Iperf 2 Activity
A means to measure network responsiveness and throughput
Brought to you by:
rjmcmahon
Activity for Iperf 2
-
Pavlo Kleymonov posted a comment on discussion General Discussion
Hello, I am currently porting iperf2 to the QNX OS. This effort has a long history. In the past, some of the work was merged into version 2.0.13. Due to the event-driven architecture of QNX, QNX team has extended the existing functionality for UDP transfers by implementing burst send/receive support using the sendmmsg() and recvmmsg() system calls. This reduces per-packet overhead and can improve UDP throughput by up to 4× in some cases. I believe this approach could also improve performance on Linux....
-
Steve Shearer created ticket #390
--txstart-time does not work correctly for UDP with --reverse
-
Robert McMahon modified ticket #389
Compilation fails with --disable-ipv6
-
Thanks for this. Fixed in master per this commit
-
fix for ticket 389: Compilation fails with --disable-ipv6
-
Compilation fails with --disable-ipv6
-
Joris Guex modified a comment on ticket #389
Patch: diff --git a/src/iperf_multicast_api.c b/src/iperf_multicast_api.c index e47e7088..b0ea7ecb 100644 --- a/src/iperf_multicast_api.c +++ b/src/iperf_multicast_api.c @@ -178,7 +178,7 @@ static int iperf_multicast_join_v4_pi(struct thread_Settings *inSettings) { static int iperf_multicast_join_v6(struct thread_Settings *inSettings) { #if (HAVE_DECL_IPV6_JOIN_GROUP || HAVE_DECL_IPV6_ADD_MEMBERSHIP) -#if HAVE_STRUCT_IPV6_MREQ +#if (HAVE_IPV6 && HAVE_STRUCT_IPV6_MREQ) struct ipv6_mreq mreq; memcpy(&mreq.ipv6mr_multiaddr,...
-
Patch did not seem to attach, let me try again:
-
Compilation fails with --disable-ipv6
-
Fix double space typo in comments (most of the time -> most of the time)
-
move toggleLevel into Reporter context only
-
set columns to 120
-
.clang-format
-
remove git test in master, only support in udp-l4s branch.
-
Auto-enable UDP L4S by default on udp-l4s branch
-
Fix memory leaks in Settings_Destroy()
-
Auto-enable UDP L4S by default on udp-l4s branch
-
Fix memory leaks in Settings_Destroy()
-
don't default bounceback to one second interval reporting
-
more on rig setup
-
more on test rig setup and discovery
-
add python code for latency testing
-
clean up indentation
-
bump date
-
bump date
-
more fixes on delay_target for --ipg
-
fix IPG test
-
--txstart-time with -R
-
Update listener to skip ineligible packets for new connection establishment (e.g., L4S UDP ACK)
-
unable to get source code
-
Sorry, this is a sourceforge issue or other. Not something I can fix.
-
Lee created ticket #388
unable to get source code
-
Chia-Yu Chang created ticket #387
Update listener to skip ineligible packets for new connection establishment (e.g., L4S UDP ACK)
-
Kari Niemelä created ticket #386
--txstart-time with -R
-
Thanks Bob
-
These are directional one way delays. They're shown on the client for convenience (vs requiring one to look at the server for OWD-TX and the client of OWD-RX.) The RTT equivalent is the BB8 PDF.
-
Thanks for the clarification. So am i correct to think that OWD-TX-PDF is effectively a PDF of RTT - but perhaps without the return trip? Could I make use of OWD-RX-PDF to factor in the return trip?
-
See the man page. PDF are histograms with binning. OWD is for one way delay ng. [ 1] 0.00-5.10 sec BB8-PDF: bin(w=100us):cnt(50)=35:1,37:1,39:1,40:3,41:4,42:1,43:1,52:1,57:1,65:1,68:1,69:1,70:1,72:2,74:1,75:5,78:1,79:2,80:4,81:3,82:1,83:1,88:2,90:2,92:1,94:1,117:1,126:1,369:1,1000:1,1922:1,3710:1 (5.00/95.00/99.7%=39/1000/3710,Outliers=4,obl/obu=0/0) where, [ 1] The traffic thread number 0.00-5.10 sec The time interval of the histogram BB8-PDF BB8 is the histogram name and the PDF indicates a histogram...
-
See the man page. PDF are histograms with binning. --bounceback-period of 0 means just bb as fast as possible and don't do any bursting. [ 1] 0.00-5.10 sec BB8-PDF: bin(w=100us):cnt(50)=35:1,37:1,39:1,40:3,41:4,42:1,43:1,52:1,57:1,65:1,68:1,69:1,70:1,72:2,74:1,75:5,78:1,79:2,80:4,81:3,82:1,83:1,88:2,90:2,92:1,94:1,117:1,126:1,369:1,1000:1,1922:1,3710:1 (5.00/95.00/99.7%=39/1000/3710,Outliers=4,obl/obu=0/0) where, [ 1] The traffic thread number 0.00-5.10 sec The time interval of the histogram BB8-PDF...
