[RC5] ogr-25

Rick Murphy rmurphy at mitre.org
Wed Aug 9 08:21:38 EDT 2000


At 01:46 PM 8/8/2000 +0100, Mark Hill wrote:
>IMO, these 6-stubs are still too big. Less than half of the PCs here
>doing OGR have yet to return a work unit today, and it's 1335. One of
>the PCs just returned a 420 Gnode work unit - it's got the fastest
>processors in the building, and still took 28 hours to complete.
>Several other 200+GN units have also come in today.


I've got to agree. There's something not right in how long these stubs are 
taking - the duration has nothing to do with the node count. For example, 
on a P3/650:

[Aug 07 20:27:01 UTC] Completed OGR stub 25/1-10-2-15-30-4 (155,200,455,141 
nodes)
                       0.00:34:18.47 - [3,473,417.95 nodes/sec]
[Aug 07 20:27:01 UTC] Loaded OGR stub 25/1-10-2-15-29-8
[Aug 07 20:27:01 UTC] Summary: 1 OGR packet (7.14 Gnodes)
                       0.00:34:18.47 - [3.47 Mnodes/s]
[Aug 07 20:27:01 UTC] 18 OGR packets (18 work units) remain in buff-in.ogr
[Aug 07 20:27:01 UTC] 1 OGR packet (1 work unit) is in buff-out.ogr
[Aug 08 09:11:32 UTC] Completed OGR stub 25/1-10-2-15-29-8 (172,332,607,587 
nodes)
                       0.12:44:38.62 - [3,756,272.44 nodes/sec]


The first stub is 155Gnodes, the second 172Gnodes. The first took 35 
minutes, the second almost 13 hours. Why such a difference? Notice that the 
nodes/sec rates are about the same, so it wasn't system load. So far, I'm 
seeing many more 10+ hour stubs on this machine than 30 minute ones. For my 
lowly Pentium 166/MMX the defaults are sending me *TWO WEEKS* worth of 
stubs per download. I'm trying to figure out how to limit the fetch to two 
stubs.
         -Rick


--
To unsubscribe, send 'unsubscribe rc5' to majordomo at lists.distributed.net
rc5-digest subscribers replace rc5 with rc5-digest



More information about the rc5 mailing list