Fred Lindberg wrote:
>
> On Tue, 18 May 1999 11:22:11 -0400, Mark E Drummond wrote:
>
> >What I am wondering about is the apparently long processing time for
> >messages. According to these stats we are processing only ~2722 messages
> >per day (~1.9 per minute) at ~23K each. Yet the average message queue
> >time is 527 seconds?! Why would a message sit in the queue for so long
> >with such a light load? Here's the numbers:
>
> First, you have a very lightly loaded machine, and it's goofing off
> most of the time ;-) This explains the 1.9 messages/min. We've had
> qmail do 1000/min on simple hardware, admittedly with ideal recipients.
>
> The queue time includes deliveries that were repeatedly deferred (host
> unreachable, user over quota try later, etc) that then finally timed
> out their queue life time(something like 7 days). Thus, the average is
> not very useful.
>
> Average ddelay per success (s): 95.132936: This shows you that the
> delivery attempts that were successful took on average 95 s to
> complete. This does not count all the unsuccessful delivery attempts.
>
> Look at other stats in the package. The time taken to complete 50% or
> 80% of the deliveries or the average time for the first 50 or 80% is a
> much more useful measure.
>
> In summary, looks ok and the numbers are a reflection of the
> recipients, not your qmail installation.
I ran zddelay on the logs after posting this and yes, 82% of all
messages are being delivered in under 1 second, 91% in under 2 seconds,
but the last 1-2% are taking _huge_ amounts of time which it skewing the
results.
Here's the run of zddelay:
Distribution of ddelays for successful deliveries
Meaning of each line: The first pct% of successful deliveries
all happened within doneby seconds. The average ddelay was avg.
doneby avg pct
0.49 0.46 10
0.49 0.46 11
0.49 0.47 12
0.49 0.47 13
0.50 0.47 14
0.50 0.47 15
0.50 0.47 16
0.50 0.47 17
0.50 0.47 18
0.50 0.48 19
0.51 0.48 20
0.51 0.48 21
0.51 0.48 22
0.51 0.48 23
0.51 0.48 24
0.52 0.48 25
0.52 0.48 26
0.52 0.49 27
0.52 0.49 28
0.52 0.49 29
0.53 0.49 30
0.53 0.49 31
0.53 0.49 32
0.53 0.49 33
0.54 0.49 34
0.54 0.50 35
0.54 0.50 36
0.54 0.50 37
0.55 0.50 38
0.55 0.50 39
0.55 0.50 40
0.55 0.50 41
0.56 0.50 42
0.56 0.51 43
0.56 0.51 44
0.57 0.51 45
0.57 0.51 46
0.57 0.51 47
0.58 0.51 48
0.58 0.52 49
0.59 0.52 50
0.59 0.52 51
0.59 0.52 52
0.60 0.52 53
0.60 0.52 54
0.61 0.53 55
0.61 0.53 56
0.62 0.53 57
0.62 0.53 58
0.62 0.53 59
0.63 0.53 60
0.64 0.54 61
0.64 0.54 62
0.65 0.54 63
0.65 0.54 64
0.66 0.55 65
0.67 0.55 66
0.68 0.55 67
0.68 0.55 68
0.69 0.56 69
0.70 0.56 70
0.71 0.56 71
0.73 0.56 72
0.74 0.57 73
0.75 0.57 74
0.77 0.57 75
0.79 0.58 76
0.81 0.58 77
0.83 0.59 78
0.86 0.59 79
0.89 0.60 80
0.93 0.60 81
0.98 0.61 82
1.03 0.61 83
1.09 0.62 84
1.16 0.63 85
1.25 0.64 86
1.34 0.65 87
1.45 0.66 88
1.59 0.67 89
1.75 0.69 90
1.92 0.71 91
2.16 0.73 92
2.47 0.76 93
2.88 0.79 94
3.49 0.83 95
4.59 0.89 96
10.00 0.99 97
10.33 1.20 98
100.18 97.06 99
608402.00 100.41 100
- med
--
_________________________________________________________________
Mark E Drummond Royal Military College of Canada
[EMAIL PROTECTED] Computing Services
Linux Uber Alles perl || die
