pcp
[Top] [All Lists]

Re: [pcp] pmlogger performance

To: PCP <pcp@xxxxxxxxxxx>
Subject: Re: [pcp] pmlogger performance
From: Stan Cox <scox@xxxxxxxxxx>
Date: Mon, 22 Jul 2013 11:09:14 -0400
Delivered-to: pcp@xxxxxxxxxxx
In-reply-to: <51E82B31.6040100@xxxxxxxxxx>
References: <51E374C4.5@xxxxxxxxxx> <51E4A12C.5000003@xxxxxxxxxxxxxxxx> <51E4A7DE.1090307@xxxxxxxxxx> <51E4A93F.4070505@xxxxxxxxxxxxxxxx> <51E4ADD0.6060806@xxxxxxxxxx> <51E4FB50.7010100@xxxxxxxxxxxxxxxx> <51E82B31.6040100@xxxxxxxxxx>
User-agent: Mozilla/5.0 (X11; Linux x86_64; rv:17.0) Gecko/20130311 Thunderbird/17.0.4
I ran sysbench by itself (raw) and then simultaneously with each of: pmlogger, sar, iostat, mpstat, vmstat, atop. Only pmlogger and sar are archiving metrics; the others are just sending the output to /dev/null. The results are a bit unpredictable. The figures are based on total time as reported by sysbench.

interval/duration   raw    pmlogger   sar   iostat   mpstat   vmstat   atop
1.5 minute .5 hour                      pmgnuplot-100-18
        1383.69    1393.15 1410.57 1382.61  1311.05  1333.19  1317.22

1 minute 2 hours             (1)        pmgnuplot-60-120.sh
        4975.39    5140.82 4569.37 4787.69  4757.82  4630.95  4478.35

5 minutes 2 hours            (2)        pmgnuplot-300-24.sh
        4813.47    4829.86 4623.23 4652.61  4869.47  4618.41  4849.66

10 minutes 2 hours   (3)                pmgnuplot-600-12.sh
        4910.76    4762.92 4948.54 4798.02  4527.14  4744.05  4734.89

(1) 8.1% faster than raw
(2) 3.9% faster than raw
(3) 3% faster than raw

The attached pmgnuplot* scripts will gnuplot the results.



Attachment: pmgnuplot-100-18.sh
Description: application/shellscript

Attachment: pmgnuplot-60-120.sh
Description: application/shellscript

Attachment: pmgnuplot-300-24.sh
Description: application/shellscript

Attachment: pmgnuplot-600-12.sh
Description: application/shellscript

<Prev in Thread] Current Thread [Next in Thread>