Hello All.
Sorry for my poor English. The subject should be "Strange Network
Performance on AIX", not "Storagne".
I realized that I don't have this problem when I use shorter
'collect_every' for bytes_in and bytes_out for 1GbE adapter.
When I used "collect_every" and "time_threshold" as follows,
============================
collection_group {
collect_every = 10
time_threshold = 120
metric {
name = "bytes_out"
value_threshold = 4096
}
.....
============================
Ganglia showed network performance which seems reasonable for 1GbE, which
is around 120MB/s.
However, when I measured performance of two trunked 10GbE interfaces while
I executed iperf, Ganglia again cannot measure correct performance, which
is, accoding to iperf, 13Gbit/s (1.6GB/s).
I have tried both of
collect_every = 10
time_threshold = 120
and
collect_every = 1
time_threshold = 12,
but in both case, Ganglia did not seem to show correct measurement.
When I execute iperf between a server with 1GbE interface and another
server with two trunked 10GbE interfaces, Ganglia showed correct
measurement (around 120MB/s) for both of servers
Is there any counter overflow issue?
Yasushi Fujimoto
System Server Services#3, Systems Services#2
Infrastructure Management Services Delivery,
Integrated Technology Service Delivery,
Global Technology Services, IBM Japan, Ltd.
Yasushi
Fujimoto/Japan/IB
m...@ibmjp To
[email protected]
2009/08/24 15:36 et
cc
Subject
[Ganglia-general] Storange Network
Performance on AIX
Hello all.
I am a system engineer working for Japan IBM and quite a newbie for
Ganglia.
I have installed RPM image (ganglia-gmond-3.0.7-1.aix5.1.ppc.rpm ) from Dr
M. Perzl's site (http://www.perzl.org/ganglia) on
IBM p5 p510 running with AIX 5.3 (5300-08-04)
However I found the report of bytes_in and bytes_out from Ganglia does not
match with the reality.
The followings are showing result from iperf, Ganglia and NMON for a case
where I executed iperf client for 300 seconds on a different machine.
iperf and NMON show the similar results which are around 120MB/s, however,
the result from Ganglia shows much slower performance, which is around
14MB/s.
Could anybody suggest me what I am missing or how I can solve the problem?
[Iperf Result]
bash-2.05b$ date; time /usr/bin/iperf -c XXXXX -t 300 -p 2000 ;date
Fri Aug 21 17:55:00 JST 2009
------------------------------------------------------------
Client connecting to hpsscws1, TCP port 2000
TCP window size: 16.0 KByte (default)
------------------------------------------------------------
[ 3] local XXX.XXX.XXX.XXX port 59968 connected with XXX.XXX.XXX.XXX port
2000
[ ID] Interval Transfer Bandwidth
[ 3] 0.0-300.0 sec 33.1 GBytes 948 Mbits/sec
real 5m0.855s
user 0m1.960s
sys 0m30.160s
Fri Aug 21 18:00:01 JST 2009
[Ganglia Result]
(See attached file: ganglia-network-hpcws1-090821-1800.PNG)
[Nmon Result]
(See attached file: nmon-result-090821-test.txt)
With my best regards.
Yasushi Fujimoto
System Server Services#3, Systems Services#2
Infrastructure Management Services Delivery,
Integrated Technology Service Delivery,
Global Technology Services, IBM Japan, Ltd.
[添付ファイル "ganglia-network-hpcws1-090821-1800.PNG" は Yasushi
Fujimoto/Japan/IBM が削除しました] [添付ファイル
"nmon-result-090821-test.txt" は Yasushi Fujimoto/Japan/IBM が削除しました]
------------------------------------------------------------------------------
Let Crystal Reports handle the reporting - Free Crystal Reports 2008 30-Day
trial. Simplify your report design, integration and deployment - and focus
on
what you do best, core application coding. Discover what's new with
Crystal Reports now. http://p.sf.net/sfu/bobj-july
_______________________________________________
Ganglia-general mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/ganglia-general
------------------------------------------------------------------------------
Let Crystal Reports handle the reporting - Free Crystal Reports 2008 30-Day
trial. Simplify your report design, integration and deployment - and focus on
what you do best, core application coding. Discover what's new with
Crystal Reports now. http://p.sf.net/sfu/bobj-july
_______________________________________________
Ganglia-general mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/ganglia-general