[hobbit] Hobbit versus Unicenter/TNG

[Henrik Stoerner]

On Thu, Feb 15, 2007 at 08:39:33AM -0500, Tom Georgoulias wrote:
> Anyway, one of the useless stats that is being thrown about by a third 
> party is the number of different data points they will monitor, and I 
> was asked to provide a similar number of my own.  I have about 9 custom 
> scripts mixed up on the server and client sides, each of which gather 
> anywhere from 4 to 40 data points per script.
> 
> Looking at the data under hobbitd, is there an easy way to determine how 
> much data (in the form of individual data points going into an RRD file) 
> monitored coming in from the clients? 

The "easiest" way is to look at the size of the RRD files. An RRD file
has a fixed size, and a 2-dataset RRD file is double the size of a
1-dataset RRD file. So if you add up the size of all the RRD files and
divide by the size of a 1-dataset RRD file (like the la.rrd files), you
have the number of datasets you're tracking for graphs.

> I found the network tests data under the bbtest column, but figuring out 
> what really matters under the hobbitd column is a bit more challenging.  
> Most of my client data comes in the "status" channel, but some is also 
> going into the data channel.

The only really interesting statistic in the "hobbitd" column is the
"Incoming messages/sec" value: This tells you how many test results are 
being processed by Hobbit every second, so to your boss this would be your
"sustained transactions/second" (TPS) number.

Another useless number to throw around is found on the "bbgen" column
statistics: The "Status messages" number there are the total number of
individual status "dots" on your web display: A "disk" status, a
"memory" status, an "http" status and so on.


Regards,
Henrik