[SunHELP] Clustering on Sun? with Netflow Collector?

[Foong, Tzeweng]

Hi there all sunhelp listers,

I am interested to find out some details regarding the state of clustering 
on sun servers. I an really preety ignorant interms of how clustering works.
In my mind clustering would allow a set of several computers to behave like 
one computer system. I know this sounds easier than it is. What I want to 
know is how well is clustering implemented on Sun Servers. I have found 
3 products that seem to do the same thing. They are :

* Sun Cluster v3.0
* Veritas Cluster Server v1.3.0
* UNIQFail

I am interested in experiences people have had with these products.
Especially
the Veritas software. In particular I am interested in the time it takes for
fail over 
to occur. I believe( but I may be wrong ) that UNIQFail can take upto 10
mins to 
detect failure and can take several munis more to fail-over to the backup
system.
I believe that SunCluster 3.0 is similar, am I wrong ?  what about veritas
cluster?

The application we are looking to run on this cluster is the Netflow
Collector software. 
It needs to be up ALL THE TIME and as such the above fail-over times are I
believe
unacceptable. Is it unavoidable? Has anyone tried clustering NetFlow
Collector software?

Other general Clustering questions I have are as follows.
* In the clustering environment does the IP of the interfaces look the same
no 
  matter which server is running/active? or do I need some kind of load
balancer/fail-over
  hardware to route/switch the UDP data to the right/active host? 

* Will that destroy the data integrity of a netflow data unit? ie will part
of a Unit be sent 
  to one Server and another part to the other server. Will Netflow Collector
be able to 
  handle this?

* Is any application (netFlow Collector) able to run tranparently in a
clustered 
  environment? How does this occur? What about the output datafiles where
are they 
  written. I believe that there are clustering agents that assist
applications to run in 
  a clustered environment. How do they fit in the scheme of things?

* Does the process only run on one sytem or are the process somehow load 
  balanced and can be run from any of the systems as required ? What about 
  the output data files? doesnt that mean it needs some kind of distributed 
   and redundant and mirrored file system. Does this beast actually exists
on 
   sun systems?

Well my brain has just about busted. Please help explain this to a
clustering 
newbie!

Thanks in advance

Tze Weng Foong