Hobbit client and large networks

[Alan Sparks]

Have used Hobbit with BB clients at a previous gig, but now looking to 
set one up with Hobbit clients at a new gig.  This will be an 
installation of close to 1000 nodes.  Having started testing using the 
clients, have some concerns and questions:

1) I am concerned that the server performs all the proc and disk 
tests... I'm concerned that several hundred servers will overwhelm the 
Hobbit server.  I see the client can be configured to run the checks 
locally, provided the configuration is copied to the clients.  Has 
anyone any practical experience with this many clients, and whether 
local checks are better in this case?

2) Most of the servers are Web and database servers... meaning that 
port, proc, and netstat reports are very large.  I've seen so far that 
the Hobbit client data sent to the server is almost 500k or larger, 
causing truncation problems and problems with the server-side checks 
(proc tests did not work until I used either client-local tests, or 
disabled the ports client data).  Can client data be disabled without 
affecting the ability to run checks?  Is there a way to streamline the 
data so that effective monitoring is possible, without sending 1/2 MB 
data for each client every 5 minutes?

Thanks in advance for any advice.
-Alan

-- 
Alan Sparks, UNIX/Linux Systems Integration and Administration
<asparks at doublesparks.net>