Way of getting "Response download time" value
Hello,
I wonder how the metric "Response download time" is generated. In the CAS server, the definition is "(...)obtained by subtracting Request Time, Server Time and Idle Time from Page Load Time".
Is it correct? Does it mean that "Response download time" is not measured but calculated by the AMD probe ?
If yes, it is possible to get wrong value?
Thank you for your answer.
Regards,
Antoine Buffotot
Antoine,
"Response download time" is actually calculated by the CAS, the AMD just provides the base metrics like Request Time. All measurements, including the most basic ones, are an effect of some sort of calculation. The value will always be correct, provided that the base metrics will be correct.
Thanks,
Wojtek.
Hi Wojtek,
Thank you for your answer, it is very helpful
Could you tell me how "Response download time" is calcukated, the formula?
The opening and the closure of the TCP session is part of the times measured?
Thank you for your answer
Antoine
The formula is as the definition states: PageLoadTime - RequestTime - ServerTime - Idle Time. The TCP session handshake will be included to hits which are the first in a session and the TCP closure will be included to hits last in a session. So yes, it is part of the measurements. However in case of a long pause between the TCP handshake and first hit, or last hit and session closure, the times will not be taken into account.
Wojtek.
Hi,
Thank you for your answer
I asked these questions because for one Customer, SYM automated checks are used and this transactions are monitored by DCRUM.
The transaction time given by SYM is around 10 ms but the operation time given by DCRUM is around 170 ms. After some tests, it appears that the real value is close to 10 ms. Moreover, 95% is for the metric "Response download time"
Why we have a too long "Response download time" in that case?
What configuration could be done to fix that?
Antoine
Update
After some tests, I have the answer of why a too long "Response download time"
The reason is there is the Acknoledgement TCP packet which is sent by the client after 200mS it received the server response. So that impacts the measures
So what configuration must be applied to fix that?
Thank you for your answer
Antoine,
In order to not include that late ACK packet, you would need to decrease the value of the 'Hit session timeout' in global HTTP configuration to less than 200ms (or whatever the delay is). However such change will dramatically affect measurements for all HTTP traffic and is not recommended.
In general, the measurements are correct the way they are. The fact that the client acknowledged the response after 200ms means, that full reception of the response did finish after those 200ms.
Wojtek.
