User Details
User Details
- User Since
- Feb 5 2016, 2:19 PM (230 w, 6 h)
Mon, Jun 29
Mon, Jun 29
olivier added a comment to D25454: cxgbe(4): changes in the Tx path to help increase tx coalescing..
First result on a 'small' server Xeon E5 2650 8 cores with a 10G Chelsio T540-CR (one port used as RX and the other as TX) :
x r362778: inet4 packets-per-second forwarded
+ r362778 with D25454: inet4 packets-per-second forwarded
+--------------------------------------------------------------------------+
| +|
| +|
| +|
| x +|
|x xx x +|
| |__A_| |
| A|
+--------------------------------------------------------------------------+
N Min Max Median Avg Stddev
x 5 11767747 11782957 11777032 11776267 5465.3431
+ 5 11905318 11905348 11905340 11905335 11.436783
Difference at 95.0% confidence
129068 +/- 5636.28
1.096% +/- 0.0483859%
(Student's t, pooled s = 3864.59)Tue, Jun 23
Tue, Jun 23
olivier requested review of D24613: skip bectl jail test if jail not installed.
Thu, Jun 18
Thu, Jun 18
olivier added inline comments to D24830: sendfile_test: Finish to implement linear probing to find a free port .
Tue, Jun 16
Tue, Jun 16
olivier updated the diff for D24830: sendfile_test: Finish to implement linear probing to find a free port .
Fixing style and almost all other remarks.
May 28 2020
May 28 2020
May 27 2020
May 27 2020
May 26 2020
May 26 2020
Fix error messages with SSH service.
May 14 2020
May 14 2020
olivier added inline comments to D24830: sendfile_test: Finish to implement linear probing to find a free port .
May 13 2020
May 13 2020
olivier requested review of D24830: sendfile_test: Finish to implement linear probing to find a free port .
May 8 2020
May 8 2020
May 6 2020
May 6 2020
Tested on igb and excellent result!
Apr 29 2020
Apr 29 2020
Skip bectl jail test if jail not installed (WITHOUT_JAIL).
Apr 17 2020
Apr 17 2020
olivier added a comment to D20869: vnet: Fix panic when shutting down jails and deleting interfaces simultaneously.
Once this patch applied to a current, my build kernel failed with a
olivier committed rS360045: Skip routing regression tests depending on if_epair if this module isn't….
Skip routing regression tests depending on if_epair if this module isn't…
olivier updated the diff for D24468: Skip sys/net/routing/test_rtsock_* regression tests if module if_epair doesn't exist.
Moving module load later.
Apr 12 2020
Apr 12 2020
Add 'service frr onerestart all' usage.
Apr 6 2020
Apr 6 2020
olivier added a comment to D24249: bridge: Change lists to CK_LIST as a peparation for epochification.
Here are my result:
x r359647 in bridge mode: inet 4 packets forwarded per second
+ r359647 with D24250 in bridge mode: inet 4 packets forwarded per second
+--------------------------------------------------------------------------+
|xx + |
|xx + + ++|
|A| |
| |_AM||
+--------------------------------------------------------------------------+
N Min Max Median Avg Stddev
x 5 3314595 3582601 3450403 3448858 99435.922
+ 5 22726995 24346441 23963008 23715416 617156.45
Difference at 95.0% confidence
2.02666e+07 +/- 644666
587.631% +/- 27.5422%
(Student's t, pooled s = 442024)Mar 30 2020
Mar 30 2020
Move groff dependency to an option.
Mar 29 2020
Mar 29 2020
Mar 13 2020
Mar 13 2020
Fix distfile fetch and while here improve portlint.
Mar 12 2020
Mar 12 2020
Mar 3 2020
Mar 3 2020
Skip if_epair regression test if module doesn't exist
Mar 2 2020
Mar 2 2020
olivier updated the diff for D23876: Skip if_epair test if module doesn't exist.
Keept the same behaviour as previously by loading the if_epair module.
Mar 1 2020
Mar 1 2020
Feb 28 2020
Feb 28 2020
olivier added a comment to D23886: ip6_output: use new routing KPI when not passed a cached route.
