What is the cpu usage of l7 filter on various rb ?
What about x86 machines ?
Anybody done any real test with real traffic, managing multiple strings ?

( I cant't put in production, with traffic initiated by others than me right now, that is why am I asking.)
[[edit]
Does the l7filter cpu usage graphs coincide with the mt cpu usage, as a general rule ?

Usage are 50%-80% CPU, but memory are 30% after few hours 40% later 60% and finally 100% (And RouterBoard 532A reboot).
10 Rules Mangle with Layer7
My cfg:
2 ubiquiti SR2 400mW
1 RB532A (64RAM, 128F)
40Clients in Ubiquiti 1
30Clients in Ubiquiti 2

Maximum clients online 40.

bear in mind that L7 will look for the specified patterns in ALL connections that hit those rules. you should make sure only unknown traffic reaches those rules, and then limit the number of L7 rules as much as possible - this is regex we are talking about, it takes a lot of resources to scan everything ...

bear in mind that L7 will look for the specified patterns in ALL connections that hit those rules. you should make sure only unknown traffic reaches those rules, and then limit the number of L7 rules as much as possible - this is regex we are talking about, it takes a lot of resources to scan everything ...

Returning to the question: Is the cpu usage of each pattern similar to those of the l7filter ?
Asking cause it's nice to know, from someone from inside.
One could try to avoid using high cpu usage filters. And it could be nice to know what are you doing before actually doing it, it's kind of risky to put in traffic a job that you can't thoroughly test by your own, with one or two machines making the traffic.

Tx in advance.

if you are talking about the rule complexity indicated on the sourceforge site next to each protocol - yes, of course routeros is affected in a similar way and you should use less complex rules, less rules, and make sure less traffic is scanned by the firewall.

Yes, that is exactly what I was talking about.
Thank you for confirming that.