A quick glance at the code makes it look like all the xodtemplate
functions spend most of their time list walking. Converting from a list to
a tree/hash should dramatically speed things up.
On Tue, 8 Feb 2005, Jason Martin wrote:
> I'm looking ahead to running a 6K host / 60K service distributed
> Nagios setup reporting up to a central Nagios host.
>
> The test procedure was to write a config file consisting of X
> hosts with 10 services per host.
>
> On a 2 proc 2Ghz Intel machine, the preflight time for 1.1 and
> 2.0cvs comes out as:
>
> Hosts v1.1(s) v2.0(s)
> 100 0 1
> 400 5 4
> 500 7 6
> 1000 23 23
> 2000 149 117
> 3000 321 251
> 4000 526 431
> 5000 760 651
> 6000 1034 912
>
> The profile for the preflight is:
> Each sample counts as 0.01 seconds.
> % cumulative self self total
> time seconds seconds calls ms/call ms/call name
> 84.24 4.33 4.33 5502 0.79 0.79 xodtemplate_find_service
> 7.78 4.73 0.40 1 400.00 449.66 xodtemplate_sort_services
> 4.09 4.94 0.21 1 210.00 210.00 pre_flight_check
> 0.78 4.98 0.04 5501 0.01 0.01 find_service
>
> It looks like xodtemplate_find_service is still taking up a log of time. Can
> anything be done to speed it up? Otherwise it looks like 6K hosts is going to be unatainable.
>
> Thanks,
> -Jason Martin
> --
> This message is PGP/MIME signed.
>
This post was automatically imported from historical nagios-devel mailing list archives
Original poster: [email protected]