Feb 23 2020
Feb 23 2020
Feb 20 2020
Feb 20 2020
olivier added a comment to D23329: Make ICMP redirect processing depend on routing daemon.
Thanks to their maintainers, the RC scripts of net/quagga, net/openbgpd and net/openbgpd6 are now publishing the "dynamicrouting" keyword too.
Feb 14 2020
Feb 14 2020
olivier added a comment to D23674: Change smarter epoch support for NIC drivers to opt-in..
So, if I've correctly understood:
Feb 10 2020
Feb 10 2020
olivier added a comment to D23329: Make ICMP redirect processing depend on routing daemon.
Ports updated to add dynamicrouting to their PROVIDE section
olivier committed rP525717: Add 'dynamicrouting' to RC's PROVIDE, to help ICMP redirect processing depend.
Add 'dynamicrouting' to RC's PROVIDE, to help ICMP redirect processing depend
olivier committed rP525716: Add 'dynamicrouting' to RC's PROVIDE, to help ICMP redirect processing depend.
Add 'dynamicrouting' to RC's PROVIDE, to help ICMP redirect processing depend
Add 'dynamicrouting' to RC's PROVIDE
- Fix starting daemon in integrated configuration mode
Feb 3 2020
Feb 3 2020
Jan 31 2020
Jan 31 2020
Fixing pkg-plist when option LDPD is set
Jan 29 2020
Jan 29 2020
olivier added inline comments to D23329: Make ICMP redirect processing depend on routing daemon.
Jan 18 2020
Jan 18 2020
olivier added a comment to D23242: Enter network epoch for network interrupts.
And on 16core with 40Gb NIC, + 4% of forwarding speed improvement (pps unit):
x 356840
+ 356840 with D23242
+--------------------------------------------------------------------------+
|x x xx x + + + + +|
| |_______AM_____| |
| |_____AM____| |
+--------------------------------------------------------------------------+
N Min Max Median Avg Stddev
x 5 30651580 31116776 30914356 30882113 172145.91
+ 5 32002188 32402450 32244079 32214895 150605.94
Difference at 95.0% confidence
1.33278e+06 +/- 235881
4.31571% +/- 0.782654%
(Student's t, pooled s = 161735)olivier added a comment to D23242: Enter network epoch for network interrupts.
First forwarding performance impact benches.
On an old PC Engines APU2 (AMD GX 4 core 1Ghz and Intel Gigabit) there is a +6% improvement of inet4 packets-per-second forwarded:
x r356840
+ r356840 with D23242
+--------------------------------------------------------------------------+
|x x xx x + + ++ +|
| |___AM___| |
| |____A_M___| |
+--------------------------------------------------------------------------+
N Min Max Median Avg Stddev
x 5 763950.5 772812.5 768790.5 768164.7 3504.7937
+ 5 812003.5 823150 818868 817803.5 4336.1657
Difference at 95.0% confidence
49638.8 +/- 5749.84
6.462% +/- 0.767994%
(Student's t, pooled s = 3942.46)Dec 31 2019
Dec 31 2019
I didn't notice any regression on my lab (using net/frr and net/bird2) with this patch applied.
Dec 16 2019
Dec 16 2019
Dec 9 2019
Dec 9 2019
Dec 5 2019
Dec 5 2019
Dec 4 2019
Dec 4 2019
Dec 2 2019
Dec 2 2019
Add mrtparse 1.7, MRT format data parser.
Nov 28 2019
Nov 28 2019
Still not committed ?
Oct 16 2019
Oct 16 2019
Sep 27 2019
Sep 27 2019
I prefer this change over my previous improvements.
olivier added reviewers for D21820: Fix coredump_phnum_test when kern.compress_user_cores != 0: olivier, ngie, asomers.
olivier added a comment to D21776: Fix sys/kern/coredump_phnum_test in case of kern.compress_user_cores=1.
Sep 24 2019
Sep 24 2019
Forgot to close this old revision.
