nv-l

Re: [NV-L] Netview Recommendation Required!

2007-07-25 04:29:50
Subject: Re: [NV-L] Netview Recommendation Required!
From: "Ahsan Ali" <ahsanali AT gmail DOT com>
To: "Tivoli NetView Discussions" <nv-l AT lists.ca.ibm DOT com>
Date: Wed, 25 Jul 2007 13:24:51 +0500
Bill,

Thanks for this input - its really helpful.

Can Windows MLMs be used with a Linux Netview server or are there some
gotchas to watch out for if we do this?

-Ahsan

On 7/20/07, Evans, Bill <Bill.Evans AT hq.doe DOT gov> wrote:


The three variables which are most important to sizing NetView are the
number of operators, the number of interfaces and the number of traps per
second/minute.  The referenced Installation Guide GC32-1842-00 discusses
this in detail in appendix A.  The book is available online.  The
calculation of the number of interfaces is more important than the number of
nodes.  X-Windows consoles consume more resources than Web Consoles.

A Very Large Network is characterized by:

more than 50000 interfaces
10 to 20 web consoles and 3-4 X-windows operators
1-5 traps per second with short bursts up to ten per second.
There are also considerations of SNMP data collection to add into the
calculation along with external connections to TEC or other systems.

The recommendation for a Very Large Network on Linux is to use a FOUR
PROCESSOR  system with THREE GHZ processors and a FOUR to SIX GB memory.

MLMs, one for every 3000 interfaces, would be a possibility and could move
your main NetView back to the MEDIUM category (10000-25000 interfaces) where
they recommend "two to four processors".  That would assume that most of the
traps could be filtered out and the forwarded ones would be in the ten per
MINUTE range and polling would be offloaded to the MLMs.  If your network is
really "nation wide", MLMs should be placed at remote sites where practical.
 They should be "local" to concentrations of monitored devices.

Alternatives for reducing the need are to slow down your network polling to
less than five minute cycles and turning off the non-critical traps at the
source.  For example, Authentication traps should probably be logged and not
handled by traps to NetView unless there is a really critical security
requirement.  If you can't control traps at the source you can also
configure NetView to discard them at the first level daemon instead of
investing resources in logging and processing them.

Good luck.  Count those interfaces and verify the number of traps expected
then figure your system size.

Bill Evans.


 ________________________________
 From: nv-l-bounces AT lists.ca.ibm DOT com [mailto:nv-l-bounces AT 
lists.ca.ibm DOT com]
On Behalf Of Usman Taokeer
Sent: Friday, July 20, 2007 7:39 AM
To: Tivoli NetView Discussions
Subject: Re: [NV-L] Netview Recommendation Required!



Thanks for your input Ahsan

So the combination would be a Linux box with NV 7.1.5? Would a dual 3.0Ghz
proc machine with 2GB memory be good enough to handle this?

Regards,
Usman



On 7/20/07, Ahsan Ali <ahsanali AT gmail DOT com> wrote:
> Take a look at Appendix A in the Linux / Unix Netview Installation guide.
>
> A trap rate of 5/sec falls in the category of Very Large Network. With
> an MLM doing some trap supression/throttling your network might be
> manageable with the sizing for a Medium sized network.
>
> >From Appendix A:
> ===============================================
> A two or four-processor system with 2 GB of memory, one or two disk
> drives, 100 MB Ethernet connection, and an appropriate video card.
>
> Example systems:
>
>    * For AIX(R) environments: IBM(R) pSeries(R) 7028 or pSeries 550
> with four processors at 1.45 - 1.6 GHz.
>    * For Windows(R) and Linux(R) environments: IBM xSeries(R) 365
> with four processors at 3.0 GHz or xSeries 366 with four processors at
> 3.6 GHz.
> ===============================================
>
> As for Linux/Unix vs Windows, it is my opinion that Linux and Unix are
> the only way to go if you want to have any sort of life outside of the
> Network Operations Center.
>
> For the experts on this list, FP01 for 7.1.5 is out so are the early
> teething problems sorted out?
>
> -Ahsan
>
> On 7/19/07, Usman Taokeer < usman.taokeer AT gmail DOT com> wrote:
> > Hi List,
> >
> > We have a 2000 node network, spanning nation-wide, media is a mix of
ISDN
> > BRI, Lease Cuircits & VSAT. Approx events/traps per sec would not be
more
> > than 5 per second. Which platform for Netview would you recommend i.e
Unix
> > or Windows and which version 7.1.4 or 7.1.5.
> >
> > Is there a need of using MLM's in this scenario or can NetView server
handle
> > these traps.
> >
> >
> > Regards,
> > Usman
> >
> >
> > _______________________________________________
> > NV-L mailing list
> > NV-L AT lists.ca.ibm DOT com
> > Unsubscribe:NV-L-leave AT lists.ca.ibm DOT com
> > http://lists.ca.ibm.com/mailman/listinfo/nv-l (Browser
> > access limited to internal IBM'ers only)
> >
> >
>
>
> --
> ---
> Due to a shortage of devoted followers, the production of great
> leaders has been discontinued.
> _______________________________________________
> NV-L mailing list
> NV-L AT lists.ca.ibm DOT com
> Unsubscribe:NV-L-leave AT lists.ca.ibm DOT com
> http://lists.ca.ibm.com/mailman/listinfo/nv-l (Browser
access limited to internal IBM'ers only)
>


_______________________________________________
NV-L mailing list
NV-L AT lists.ca.ibm DOT com
Unsubscribe:NV-L-leave AT lists.ca.ibm DOT com
http://lists.ca.ibm.com/mailman/listinfo/nv-l (Browser
access limited to internal IBM'ers only)




--
---
Due to a shortage of devoted followers, the production of great
leaders has been discontinued.
_______________________________________________
NV-L mailing list
NV-L AT lists.ca.ibm DOT com
Unsubscribe:NV-L-leave AT lists.ca.ibm DOT com
http://lists.ca.ibm.com/mailman/listinfo/nv-l (Browser access limited to 
internal IBM'ers only)