#246 Bounceback responsiveness needs better outputs

[Robert McMahon]

~~~
Hi Sriram,
Thanks for pointing this out. There are two things here:
The RPS calculation is using the average (3.99ms)
The PDF outputs are in unit time and not RPS
We probably should change these.
Should the RPS output be based on max latency (or worst RPS) and the histogram output units in RPS?
Thanks,
Bob
On Fri, Mar 17, 2023 at 12:07 AM Sriram Neelakandan wrote:
Thanks Bob, Nice Video!
A lot of press & education around "RPM under Working Conditions" is needed.
===
BTW .. a single RPM number as Stuart says, should reflect the 99.9% or say 99.7% RPM number.
I started looking at a UTF report that Battu is helping to add to include Bounce back reporting...
12:06:06.551 LOG blr30tst1 [ ID] Interval Transfer Bandwidth BB cnt=avg/min/max/stdev Rtry Cwnd/RTT RPS
12:06:15.605 LOG blr30tst1 [ 1] 0.00-10.03 sec 0.019 MBytes 0.016 Mbits/sec 100=3.991/2.811/4.692/0.224 ms 0 14K/3809 us 251 rps
12:06:15.606 LOG blr30tst1 [ 1] 0.00-10.03 sec OWD Delays (ms) Cnt=100 To=1.024/0.759/1.727/0.170 From=2.963/1.562/3.530/0.279 Asymmetry=1.939/0.314/2.516/0.403 251 rps
12:06:15.606 LOG blr30tst1 [ 1] 0.00-10.03 sec BB8(f)-PDF: bin(w=100us):cnt(100)=29:1,32:1,37:1,38:4,39:2,40:52,41:33,42:1,47:5 (5.00/95.00/99.7%=38/47/47,Outliers=5,obl/obu=0/0)
Now the complete 10 sec run report has MAX at 4.692ms.. and an average 3.99ms => 251 RPS.
I expected my MAX- say 99.7% RPS / Lowest RPS = 1000/4.7 = 213 RPS.
Why is the 99.7% at 47 RPS ? or even the 5% at 38 RPS ?
Am I missing something ? or Is there something incorrect in the BB8(f)-PDF computation ?
thanks
Sriram~~~

[Robert McMahon]

fixed in master as of March 19, 2023