Fix coredump_phnum_test in case of kern.compress_user_cores=1
Sep 20 2019
Sep 20 2019
olivier added a comment to D21122: Proof-of-concept hack: remove mp_ring from TX path.
After removing those 2 lines:
3875 if (txq->ift_in_use) 3876 txq->ift_br->drain(txq->ift_br, NULL, 0, 0);
The result is now different (7% improvement):
x bypass_mpring=0 (default) and tx_abdicate=1: inet4 packets-per-second forwarded
+ bypass_mpring=1 and tx_abdicate=1: inet4 packets-per-second forwarded
+--------------------------------------------------------------------------+
|x x x x x + ++ + +|
||_______M___A___________| |
| |____M_A_____| |
+--------------------------------------------------------------------------+
N Min Max Median Avg Stddev
x 5 1047806.5 1091725 1059590 1066031.6 17660.615
+ 5 1132410 1156922 1141373 1144724.6 9694.1874
Difference at 95.0% confidence
78693 +/- 20776.4
7.38186% +/- 2.0604%
(Student's t, pooled s = 14245.6)Sep 19 2019
Sep 19 2019
olivier added a comment to D21122: Proof-of-concept hack: remove mp_ring from TX path.
So new result with new patch's version:
x bypass_mpring=0 (default) and tx_abdicate=1: inet4 packets-per-second forwarded
+ bypass_mpring=1 and tx_abdicate=1: inet4 packets-per-second forwarded
+--------------------------------------------------------------------------+
|+ ++ + + x x x x |
| |_______A_M_____||
| |______M_A________| |
+--------------------------------------------------------------------------+
N Min Max Median Avg Stddev
x 5 1043042 1084017 1074558 1070992.4 16759.158
+ 5 938950.5 984677 959011 963327.5 17971.342
Difference at 95.0% confidence
-107665 +/- 25341.6
-10.0528% +/- 2.25866%
(Student's t, pooled s = 17375.8)Sep 17 2019
Sep 17 2019
olivier added a comment to D21122: Proof-of-concept hack: remove mp_ring from TX path.
tx_abdicate was not set (default value of 0) for these benches.
If I'm setting tx_abdicate, here are the new results:
x bypass_mpring=0 (default) and tx_abdicate=1: inet4 packets-per-second forwarded
+ bypass_mpring=1 and tx_abdicate=1: inet4 packets-per-second forwarded
+--------------------------------------------------------------------------+
| + |
| + |
|+ ++ x x xx x|
| |____MA_____| |
| |___A_M__| |
+--------------------------------------------------------------------------+
N Min Max Median Avg Stddev
x 5 1060792 1099642 1069370 1073776.1 15108.116
+ 5 906915.5 934633 930094 926433 11078.667
Difference at 95.0% confidence
-147343 +/- 19320.7
-13.722% +/- 1.64298%
(Student's t, pooled s = 13247.5)Wit tx_abdicate is enabled there is about -13% of performance reduction once bypass_mpring is enabled.
Sep 12 2019
Sep 12 2019
olivier added a comment to D21122: Proof-of-concept hack: remove mp_ring from TX path.
Following the code with TUNABLE_INT_FETCH("dev.%s.%d.iflib.bypass_mpring"), I've added these line in the /boot/loader.conf.local:
dev.igb.0.iflib.bypass_mpring=1 dev.igb.1.iflib.bypass_mpring=1 dev.igb.2.iflib.bypass_mpring=1 dev.igb.3.iflib.bypass_mpring=1 dev.igb.4.iflib.bypass_mpring=1 dev.igb.5.iflib.bypass_mpring=1
New results with this new version on atom C2558 (4 cores).
Aug 28 2019
Aug 28 2019
olivier committed rP510039: Custom OSPF "wrong LSA collision detection patch" is in conflict with the fix.
Custom OSPF "wrong LSA collision detection patch" is in conflict with the fix
Aug 27 2019
Aug 27 2019
olivier added a comment to D21122: Proof-of-concept hack: remove mp_ring from TX path.
Yes you've right! The correct label should be "bypass_mpring" disabled/enabled... which revert the logic.
