
>
> On Fri, Aug 6, 2010 at 4:37 PM, Gary Dusbabek <gd...@gmail.com> wrote:
>> I think if there is a demonstrable increase in performance or other
>> obvious wins, then you should go ahead and submit jiras (and patches).
>>
>> Gary.
>>
>> On Fri, Aug 6, 2010 at 15:25, Jason Fager <jf...@gmail.com> wrote:
>>> Hi all,
>>>
>>> I'm doing some work with the Java Debugger Interface, and I'm using
>>> Cassandra as a test bed. I've found a few minor contention spots
>>> (example: MessageDigest.getInstance() being called for every
>>> invocation of FBUtilities.hash()), but before I start possibly
>>> spamming your jira with them, I wanted to gauge the general interest
>>> in knowing these kinds of minor issues. I'm happy to report them as
>>> they come up while I'm doing my work, but if you consider them too
>>> minor to worry about until they actually show up as impacting
>>> performance during profiling, I don't want to annoy anyone. Let me
>>> know.
>>>
>>> - Jason
>>>
>>
Re: interest in minor contention issues?
Posted by Jason Fager <jf...@gmail.com>.
My main concern is that I'm not primarily interested in Cassandra at
this point, I'm interested in my JDI project, and formal benchmarking
of Cassandra on each item I find is probably more work than I want to
put in at this time. If there's an easy benchmark that's already
generally agreed upon, I'll be happy to run it first, but otherwise,
I'm basically asking if the dev group is interested in reports that
won't reach the threshold of having a demonstrated performance impact,
or if those kinds of reports are generally just regarded as annoying.
On Fri, Aug 6, 2010 at 4:37 PM, Gary Dusbabek <gd...@gmail.com> wrote:
> I think if there is a demonstrable increase in performance or other
> obvious wins, then you should go ahead and submit jiras (and patches).
>
> Gary.
>
> On Fri, Aug 6, 2010 at 15:25, Jason Fager <jf...@gmail.com> wrote:
>> Hi all,
>>
>> I'm doing some work with the Java Debugger Interface, and I'm using
>> Cassandra as a test bed. I've found a few minor contention spots
>> (example: MessageDigest.getInstance() being called for every
>> invocation of FBUtilities.hash()), but before I start possibly
>> spamming your jira with them, I wanted to gauge the general interest
>> in knowing these kinds of minor issues. I'm happy to report them as
>> they come up while I'm doing my work, but if you consider them too
>> minor to worry about until they actually show up as impacting
>> performance during profiling, I don't want to annoy anyone. Let me
>> know.
>>
>> - Jason
>>
>
Re: interest in minor contention issues?
Posted by Rob Coli <rc...@digg.com>.
On 8/6/10 1:37 PM, Gary Dusbabek wrote:
> I think if there is a demonstrable increase in performance or other
> obvious wins, then you should go ahead and submit jiras (and patches).
+1, when you have hundreds of threads, even minor contention issues may
be significant.
=Rob
Re: interest in minor contention issues?
Posted by Gary Dusbabek <gd...@gmail.com>.
I think if there is a demonstrable increase in performance or other
obvious wins, then you should go ahead and submit jiras (and patches).
Gary.
On Fri, Aug 6, 2010 at 15:25, Jason Fager <jf...@gmail.com> wrote:
> Hi all,
>
> I'm doing some work with the Java Debugger Interface, and I'm using
> Cassandra as a test bed. I've found a few minor contention spots
> (example: MessageDigest.getInstance() being called for every
> invocation of FBUtilities.hash()), but before I start possibly
> spamming your jira with them, I wanted to gauge the general interest
> in knowing these kinds of minor issues. I'm happy to report them as
> they come up while I'm doing my work, but if you consider them too
> minor to worry about until they actually show up as impacting
> performance during profiling, I don't want to annoy anyone. Let me
> know.
>
> - Jason
>