[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

20040610: LDM changes on Rossby, latencies (cont.)

>From: Mike Voss <address@hidden>
>Organization: SJSU
>Keywords: 200406052141.i55LfstK011909 IDD HDS latency

Hi Mike,

>There is a bit more to the story. It turns out that rossby, with it's 512 
>Mb of RAM does not perform well with the large (2GB) queue. The main clue 
>was that as soon as I went to the large queue, I started getting latencies. 

This makes sense.

>"TOP" command showed that the "iowait" was up around 90%. And sure enough, 
>as soon as I take the queue down to 400 MB the data flows in and there is 
>small "iowait's" showing in TOP. So, I'm feeding CRAFT from OU as you 
>intended. And, the queue size thing should be irrelevant once I get a new 
>machine with 3 GB of RAM, etc.


By the way, I was forced to logon to rossby the night before last and
change its requests from thelma.ucar.edu to emo.unidata.ucar.edu.  With
the power outage at the NCAR Mesa Lab Tuesday night, the original
thelma (Sunfire 480R) went down, and we were forced to press a machine
housed at the UPC into service as its replacement.  That machine, a
Sunfire 280, was not able to meet the demand that was being serviced by
the original thelma.  I reviewed all of the machines feeding from
thelma that I had logins to and took the liberty of changing their
requests to point to emo.  I intended to write you an note explaining
the change yesterday, but I was unable to get to it.


Thanks for the update!


NOTE: All email exchanges with Unidata User Support are recorded in the
Unidata inquiry tracking system and then made publically available
through the web.  If you do not want to have your interactions made
available in this way, you must let us know in each email you send to us.