high network utilisation

[amithp]

Hi the database server SD64B, HPUX 11.31

has very high network utilisation.

Please advise how to check the culprit?



Glance C.04.73.400 04:16:31 proudb07 ia64 Current Avg High
------------------------------------------------------------------------------------------------------------------------------
CPU Util SSU U | 9% 11% 36%
Disk Util F F | 14% 35% 100%
Mem Util S SU UF F | 82% 85% 96%
Networkil U UR R | 70% 74% 90%
------------------------------------------------------------------------------------------------------------------------------
PROCESS LIST Users= 12
User CPU % Thrd Disk Memory Block
Process Name PID Name (1600% max) Cnt IOrate RSS/VSS On
--------------------------------------------------------------------------------
rman 6619 oracle 97.1 1 0.0 264.7mb 267.7mb PRI
perl 1566 root 8.0 1 0.0 3.0mb 8.3mb died

[Steven E. Protter · Accepted solution]

Shalom,

Your network utilization is above average, but as long as it never hits 100% you will not have a problem.

This looks like a superdome.

You can increase band width by purchasing and installing APA, Auto Port aggregation. This permits you to combine two network ports as one and increase band width. This is network bonding for HP-UX.

Based on the data submitted, you have a lot of data being read by oracle across the network. Is the database hosted on NFS? Are there a lot of inefficient queries in the database? Consult your dba.

SEP

[Zinky]

Amith,

My guess is your on an rx**40 server and using Gigabit - right? And you're using GlancePlus/OVPA with out of the box settings.

CAn you hit "l" (lower case in glance and post please? It will show you you're "LAN Interface Stats".

OOTB (Out of the Box), there are certain metrics (DiskIO and Network are some) that are not really representative of what you are experiencing. Network is usually shown as 100% in Glance if any of your LAN interfaces show in or out packets exceeding 1000 packets/second even if byte throughput is in the KBytes/second.

[Zinky]

In my ecosystem -- my Network is practically 100% reported as utilized as we have a lot of inter-server transactions in the order of 5,000 packets per second in small packets (transactional). But our applications are happy - transactions are fine.

[rick jones]

I would suggest looking to see if you have packet losses on the transmit queue(s) of your nics - lanadmin -g mibstats (I think the command changed in 11iv3, but typing lanadmin will point you at the new command). If you aren't dropping packets on outbound, you aren't oversaturated.

[amithp]

thank you all of you