[smokeping-users] decreasing ping interval and improving performance

[Tobias Oetiker]

Hi Jeff,

Yesterday Jeff Williams wrote:

> Hey folks,
>
> Our network group is looking to get some more performance from
> smokeping. I'm looking for some tips on how to optimize
> performance and I welcome any suggestions. Currently, we have
> about 6400 ip addresses that are pinged 30 times across a 5
> minute interval. When I reconfigured smokeping last time, I set
> up 30 FPing probes and distributed the ip addresses across those
> 30. Right now, there is no offset on those FPing probes, but I
> have experimented with that in the past.

If you do not define an offset, they should spread themselfs out
over the interval automatically.

> We are consistently
> adding branches, which means more ip addresses to ping, so I have
> a script that runs every morning to retrieve the latest list of
> ip addresses from a database and write the config for smokeping.
> Many of our branches are on DSL or cable with some dialup, but
> many are moving to T1. My feeling is that some smokeping
> performance issues may be solved simply by having lower latency
> links. Our network group would like to be able to do 5 pings
> every 30 seconds for each of the 6400+ ip addresses. As part of
> the process of improving performance, we are going to move
> smokeping from a virtual machine to a physical machine. However,
> we are currently running it in Xen paravirtualized, so
>

I have not yet done setups of this scale, but having multiple
fpings running in paralell is certainly a goood thing ... I would
not think that you are disk bound with only 6400 targets ...
especially not if you have plenty of memory ... Also moving off xen
to real HW should help ...

If you use rrdtool 1.3 you may be able to reduce your cache
consumtion and improve performance ...

> Are any folks on this list doing a similar number of pings in a
> fairly short amount of time? Is there a better way to spread out
> the ping load other than creating more FPing probes? We seem to
> be limited by I/O and memory and those probes currently use about
> 3.5 GB of memory last time I checked. Another thought we had was
> to use SAN disk instead of internal disk to decrease latency and
> improve throughput.

that does sound odd ... your smokeping process uses
3.5 GB or RAM ???

> Also, I just wanted to say thanks to all of the developers and
> testers of smokeping for a great utility. We have been using it
> for over 2 years now and have been pleased with its reliability
> and performance. We looked at comparable commercial solutions
> that could match the performance that we get from smokeping and
> were shocked at their sticker price.

Glad to hear smokeping makes you happy. If you want to make sure it
continues to evolve you could always become a sponsor and ensure
its continued development ...

cheers
tobi

-- 
Tobi Oetiker, OETIKER+PARTNER AG, Aarweg 15 CH-4600 Olten
http://it.oetiker.ch tobi at oetiker.ch ++41 62 213 9902