QUESTION Is dahdi_test still a valid measurement of stablility?

[masterdk]

So under a virtualized environment(HyperV or VirtualBox), dahdi_test usually yield an average of 99.7xxx

So i went ahead and installed the same PIAF green on a physical machine and ran the dahdi_test. At the console it resulted an average of 99.718660%, this was a shocker to me as its a pretty beefy intel xeon e5-2620 with 4GB ram. Then i SSH into the system, and ran the same test again, it netted an average of 99.978612%.

I'm able to reproduce the difference between console vs ssh easily. These are "clean" machines without anything configured or installed. No T1 cards or SIP setup yet. Is there a reason why the numbers are much lower when doing it on the console?


On a side note, my super old PIAF 1.7.5.5 (with a Dual T1 PRI PCI-e Card) running on a Core2Duo nets an average of 99.996706% average... What reason can it be that my new servers have lower score then my old trust system?


So my bottom line question, does the dadhi_test mean anything anymore? Back when I first setup the system, dadhi_test needed to be 99.95% or better, otherwise i might have drop calls, or random issues.

 

[phoneguy]

The reason is because your old server has a DAHDI card and your others do not so dahdi timing off of dahdi dummy which is what is used when you have no DAHDI card will always be less. Yes those scores will give you problems.

 

[krzykat]

I get Best: 99.999% -- Worst: 98.892% -- Average: 99.774594%
from a server with rentPBX and haven't realized any recognizable problems.
My Prox server: Best: 100.000% -- Worst: 99.602% -- Average: 99.947297%

 

[phoneguy]

Well any heavy hitting of the server especially MoH and anything that plays audio or conferences calls together will have issues at those times that drop below 99.97. Its been proven over and over again.

 

[masterdk]

Is there an explanation why the results are lower on my console vs ssh with the dahdi dummy?

So even if i went with all SIP trunks, with a bad timing, it will still affect MoH & Conference calls ?(If soo damn it all lol)...

What is the definition of "heavy hitting"? Is it when we have 15+ concurrent users on conference calls? Or is the required magnitude in 50+ users?