Latency differences between ACT 5.2 and 6.4 versions
Opened this issue · 1 comments
I am running ACT benchmark on an NVMe SSD, and seeing odd results between different versions of ACT. Specifically, after act_prep has run, and the drive has reached a steady state, running 40x scaling load via ACT version 5.2 vs. version 6.4 produce very different latency results, running each on the same system and disk. Here is a snippet of the latency results between the two versions. I am using the default config for the specific version in each run, changing only the target read-reqs-per-sec and write-reqs-per-sec to match the target load.
6.4 ACT Results @40x
reads
%>(ms)
slice 1 2 4 8 16 32 64
----- ------ ------ ------ ------ ------ ------ ------
1 70.96 28.44 3.76 0.05 0.00 0.00 0.00
2 71.34 28.79 3.93 0.06 0.00 0.00 0.00
3 71.18 28.79 3.87 0.05 0.00 0.00 0.00
4 71.13 28.71 3.84 0.05 0.00 0.00 0.00
5 71.25 28.90 3.99 0.08 0.00 0.00 0.00
6 71.16 28.69 3.88 0.08 0.01 0.00 0.00
7 71.17 28.63 3.76 0.05 0.00 0.00 0.00
8 71.27 28.79 3.88 0.05 0.00 0.00 0.00
9 71.26 28.94 4.01 0.06 0.00 0.00 0.00
10 71.19 28.80 3.94 0.06 0.00 0.00 0.00
----- ------ ------ ------ ------ ------ ------ ------
avg 71.19 28.75 3.89 0.06 0.00 0.00 0.00
max 71.34 28.94 4.01 0.08 0.01 0.00 0.00
5.2 ACT Results @40x
reads device-reads
%>(ms) %>(ms)
slice 1 2 4 8 16 32 64 1 2 4 8 16 32 64
----- ------ ------ ------ ------ ------ ------ ------ ------ ------ ------ ------ ------ ------ ------
1 0.33 0.02 0.01 0.00 0.00 0.00 0.00 0.07 0.01 0.01 0.00 0.00 0.00 0.00
2 0.33 0.02 0.01 0.00 0.00 0.00 0.00 0.08 0.02 0.01 0.00 0.00 0.00 0.00
3 0.26 0.00 0.00 0.00 0.00 0.00 0.00 0.04 0.00 0.00 0.00 0.00 0.00 0.00
4 0.28 0.02 0.01 0.00 0.00 0.00 0.00 0.06 0.02 0.01 0.00 0.00 0.00 0.00
5 0.27 0.01 0.01 0.00 0.00 0.00 0.00 0.05 0.01 0.01 0.00 0.00 0.00 0.00
6 0.29 0.02 0.01 0.00 0.00 0.00 0.00 0.07 0.02 0.01 0.00 0.00 0.00 0.00
7 0.28 0.00 0.00 0.00 0.00 0.00 0.00 0.05 0.00 0.00 0.00 0.00 0.00 0.00
8 0.29 0.01 0.01 0.00 0.00 0.00 0.00 0.07 0.01 0.00 0.00 0.00 0.00 0.00
9 0.30 0.02 0.01 0.00 0.00 0.00 0.00 0.07 0.02 0.01 0.00 0.00 0.00 0.00
10 0.27 0.01 0.00 0.00 0.00 0.00 0.00 0.05 0.01 0.00 0.00 0.00 0.00 0.00
----- ------ ------ ------ ------ ------ ------ ------ ------ ------ ------ ------ ------ ------ ------
avg 0.29 0.01 0.01 0.00 0.00 0.00 0.00 0.06 0.01 0.01 0.00 0.00 0.00 0.00
max 0.33 0.02 0.01 0.00 0.00 0.00 0.00 0.08 0.02 0.01 0.00 0.00 0.00 0.00
I have attached the full results output for both the 5.2 and 6.4 run. I am looking for some pointers on what might be the cause for the significant difference in latency being reported between the two test runs on the versions noted.
ACT-5.2-40x-results.txt
ACT-6.4-40x-results.txt
ACT 5.2 used 4 * <number of cores> threads for the transaction queues by default.. The equivalent tin 6.4 is the service-threads which now uses 5* <number of cores> by default. Could you try manually tuning the number of threads down in 6.4?