the cacti team wrote on the web page, that they wanted some benchmarks, so I did some during upgrade:
We are using cacti for montoring some san switches with a resolution of 10sec and some with 1 sec.
I did some tests and saw, that cactid 0.8.6b is roughly 3 times slower than the older cactid from 0.8.5a. That's the reason that we changed the intervall to 20sec and probably wan't upgrade the cacti with 1sec interval
Also it looks like the bottleneck of 0.8.5a is the I/O of the disks, with 0.8.6b it looks like CPU is the bottleneck.
Environment for both tests:
Hosts Polled: 32
RRA: 4000
OIDs: 7000
RTT ICMP: 0.5ms +-0.2ms
Concurrent Processes: 1
Processor: UltraSPARC-IIIi
Processor Speed: 1GHz
Number of Processors: 2
Memory: 2GByte Memory
Disks: 2x SCSI (0.5 of the rra's are on disk 1 and the rest on disk 2)
Database: mysql 4.0.16
First Test: cacti & cactid 0.8.5a,
Threads: 10
Mean time: 5.5+- 0.5 sec
Output from
Code: Select all
time time ./cactid:
user 0m2.160s
sys 0m2.550s
First Test: cacti & cactid 0.8.6b,
Threads: 20
Mean time: 16+- 4 sec
Output from
Code: Select all
time php poller.php
user 0m5.200s
sys 0m2.200s
Let me know, if you need some more information.
Roger