Do I need my APE?

I inherited an environment that was first setup about 6-7 years ago. It had an APE added to it probably 6 years ago right when I started and my understanding was it was due to sluggish polling and behavior, but I never got into the conversation with the admin at the time. I'm now digging into doing a full on upgrade to our environment with migrations to new servers and I'm seriously questioning why I may need this APE. We have 7,768 elements and the documentation I'm seeing says APE for use with over 12,000 elements. What would the ramifications be to move everything over to the main poller and power down the APE before moving ahead with upgrades? The elements aren't even split evenly, I have almost 5800 on the main poller and it seems to operate just fine.

Parents
  • So the 12,000 benchmark partially depends on your polling interval, that number assumes default values of 2m ping and 10m stats.  Collecting stats at a 5m interval would be about to say each poller can only handle 6k elements.  If you go to the Settings > Polling Engines you can see the current utilization in % terms to give you a good idea of where you are at on that stuff.  

    Additionally, APE's allow you to offload polling work from the main polling engine, so sometimes people will do it for performance reasons, for example I have a relatively large environment with 40k elements across many pollers and I try to avoid having my MPE do much polling, i just let it do application and alerting work, my APE's do 99% of the polling, my additional web servers are in a load balanced pool that does not include the MPE so users rarely actually touch the primary app server.  It doesn't get bogged down, I don't have monitoring hiccups and everyone is happy.  In smaller environments you might not find that layout to be cost effective, but that might explain some reasons you may have ended up with the APE. 

    Dropping the APE and running the MPE by itself at 8k elements at default polling intervals is totally doable, just don't cheap out on resources or you may find yourself back in here next week complaining about a slow web interface.

Reply
  • So the 12,000 benchmark partially depends on your polling interval, that number assumes default values of 2m ping and 10m stats.  Collecting stats at a 5m interval would be about to say each poller can only handle 6k elements.  If you go to the Settings > Polling Engines you can see the current utilization in % terms to give you a good idea of where you are at on that stuff.  

    Additionally, APE's allow you to offload polling work from the main polling engine, so sometimes people will do it for performance reasons, for example I have a relatively large environment with 40k elements across many pollers and I try to avoid having my MPE do much polling, i just let it do application and alerting work, my APE's do 99% of the polling, my additional web servers are in a load balanced pool that does not include the MPE so users rarely actually touch the primary app server.  It doesn't get bogged down, I don't have monitoring hiccups and everyone is happy.  In smaller environments you might not find that layout to be cost effective, but that might explain some reasons you may have ended up with the APE. 

    Dropping the APE and running the MPE by itself at 8k elements at default polling intervals is totally doable, just don't cheap out on resources or you may find yourself back in here next week complaining about a slow web interface.

Children
No Data