Aug 26 2019
Aug 26 2019
olivier added a comment to D21122: Proof-of-concept hack: remove mp_ring from TX path.
Flamegraph (fixed labels):
Aug 24 2019
Aug 24 2019
Aug 15 2019
Aug 15 2019
olivier added a comment to D21122: Proof-of-concept hack: remove mp_ring from TX path.
On another hardware (Atom 4cores):
x inet4 forwarding of small packet size, bypass_mpring=0 in packets-per-second
+ inet4 forwarding of small packet size, bypass_mpring=1 in packets-per-second
+--------------------------------------------------------------------------+
|+ + ++ + x x x x |
| |______A__M____||
| |_______________A_____M________| |
+--------------------------------------------------------------------------+
N Min Max Median Avg Stddev
x 4 927858.5 983285 971934 963752.88 24606.11
+ 5 749042 872241 859294 839484.8 51106.313
Difference at 95.0% confidence
-124268 +/- 66405
-12.8942% +/- 6.76603%
(Student's t, pooled s = 41856.6)Bigger impact her: -12%
Aug 14 2019
Aug 14 2019
olivier added a comment to D21122: Proof-of-concept hack: remove mp_ring from TX path.
Results on PC Engines APU2 (notice that I had to set back iflib.tx_abdicate=0, because if=1, all packets are dropped in another dimension):
x inet4 forwarding of small packet size, bypass_mpring=0 in packets-per-second
+ inet4 forwarding of small packet size, bypass_mpring=1 in packets-per-second
+--------------------------------------------------------------------------+
|+ + + + + x x x xx|
| |____A___||
| |_________A____M_____| |
+--------------------------------------------------------------------------+
N Min Max Median Avg Stddev
x 5 462506 464435 463851 463697.4 759.63235
+ 5 451908 456286 455316 454538 1785.9727
Difference at 95.0% confidence
-9159.4 +/- 2001.51
-1.9753% +/- 0.430346%
(Student's t, pooled s = 1372.36)Aug 10 2019
Aug 10 2019
Jul 15 2019
Jul 15 2019
Fix filename to avoid skipping lib/libc/net/servent test.
olivier added a comment to D20943: Fix test for lib/libc/net/servent_test.
Update following Enji's advice.
Jul 14 2019
Jul 14 2019
Jul 12 2019
Jul 12 2019
Jun 27 2019
Jun 27 2019
May 28 2019
May 28 2019
Restrict python versions used to build net/frr7
May 13 2019
May 13 2019
olivier added a comment to D20224: Eliminate rwlock from fast path processing in BPF code.
Let's try the benefit of this patch on a 8-core Atom with a 10G Chelsio, during a forwarding bench with tcdpump running:
x r347519: inet4 smallest size packets-per-second forwarded with tcpdump running
+ r347519 and D20224: inet4 smallest size packets-per-second forwarded with tcpdump running
+--------------------------------------------------------------------------+
| xx x + + + ++|
||___M__A______| |
| |___AM___||
+--------------------------------------------------------------------------+
N Min Max Median Avg Stddev
x 5 3682162 3816554 3684703 3711296.2 58889.489
+ 5 4179733.5 4274492 4240164 4234216.7 39505.425
Difference at 95.0% confidence
522920 +/- 73130.8
14.09% +/- 2.16578%
(Student's t, pooled s = 50143)Apr 30 2019
Apr 30 2019
Fix build with GCC-based architectures.
Apr 23 2019
Apr 23 2019
olivier committed rS346605: Skip test component_selection:run_latest_genid if gmirror/gnop GEOM classes.
Skip test component_selection:run_latest_genid if gmirror/gnop GEOM classes
olivier updated the diff for D19958: Skip test component_selection:run_latest_genid if gmirror/gnop GEOM classes aren't available.
Applying ngie's comments
olivier added inline comments to D19958: Skip test component_selection:run_latest_genid if gmirror/gnop GEOM classes aren't available.
Apr 19 2019
Apr 19 2019
Apr 18 2019
Apr 18 2019