[Beowulf] detection and diagnosis of PCI bus saturation

Roy L Butler roy.butler at jpl.nasa.gov
Fri Sep 8 14:57:38 PDT 2006


Hernando,

H.Vidal, Jr. wrote:
> Hello.
> 
> Does anybody here have any thoughts, experiences, or references
> to diagnostic methods when looking into potential PCI bus saturation?
> 
> If one is moving data across the bus, either under CPU copies or under
> DMA/bus master, are there canonical methods to detect, perhaps
> in bridge chip registers or other, the state and/or counts of
> bus collisions and/or saturation?
> 
> Your help would be greatly appreciated. Thanks.
> 
> Hernando Vidal, Jr.
> Tesseract Technology
> _______________________________________________
> Beowulf mailing list, Beowulf at beowulf.org
> To change your subscription (digest mode or unsubscribe) visit 
> http://www.beowulf.org/mailman/listinfo/beowulf
> 

Solaris provides "busstat", for their hardware that supports it - 
definitely SPARC, possibly Intel.  I'm guessing that's not your OS 
though...  The source is available, but it relies on counters maintained 
by the drivers, so it's probably not real portable.

For Linux, something could probably be written on top of IBM's kprobes 
(again, for hardware that supports it).  I think the place to start 
looking would be around in the I/O MMU code.

For OS X, Dtrace will be bundled in some future release, putting it on 
the same footing as kprobes, but then I don't suspect this is your 
platform for HPC (big locks, semi-microkernel).

For Windows, there's perfmon, right?  Just kidding. :)


Roy




More information about the Beowulf mailing list