Question with network bottlenect

[Ming.Dynasty]

Hi all

I have a question with network bottlenect. from glace it show the network average usage over 80%. my question is

1) if that means i'm facing a network bottlenect?if yes. any good suggestion for this?

2) I'm using 1GB fiber card,it the 80% means, the rate is 1000/8 *0.8=100MB/sec ?

thanks man. points will be given for all reply.

[Yogeeraj_1]

hi

Symptoms of a bottleneck would be a sustained 100% utilisation of the network interface card.

I would suggest that you run a measureware report and plot a graph of the current utilisation over a range of days.

Example of the measureware report may look like as follows: ( /opt/perf/bin/extract -xp -r /home/yogeeraj/zgbltemplate -g -b today 00:00 -e today 23:59 )
Memory Cache Peak Pk FS Pg Out Swap Memory Network Netwk Netwk
Year Date Time CPU % % Swap % Hit % Disk % Sp % Rate Out Rt Queue Pkt Rt OutQ Queue
2006 27/10/2006 0:00 6.69 31.15 11 97.63 3.53 97.1 0 0 0 56 0 0
2006 27/10/2006 0:05 3.19 31.04 11 99.74 1.66 97.1 0 0 0 34 0 0
2006 27/10/2006 0:10 3.26 31.06 11 99.85 1.33 97.1 0 0 0 35 0 0
2006 27/10/2006 0:15 5.44 31.11 11 99.93 1.82 97.1 0 0 0 129 0 0
2006 27/10/2006 0:20 1.8 31.12 11 99.96 0.54 97.1 0 0 0 32 0 0
2006 27/10/2006 0:25 1.22 31.12 11 100 0.36 97.1 0 0 0 28 0 0
2006 27/10/2006 0:30 1.4 31.12 11 99.9 0.58 97.1 0 0 0 32 0 0
2006 27/10/2006 0:35 1.69 31.12 11 99.82 0.5 97.1 0 0 0 22 0 0
2006 27/10/2006 0:40 1.45 31.13 11 100 0.61 97.1 0 0 0 20 0 0
2006 27/10/2006 0:45 11.98 31.14 11 100 0.44 97.1 0 0 0 18 0 0
2006 27/10/2006 0:50 17.43 31.31 11 99.92 0.67 97.1 0 0 0 19 0 0
2006 27/10/2006 0:55 1.83 31.34 11 99.73 0.55 97.1 0 0 0 29 0 0
2006 27/10/2006 1:00 2.13 31.43 11 99.95 0.83 97.1 0 0 0 36 0 0
2006 27/10/2006 1:05 4.65 31.25 11 98.66 14.31 97.1 0 0 0 573 0 0
2006 27/10/2006 1:10 4.73 31.29 11 100 16.62 97.1 0 0 0 666 0 0
2006 27/10/2006 1:15 3.91 31.19 11 100 15.35 97.1 0 0 0 610 0 0
2006 27/10/2006 1:20 2.66 31.2 11 99.91 15.24 97.1 0 0 0 612 0 0
2006 27/10/2006 1:25 3.19 31.2 11 99.75 15.49 97.1 0 0 0 616 0 0
2006 27/10/2006 1:30 6.17 31.16 11 99.82 14.02 97.1 0 0 0 572 0 0
2006 27/10/2006 1:35 3.49 31.26 11 99.34 14.13 97.1 0 0 0 572 0 0
2006 27/10/2006 1:40 3.19 31.24 11 99.8 15.23 97.1 0 0 0 634 0 0

if you need more assitance, please let us know

kind regards
yogeeraj

[Ming.Dynasty]

Hi yogeeraj

what comman i can use to get network usage? I use sar -d to get disk busy% to plot a graph for disk, but what comman i can use for traffic?

[Yogeeraj_1]

hi,

to view the utilisation online, try:

glance -l

or

lanadmin

(see man lanadmin for more information)

otherwise, try the measureware option as specified in my previous post

kind regards
yogeeraj

[Ming.Dynasty]

Hi
I did not have this file in my /opt/perf/bin folder. and also ,I need a comand to get traffic% into a txt file, then I can use excel to plot a graph. lanadmin and glance -l can not work

[Yogeeraj_1]

hi,

Sorry, Measureware is another product that must be installed before you can actually use it.

# swlist -l product |grep -i measure
MeasureWare C.03.58.00 MeasureWare Software/UX

# which extract
/opt/perf/bin/extract
#


kind regards
yogeeraj

[Jaime Bolanos Rojas. · Accepted solution]

Ming,

If you have a network bottleneck glance may help you, go to the Network interface by screen, first indication of a bottleneck is a high number of collisions, you can check this out also with lanscan.

Also watch for time out requests, retries and packet errors or other indications that you are having a bottleneck in network.

The use of the netstat command also helps.

Network bottleneck sometimes are the most difficult ones to find becuase of the lack of tools that would let us know exactly the problem.

Regards,

Jaime.

[John Collier]

Being a bit of a network-type guy, I am fairly interested in the general layout of the network in reference to the interface in question.

It would help me greatly to know what is connected to this port, how it is connected, what kind of data is being pushed/pulled from this interface (general information, nothing to compromise a job or company).

To get some trend type information, you may consider putting in a cron job to pull the stats off of this interface every 2 - 5 min for the next 48 - 72 hours (or longer if you can) and have that all piped to a text file.

While 80% seems extreme, it should NOT be service effecting if it does not try to exceed that regularly.

It may simply be that the box is plugged into the network in a less-than-optimal position. The information requested could go a long way to helping us determine that.