Remote poller + boost

[adankert]

I have a remote poller with approx 35ms latency to the database/main instance.
I use a 1 minute poll interval.

When boost is active the remote poller is not able to finish it's poll cycle in a minute and graphing etc. fails as a consequence. This happens even if the remote poller only polls one single device!

If I disable Boost everything works fine. I can poll ~100 devices in around 30 seconds.

I have verified the database connection that works fine and both cacti and spine seems happy with the connection. Using tcpdump + wireshark I can see that queries are run OK, no network hickups detected.

Is there something that needs to be tuned to the ~30ms latency from remote poller to database server? Any config options I may have missed?
Running without Boost will not scale to the full network I need to poll, so that is not a long term option

(I had hoped to place a remote poller around 140ms away from main instance, this is just a first test).

Any help/hints/info is grately apreciated!

[Osiris]

Cacti 1.2 will fundamentally change the remote data collector architecture allowing much increased performance.

[adankert]

Cacti 1.2 will fundamentally change the remote data collector architecture allowing much increased performance.

Sounds good, fingers crossed....