Slides and Notes 11-August-2011

The end of the ingest of E2007 has been going swimmingly: 50-60MB/sec transfer speeds (>4TB/day).

Note below that for an interval sam stopped reading and then stopped writing while the reading speed doubled (as did the ethernet transmit rate!). This would have transferred > 190GB, but the largest file I find in the log is only 1.3GB??

In the log file it says that at 1:35 if found 66 collections unwritten, and by 1:59 there was only 1. In the meantime it acquired another 64 which were also written out, so that explains the data quantity if not the reason for the change. At 1:32 it started to throttle, presumably for lack of space in the collector area. The obvious first guess--that read+write is a constant--can be seen to be wrong by looking at the output of iostat.

And chuma handles the load easily.

If I look at the output of iostat (while the above transmission is still going strong), I can see that the read speed hits as high as 140MB/sec and the write speed over 180MB/sec, and the sum of the two can be over 250MB/sec.

Somewhat later I started another process on klausz to copy large files to maloof E2008. As you can see chuma's CPU rises, and the sum of klausz' and maloof's CPU usage is a little higher than that of sam alone.

And from iostat:


Summary:


Modified 11-August-2011 at 10:47

http://icecube.wisc.edu/~jbellinger/StorHouse/11Aug2011
Previous notes Next notes Main slide directory

Please contact jnbt@hep.physics.wisc.edu if you have trouble accessing the information on this page.