[Orca-users] Installing Orca 0.27 and Procollator on Linux

[Jarle Bjørgeengen]

On Mon, 2006-06-19 at 10:23 +1000, Ian Holsman wrote:
> On 19/06/2006, at 8:18 AM, Kai 'wusel' Siering wrote:
> 
> > csarid wrote:
> >
> >> Another question.. is the idea to run procollator on linux to  
> >> collect the data and run all the reporting from a solaris system  
> >> running orcallator?
> >
> > I wouldn't do this unless Sun comes up with competitive CPUs again ...
> > I'm running the Orca backend for some hundred Solaris- and Linux-
> > machines on a Dual Xeon (HT enabled); beware that you'll need vast
> > amounts of I/O performance as well as decent computing powers.
> 
> I'm doing the same thing as you, and am frankly frustrated at the amount
> of I/O required by RRD.
> 
> have you got any specific tuning tips on how to set up the backend  
> machine(s)?
> and what you have found works?
> 
> personally I think a RRD-type system could be designed which isn't as  
> I/O intensive, but I haven't got the time
> to invest in creating it.
> 
> 
You can help things a little bit by putting the rrd files on a ramdisk ,
and rsync-the rrd-files to disk after updating them. Then you avoid
having the tremondous amont of small i/o 's directly to disk, and let
rsync do that in a more efficient way, with larger and fewer chunks of
data. 



-- 
Jarle Bjørgeengen <jarle.bjorgeengen at usit.uio.no>
Gruppe for basis systemdrift (BSD), SAPP, USIT
Tel. +47 22852930 (office)