ABCDEFG
1
DateBuildTesterTest GoalsDateResults ObservedDiscussion
2
8/27/2019 16:55:08DB6Json
Find saturation level. Observing nodes overall stability
Aug 27 2019, 12.00 am UTC
Published 120k blocks at 230 BPS and saw 230 CPS.
Nodes confirming at 209ms median.
https://imgur.com/a/HAPu7rb
https://imgur.com/a/AT19b5r
3
8/27/2019 16:55:11DB6Json
Find saturation level. Observing nodes overall stability
Aug 27 2019, 8.10 pm UTC
Published 160k blocks at 230 BPS and saw 230 CPS.
Nodes confirming at 170ms median.
4
8/28/2019 8:31:05DB6Json
Find saturation level. Observing nodes overall stability
Aug 28 2019, 1pm UTC
Published 200k blocks at 363 BPS (peak above 500).
CPS peaked at 300 but network saturated around 200, some nodes at 100
Confirming the first 100k for the average node took 15min
https://imgur.com/a/eoGsOhz
5
8/30/2019 15:41:23DB6Json
Find saturation level. Observing nodes overall stability
Aug 29 2019, 8am UTC
50k blocks at 250 BPS => 250 CPS
50k blocks at 280 BPS => Started at 280 but sligtly saturated network
https://t.ly/kMEYm
https://t.ly/lDq5L
6
8/31/2019 12:45:48DB7JsonPerformance of single account spamAug 31 2019, 6am UTC25k at 30BPS live spam. CPS around 20
https://t.ly/1nxxl
7
9/1/2019 8:53:45DB7Json
Find saturation level. Observing nodes overall stability
Sep 1 2019, 1pm UTC
200k blocks at 561 BPS. CPS peak at 400, network saturation around 200.
Confirming the first 100k for the average node took 10min
First 100k was 50% faster than DB6 and 30 peers instead of 20
https://t.ly/pnNMO
8
9/3/2019 3:11:32DB8JsonCompare with v19 RC6Sep 3 2019, 7am UTC
300k blocks at 220 BPS. CPS at 220 with median confirmation time at 130ms
https://t.ly/O09g8
9
9/3/2019 14:33:37DB8JsonSearching for saturation sweet spotSep 3 2019, 8.30 pm UTC200k blocks at 300 BPS. CPS 300 with 170ms median conf time
10
9/5/2019 22:03:05DB9Json/SraymanRetesting and capturing statsSep 5 2019, 6:00 pm UTC200k blocks at 300 BPS. CPS started at 300, dipped to 214 CPS before going back up. First 4 minutes stats->counters looked good, after 4 minutes confirm_req, tcp_write_drop etc all started going up that coincided with the dip to 214 CPS.
11
9/7/2019 15:29:12DB10Json/SraymanRetesting and capturing statsSep 6, 2019, 1:25 pm UTC200k blocks at 300 BPS. Similar to DB 9 saturation hit and dipped to down again.
12
9/7/2019 15:34:27DB10Json/SraymanRetesting and capturing statsSep 6, 2019, 1:50 pm UTC350k blocks at 300 BPS with some local fanout changes only. Network maintained steady CPS
13
9/7/2019 15:34:27DB11Json/SraymanRetesting and capturing statsSep 7, 2019, 1:00 pm UTC300k blocks at 300 BPS. Similar to DB9 and DB10 results
14
9/8/2019 13:08:23DB11JsonRetesting and capturing statsSep 8, 2019, 6:00 pm UTC200k blocks at 300 BPS. No saturation all confirmed quickly. Similar to DB8 test result on Sep. 3rd.
15
9/9/2019 19:52:20DB11JsonRetesting and capturing statsSep 9, 2019, 5:12 am UTC200k single account fanout. 450 CPS
16
9/9/2019 19:53:35DB11JsonRetesting and capturing statsSep 9, 2019, 12:40 pm UTC200k blocks at 350 BPS. Measuring peak max active with high tcp_write_drop
17
9/9/2019 19:55:07DB11Json/SraymanRetesting and capturing statsSep 9, 2019, 4:11 pm UTC200k blocks at 350 BPS. Recording node using 1024 socket queue_size_max instead of 128. Very little impact, still high tcp_write_drop and max active
18
9/10/2019 14:14:21DB12robotnTesting of prioritization with DB12 and PR #2296Sep 10, 2019, 7:05 pm UTCSmall sample size, but sent about 10 low pow blocks and 10 high pow blocks. No major difference in confirmation times between low work and high work blocks, and during full saturation 15x+ blocks took 300-500 seconds to be confirmed. My memory from testing this on v19 might not be accurate, but I do seem to remember that high difficulty blocks would "always" get confirmed within around 100 seconds? Even when average confirmation times got really long? I can do a better structured test with more sends during another saturation. Edit: Forgot that blocks was sent with default watch_work = true, so low pow blocks was probably reworked to a higher difficulty when they did not get confirmed. For next test I will set watch_work = false.
19
9/12/2019 20:55:49DB14JsonRetesting and capturing statsSep 12 2019, 6:15 pm UTC200k at 350 BPS. Network performed well. Some blocks took up towards 10 seconds to confirm, but average was still < 1 second
20
9/12/2019 20:58:27DB14JsonRetesting and capturing statsSep 12 2019, 7:15 pm UTC200k at max speed. Bottleneck was on sending node slowing down after a while and could not sustain 350 BPS. Held 370-380 CPS and peaked at 400 CPS before slowing down. Little peak of CPS after blocks stopped being broadcast.