Bassem, Thanks for writing into the mailing list (ML) for help. While I mostly lurk on this list, the notion that Alan isn't helpful is wildly wrong. Just because his answers are succinct does not mean they are rude. And to be honest, this is *your* system and *your* design isn't working. The quality of your questions to the ML and the quality of the answers you provide to the ML will be directly proportional to the type/amount/quality of help you receive. For instance, "top" posting to his reply while he (and the rest of the FR-users ML) uses "inline". This is slightly disrespectful to those whose help you are requesting. I'll continue replying inline... On Thu, Jan 18, 2018 at 10:34 AM, Bassem Mettichi <mettichi@gmail.com> wrote:
Hello Alan,
i don't understand why your answers are always like this, if you don't like to give a hand you don't need to write me any thing you didn't answered any question i have posted on the this list always you blame me i don't like why, you have to trusted me, iam using Redhat 7.1, on this machine i have 4VCPU + 8GB RAM, iam using freeradius with rediswho for accounting and freeradius with rlm_perl for authentication,
Alan already stated that perl is slow. If you are going to insist on using perl, then you need to do some analysis on if you can be speed up what perl is doing. There was already advice on this: "b) make it do less work per-packet"
iam doing a performnce tests because the Customer has a lot of trafic that is why i have stopped using MYSQL and iam using now redis database, the Customer need 10000 accounting start / seconde but i have now reached only 6000 accounting start / second because of the high cpu
Is the CPU usage due to perl above? Is it starving the CPU for redis?
use of radiusd process on the system when i execute top
top isn't going to be the best way to measure where the CPU is spending its time.
command i see only radiusd running and used the totality of the 4 VCPU please if you can't help me don't tell any thing, i don't need for your blame.
For the sake of my own curiosity, does anyone know the throughput for FR dumping accounting packets to disk? Respectfully, -m