While working on the PX paper for Hotsos, I’ve been trying to show the same issue using different tools that people might have available to them and struggled to get OraSRP working on Solaris 10. From digging around the net I started to realised that the Python distribution included in Solaris 10 isn’t the best. Alex G suggested using the Windows version and the results I was getting didn’t look right.
Anyway, because I’m pushed for time, I abandoned it because I already had plenty of material and didn’t want to waste time fixing it.
Well blow me if, just in the nick of time, the new beta of version 3 has been released. I’m blessed with good fortune. I’ve downloaded the Windows Installer version and, at this early stage, it looks pretty good and is confirming the results I’m seeing with other tools.
For those of you unfamiliar with OraSRP, it’s a free Oracle resource profiler which reads event 10046 trace files and creates a useful and readable HTML resource profile. Now I just need to incorporate it into the paper.
Thanks to Egor Starostin for his tool and thanks to Yuri van Buren for drawing my attention to it.
Phwoar! I just looked at the OraSRP tool and downloaded it. That looks good! I shall be taking some time out this afternoon to have a play with that I recon 🙂
It’s a pretty nice tool … but it would have been cooler if it was written in something I could read … like perl or java 😉
I’m looking forward to testing this. The last time I used OraSRP, I have several problems. Off the top of my head:
– It could not finish processing certain files. I ran it on several platforms (Solaris, Linux, windows). There was a file that it could not process for 2 days.
– I compared output to the Hotsos profiler. There were some numbers that were too far apart… so far that it can not be contributed to rounding off. This does not mean, of course that OraSRP was wrong. I’m just saying the outputs of the two profilers were vastly different in some sections.
James
s/contributed/attributed
🙂
It could not finish processing certain files. I ran it on several platforms (Solaris, Linux, windows). There was a file that it could not process for 2 days.
Based on a very small experience so far, the speed seems blinding! Very impressive.
I compared output to the Hotsos profiler. There were some numbers that were too far apart… so far that it can not be contributed to rounding off. This does not mean, of course that OraSRP was wrong. I’m just saying the outputs of the two profilers were vastly different in some sections.
Interesting. In truth, I’m already getting some strange results but I suspect that’s more to do with my strange tests than the tools. I’ll reserve judgment for now ..
Based on a very small experience so far, the speed seems blinding! Very impressive.
The one that I tried was version 2.x, written in python. I just browsed the site and it looks like this new one is written in D and it is compiled.
But my experience really had nothing to do with speed. What happened is that I gave it a big file (60+ million lines) with lots of BINDs. It ate up all the memory so it could never finish.
There’s certainly still an option to limit the extent of bind analysis to avoid memory problems
Coincidentally I was playing around with this last week in preparation for a phone interview (I thought they might ask about performance problem diagnosis, which they did). I mentioned OraSRP and they hadn’t heard of it and made a note of the name – I think that might have impressed them.
The browser interface is particularly handy when the database is running in a Linux VM or some other less than accessible server. It can also run as a commmand-line utility though.
It can also run as a commmand-line utility though.
I confess that’s how I use it, although I do like the HTML output rather than text.
Just FYI – I posted my initial impression in my blog:
http://jforonda.blogspot.com/2007/02/orasrp-v30b-very-short-initial.html