Tips & Tricks for Checking Health of Network...

[Shahbaz_1]

Hi Friends,
We are working on Oracle Applications, here the availability and performance is very crucial.
Most of the time the network becomes the performance bottleneck.

How to check the health of the network?
and How to know that the problem with Unix-Server or Network?

Please give me your tips ..


Thanks in Adv

Regards
Shah

[Bill McNAMARA_1 · Accepted solution]

netstat (-an)
netstat -i
lanadmin (comprehensive lan diag tool)

nettl -status
netfmt -f (Nettl log files)

Check syslogs.log for lan errors.


Later,
Bill

[Leif Halvarsson_2]

Hi
To just to check network performance you can use ftp , get and put files from the server to the clients. To monitor your network you need a more advanced tool as HP OpenView.

[U.SivaKumar_2]

Hi,
Monitor your hosts , routers with ping command.
Find the average response time evry 15 min.
Keep Record of your routers interface utilisation, cpu load etc.
Follow a proper IP addressing scheme to avoid
duplicate IP conflicts in mission critical
environments.
Use packet sniffer to analyse network traffic
utilisation protocol wise.

regards,
U.SivaKumar

[Steven Sim Kok Leong]

Hi,

If you want to check Oracle database response, use tnsping that is provided by Oracle from the client (that is encountering slow Oracle responses to queries) to the Oracle database server:

# tnsping

A highly accurate tool to measure network throughput and performance is the iperf tool downloadable at:

http://dast.nlanr.net/Projects/Iperf/

Note that while Oracle's tnsping measures response at the application level (OSI level 7), iperf measures till the socket level (OSI level 4).

Hope this helps. Regards.

Steven Sim Kok Leong

[sven verhaegen]

some very simple test involve atcp and udp type of communication , appart fom al the logfiles and netstat commands best is jsut to do a FTP test , take your lancard speed e.g. 100mb , take some reserve on the speed as no card really hits 100mb , usually it goes 50 to 70% of that speed , take a big fiel and :

from your own machine ftp it to and from a directory , time it

go on the local network to another server and do the same , time it again

go to a remote location over a router or a firewall/proxy and do the same test and again time

look for any strange behaviour or the occurence of the problem , the closewr the time issue or communivcation isue is the mor elikely it is on your system , if it happen only remotely it is an interlaying device , if it also happens on the system itself , try copy the file manuall without ftp , then it could be a system and not a network issue if you see also see speed/transfer issues

the same can be done with udp like protocols just in case it is only a udp related problem , this is how network problem are normally handled and investigated by responce centers , just brake it up in steps and go step by step untill you identify the culprit

[steven Burgess_2]

Hi Shah

Have attached a document regarding performing nettl tracing

The functions created trace files which can be used for troubleshooting events on the network

HTH

Steve

[steven Burgess_2]

Hi

And of course , you have linkloop for checking physical layer connectivity

linkloop -t 30 -i 4 -s 30 0x08000984bb16

man linkloop for more details

HTH

Steve

[steven Burgess_2]

Hi

I should have put these in the one reply.

A lot of performance issues can be down to name resolution.

Think about how you have your /etc/nsswitch is setup. It is always recommended to resolve locally before consulting dns

entry in /etc/nsswitch would look something like

hosts: files [NOTFOUND=continue] dns

A quick test for resolve response times is to use nslookup

HTH

Steve

[Sebastian Galeski_1]

Hi
go to: http://www.netperf.org/
there You can find a lot information.
Use netperf to look how your network work

regards Seba