I've managed to break the back of this, but still have some questions.
We require aggregate graphs with total 95 percentiles for some customers of ours, so I've been testing on an aggregate graph today.
1 minute graphing (works great) Poller interval still default etc.
I have addedinbound traffic
HRULE: |95:bits:0:aggregate_current:2| Mbits inOutbound traffic
HRULE: -|95:bits:0:aggregate_current:2| Mbits in
Told it to calculate similar data sources, and all is well, I get my lines. They appear to be accurate from what I can tell.
However, when I try and view the day/week/month/year views, initially only the day and week would draw, with the other graphs timing out. Our Linux chaps helped me look at this and have altered some of the PHP settings.
We have changed
memory_limit from 128 > 512
max_execution_time from 60 > 180
And the graphs now draw! But they take an absolute age to draw. I have read that the performance of aggregate 95th is quite high performance from the cacti documentations and the caveat about having to tweak PHP memory settings. At least the graphs do draw now, but my questions are threefold
[*]Have I added 95th the correct way? is aggregate_current the correct definition, or should I be using something else? Is this the "proper way" to do this?
[*]If it is correct, is it just going to be this slow
[*]Are there other tweaks to speed this up? Do I have any options.
The machine we're running Cacti on is a powerhouse, 16 core Xeons with 72Gig of ram, so we have the throughput to play with.
Any advice would be much appreciated.
Picture of graph output is below