Re: PROBLEM: Sending mail-attachment > 45k with Netscape via sendmail

Pete Wyckoff (pw@dancer.ca.sandia.gov)
Thu, 25 Feb 1999 10:16:24 -0800

I played with your client/serv codes and, sure, a READ_SIZE of 15 is
much slower than one of 100, but I don't notice any stalls. Running
strace with timestamps shows the alternating calls to select and read,
but the time between calls is only ever a few ms.

Here's the client-side timings for a range of sizes. There's a nice
exponential decay in user and system times as the receive size goes up,
which makes sense. Is this a case of "don't do that", or do you see
real stalls?

size 10 0.69 user 24.78 system 0:25.48 elapsed 99% CPU
size 20 0.33 user 12.34 system 0:12.69 elapsed 99% CPU
size 30 0.22 user 8.13 system 0:08.36 elapsed 99% CPU
size 40 0.15 user 6.11 system 0:06.29 elapsed 99% CPU
size 50 0.11 user 2.38 system 0:02.76 elapsed 90% CPU
size 60 0.15 user 1.25 system 0:02.30 elapsed 60% CPU
size 70 0.12 user 1.36 system 0:02.30 elapsed 64% CPU
size 80 0.13 user 0.85 system 0:02.30 elapsed 42% CPU
size 90 0.14 user 0.86 system 0:02.30 elapsed 43% CPU
size 100 0.02 user 0.66 system 0:02.30 elapsed 29% CPU

-- Pete
---------------------------------------------
Pete Wyckoff | wyckoff@ca.sandia.gov
Sandia National Labs | 925 294 3503 (voice)
MS 9011, P.O. Box 969 | 925 294 1225 (fax)
Livermore, CA 94551 |

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/