Hi. Thank you for the comments. Overall there is no immediate answer to your question, because the Mserver, OS and hardware characteristics are unkown. On a small (ATOM) system there may be swapping between client/server cost, or network-re-establishment cost in a distributed client/server (cloud) setup. Also some Linux versions show unstable behavior under some page mapping schemes. On 5/20/13 7:11 PM, Stephen P. Morgan wrote:
Hi. I'm working with Masood. When profiling an explained (MAL) version of TPCH Query 1 with stethoscope, we noticed something strange, and would appreciate your comments. First, the MAL program seems to take longer by about 75% (10 to 15 seconds on a given server) to produce output when run at TPCH scale factor 100. (We have similar results at other scale factors.) Second, the program doesn't produce output until about 10-15 seconds after completing sql.exportResult() when run as MAL, looks like a reload/re-establishment of processes.
but does so immediately after completing sql.exportResult() when run as SQL. There doesn't seem to be anything going on in that final 10-15 second interval in the MAL program, at least according to stethoscope. The stethoscope does not show OS behavior, but this can be triggered using the proper parameters. See stethoscope --help, which gives you access to the basis OS performance parameters. What might be happening during that time?
Perhaps naively, we imagined that the MAL and SQL timings would be about the same. They use the same protocol for interaction.
NOTE 1; be aware of the difference between hot and cold runs. NOTE 2; be aware that even starting a browser at the same time on the test machine can lead to resource (memory/cpu) fights and therefore possible performance differences. regards, Martin
Thanks,
Steve
------------------------------------------------------------------------ *From:* Martin Kersten
*To:* users-list@monetdb.org *Sent:* Fri, May 17, 2013 5:35:50 PM *Subject:* Re: Using something like "trace" within the MAL interpreter Hi
The best way to trace is to use the 'stethoscope' http://www.monetdb.org/Documentation/Manuals/MonetDB/Profiler/Stethoscope
For MAL programs one can use something like: profiler.activate("ISTestmrw"); profiler.setAll(); profiler.openStream(); profiler.start(); ...MAL program to follow.... profiler.stop();
regards, Martin
On 5/18/13 2:29 AM, Masood Mortazavi wrote:
Hi, The "trace" commands works well when mclient is in the SQL interpreter mode. How about when mclient is in MAl interpreter mode? Is there a way to use the "trace" command in this mode? Best regards, Masood
_______________________________________________ users-list mailing list users-list@monetdb.org mailto:users-list@monetdb.org http://mail.monetdb.org/mailman/listinfo/users-list
_______________________________________________ users-list mailing list users-list@monetdb.org mailto:users-list@monetdb.org http://mail.monetdb.org/mailman/listinfo/users-list
_______________________________________________ users-list mailing list users-list@monetdb.org http://mail.monetdb.org/mailman/listinfo/users-list