That is why I said it is tricky to talk about NetView limitations. Really
you can identify bottlenecks that your particular configuration is having.
A person with experience can suggest resolutions to your bottleneck.
You have unmanaged all but one interface per system. Where does this help?
1. It helps the traffic on the interface card
2 It reduces the CPU load of the netmon process (polling)
Where does this not help?
1. Unsolicited traps -- trapd process
2. Memory used by ovwdb -- managed or not, they take the same memory in
database
3. Memory taken by the open maps
4. netmon configuration checking, still reads data for managed and
unmanaged interfaces.
5. snmpcollections (if you have them configured)
6 ....etc
You can have bottlenecks on a NetView server that also always has
solutions. The solution may be tuning, configuration or hardware.
1. Interface card traffic. (tune your AIX settings)
2. Can netmon keep up with all polling. (can be CPU bound, queue sizes,
adjusting timeouts, adjusting retries)
3. trap storms kill your processors (put in MLM, configure as trap
filter)
4. ovwdb slow (verify cache size is 120%)
5. cpu bound (put a local DNS on NV server, load DNS with EVERY interface
in your network)
6. cpu bound (everything else done, move from a 1-way to a 2-way or 4-way
hardware box)
7. In AIX, database cache writes affect operators, pause every minute
(strip database filesystem over multiple drives)
8. .... this list could go on for a while.
Kind regards,
Stephen Hochstetler shochste@us.ibm.com
International Technical Support Organization - Austin
Office - 512-436-8564 FAX - 512-436-9326
ITSO redbooks at http://www.redbooks.ibm.com
|