-
See the man page. PDF or histograms with binning. --bounceback-period of 0 means just bb as fast as possible and don't do any bursting. [ 1] 0.00-5.10 sec BB8-PDF: bin(w=100us):cnt(50)=35:1,37:1,39:1,40:3,41:4,42:1,43:1,52:1,57:1,65:1,68:1,69:1,70:1,72:2,74:1,75:5,78:1,79:2,80:4,81:3,82:1,83:1,88:2,90:2,92:1,94:1,117:1,126:1,369:1,1000:1,1922:1,3710:1 (5.00/95.00/99.7%=39/1000/3710,Outliers=4,obl/obu=0/0) where, [ 1] The traffic thread number 0.00-5.10 sec The time interval of the histogram BB8-PDF...
-
I am trying to get something like a histogram of TCP RTT and I've been poking around at various things. Could someone point me to a description of OWD-TX-PDF and OWD-RX-PDF please, I cant seem to find a reference to what it is a PDF of exactly. Also I see in the above, --bounceback-period 0, - what does this do?
-
add back in the CE counter on the udp l4s client
-
fix for setting TXTIME
-
set the SO_TXTIME on a socket when --send-delay and udp
-
set the SO_TXTIME on a socket when --send-delay and udp
-
fix on control len in cmsg
-
fix on control len in cmsg
-
use new writemsg calls for udp l4s
-
add code for writemsg_delay_tos that uses cmsg and SO_TXTIME
-
add code for writemsg_delay_tos that uses cmsg and SO_TXTIME
-
update send-delay help
-
deprecate --tcp-tx-delay and replace it with --send-delay which will work for both TCP and UDP
-
update send-delay help
-
deprecate --tcp-tx-delay and replace it with --send-delay which will work for both TCP and UDP
-
configure support for checking SO_TXTIME +more
-
configure support for checking SO_TXTIME +more
-
on recvn and have MSG_DONTWAIT then use recv to distinguish peer close event
-
on recvn and have MSG_DONTWAIT then use recv to distinguish peer close event
-
port ranges and reverse not working
-
Mutex bug in reporter EndJob - seems to be a race
-
don't disable udp fin /w l4s - fix later
-
Option to dump raw latency
-
UDP buffer configuration fix for reverse mode
-
should be fixed. Try the --udp-l4s branch
-
bump date
-
fix v4 resize
-
fix v4 resize
-
UDP buffer configuration fix for reverse mode
-
resize client mBufLen on reverse
-
resize client mBufLen on reverse
-
fix mBuf size regression w/reverse
-
fix mBuf size regression w/reverse
-
more on l4s reverse bug fixes
-
fixes for --reverse with l4s to default to capacity seeking
-
default l4s --reverse to capacity seeking
-
default l4s --reverse to capacity seeking
-
fix CE accounting bug
-
remove a space
-
default udp l4s to capacity seeking behavior
-
default udp l4s to capacity seeking behavior
-
remove inP variance with l4s
-
l4s output handlers
-
l4s output handlers
-
l4s output handlers
-
bump date
-
initialize the udp_l4s_pkt_ack to zero properly since a stack variable
-
initialize the udp_l4s_pkt_ack to zero properly since a stack variable
-
report empty packets too with UDP L4S send
-
Yunha Choi posted a comment on discussion General Discussion
Oh, I can check this in the server header output. Server listening on UDP port 5001 with pid 43577 Read buffer size: 1.44 KByte (Dist bin width= 183 Byte) Enabled receive histograms bin-width=0.100 ms, bins=10000 (clients should use --trip-times) UDP buffer size: 208 KByte (default)
-
Thanks for your reply. When I run iperf -s -u -e --histograms, and iperf -u --trip-times -e the result shows: [ ID] Interval Transfer Bandwidth Jitter Lost/Total Latency avg/min/max/stdev PPS Rx/inP Read/Timeo/Trunc NetPwr [ 1] 0.00-15.00 sec 562 MBytes 315 Mbits/sec 0.030 ms 51/401246 (0.013%) 0.236/0.174/1.132/0.066 ms -0 pps 401195/6(0) pkts 401196/0/0 166335 [ 1] 0.00-15.00 sec T8(f)-PDF: bin(w=100us):cnt(401194)=2:81670,3:307775,4:5088,5:1025,6:1039,7:1775,8:1690,9:942,10:167,11:20,12:3 (5.00/95.00/99.7%u=0/0)...
-
cosmetic change to CE duration header line
-
move CE active counter to after duration values
-
L4S CE counter improvements
-
add -b 0 support for TCP
-
fix man page typo
-
add -b 0 support for TCP
-
fix man page typo
-
Jean Tourrilhes created ticket #382
Option to dump raw latency
-
zero tests in ce duration output
-
zero tests in ce duration output
1 >
