We have a bunch of virtual interfaces. We turned on basic net accounting:
# acctadm
...
Net accounting: active
Net accounting file: /var/adm/exacct/net
Tracked net resources: basic
Untracked net resources: src_ip,dst_ip,src_port,dst_port,protocol,dsfield
# du -sh /var/adm/exacct/net
19M /var/adm/exacct/net
#
# timex /usr/sbin/dladm show-usage -f /var/adm/exacct/net
LINK DURATION IPACKETS RBYTES OPACKETS OBYTES BANDWIDTH
vnic21 101457 190292 10851134 1638 81556 0 Mbps
vnic20 101457 226535 15074788 62121 72151629 0.006 Mbps
vnic19 101457 190789 10908808 2276 119269 0 Mbps
vnic17 101457 190932 10913464 2469 128502 0 Mbps
vnic16 101457 942756 68233502 1312018 182318767 0.149 Mbps
vnic15 101457 190722 10922199 2200 120894 0 Mbps
vnic14 101457 190740 10912264 2115 132574 0 Mbps
vnic13 101457 202863 12786935 14724 1863878 0.001 Mbps
vnic12 101457 1827891 727021611 1648850 814314455 0.121 Mbps
vnic11 101457 190428 10874206 1793 93119 0 Mbps
vnic10 101457 190649 10900207 2128 111431 0 Mbps
vnic9 101457 356775 102299396 133283 54158207 0.012 Mbps
vnic8 101457 191048 10977160 2674 146583 0 Mbps
vnic7 101457 190198 10869550 1555 82795 0 Mbps
vnic6 101457 217035 14549357 29706 3606906 0.001 Mbps
vnic5 101457 190711 10928353 2182 120391 0 Mbps
vnic1 101457 2209827 332153102 2355068 368743032 0.055 Mbps
vnic0 101457 1229364 192987126 881432 102293324 0.023 Mbps
vnic4 101457 190938 10920866 2454 128011 0 Mbps
vnic3 101457 191156 10984794 2787 156696 0 Mbps
vnic2 101457 190550 10887023 1862 116420 0 Mbps
e1000g0 101457 1352037 174933145 1700086 170802132 0.148 Mbps
pnic1 101457 189734 11378082 0 0 0 Mbps
pnic0 101457 189403 11364180 331 13902 0 Mbps
real 2:12.73
user 2:11.81
sys 0.26
That is quite a long time! I created a brand new accounting file and it is at
19mb and just parsing that takes 2+ minutes. I was planning to use this for
capturing bandwidth usage over a period of one month.
I can''t imagine how long this will take to get the metrics for a file
that might be 500mb.
How can we speed this up? Its nice to have this feature, but it is pointless if
it can''t be used or doesn''t scale to these relatively low
numbers.
--
This message posted from opensolaris.org
venugopal iyer
2009-Jul-13 16:56 UTC
[crossbow-discuss] extended accounting and dladm show-usage
Hi, Anil: I think part of the time taken is due to the amount of data that gets written to the file for each link. Currently, we write a record for each link every 20 secs, which would result in a lot of records over a period of time. In the future we''d like to be able to administratively configure the interval, on a per-link basis (also enabling/disabling accounting on a per-link basis). I am not sure the kind of system you are using, but I guess, the processing time would also depend on that. We are planning on making changes in the stats area, in the meantime let me have a look at this issue .. thanks for the feedback. -venu On Sat, 11 Jul 2009, Anil wrote:> We have a bunch of virtual interfaces. We turned on basic net accounting: > > # acctadm > ... > Net accounting: active > Net accounting file: /var/adm/exacct/net > Tracked net resources: basic > Untracked net resources: src_ip,dst_ip,src_port,dst_port,protocol,dsfield > > # du -sh /var/adm/exacct/net > 19M /var/adm/exacct/net > # > # timex /usr/sbin/dladm show-usage -f /var/adm/exacct/net > LINK DURATION IPACKETS RBYTES OPACKETS OBYTES BANDWIDTH > vnic21 101457 190292 10851134 1638 81556 0 Mbps > vnic20 101457 226535 15074788 62121 72151629 0.006 Mbps > vnic19 101457 190789 10908808 2276 119269 0 Mbps > vnic17 101457 190932 10913464 2469 128502 0 Mbps > vnic16 101457 942756 68233502 1312018 182318767 0.149 Mbps > vnic15 101457 190722 10922199 2200 120894 0 Mbps > vnic14 101457 190740 10912264 2115 132574 0 Mbps > vnic13 101457 202863 12786935 14724 1863878 0.001 Mbps > vnic12 101457 1827891 727021611 1648850 814314455 0.121 Mbps > vnic11 101457 190428 10874206 1793 93119 0 Mbps > vnic10 101457 190649 10900207 2128 111431 0 Mbps > vnic9 101457 356775 102299396 133283 54158207 0.012 Mbps > vnic8 101457 191048 10977160 2674 146583 0 Mbps > vnic7 101457 190198 10869550 1555 82795 0 Mbps > vnic6 101457 217035 14549357 29706 3606906 0.001 Mbps > vnic5 101457 190711 10928353 2182 120391 0 Mbps > vnic1 101457 2209827 332153102 2355068 368743032 0.055 Mbps > vnic0 101457 1229364 192987126 881432 102293324 0.023 Mbps > vnic4 101457 190938 10920866 2454 128011 0 Mbps > vnic3 101457 191156 10984794 2787 156696 0 Mbps > vnic2 101457 190550 10887023 1862 116420 0 Mbps > e1000g0 101457 1352037 174933145 1700086 170802132 0.148 Mbps > pnic1 101457 189734 11378082 0 0 0 Mbps > pnic0 101457 189403 11364180 331 13902 0 Mbps > > real 2:12.73 > user 2:11.81 > sys 0.26 > > That is quite a long time! I created a brand new accounting file and it is at 19mb and just parsing that takes 2+ minutes. I was planning to use this for capturing bandwidth usage over a period of one month. > > I can''t imagine how long this will take to get the metrics for a file that might be 500mb. > > How can we speed this up? Its nice to have this feature, but it is pointless if it can''t be used or doesn''t scale to these relatively low numbers. > -- > This message posted from opensolaris.org > _______________________________________________ > crossbow-discuss mailing list > crossbow-discuss at opensolaris.org > http://mail.opensolaris.org/mailman/listinfo/crossbow-discuss >
I doubt I am hitting into any kind of bottleneck on the system. This is a x4150 quad core that has been pretty idle. Usually, when people take network measurements on the network gear, they sample at every 5 minutes. 20 seconds is indeed too short. The new changes you guys are planning to implement sounds like the right thing to do. But, right now I am using 2009.06 and it''ll be nice if the default of 20 can be changed to something higher as an immediate relief, that might be the easiest fix. When are these changes to the stats code going to be implement and available? Thanks! -- This message posted from opensolaris.org
venugopal iyer
2009-Jul-13 18:59 UTC
[crossbow-discuss] extended accounting and dladm show-usage
Thanks, Anil. We don''t have any details, incl. dates, for the changes as yet. We will post it on this list when we do. In the meantime we will check how we can make incremental fixes available for this (or else a workaround). -venu On Mon, 13 Jul 2009, Anil wrote:> I doubt I am hitting into any kind of bottleneck on the system. This is a x4150 quad core that has been pretty idle. Usually, when people take network measurements on the network gear, they sample at every 5 minutes. 20 seconds is indeed too short. > > The new changes you guys are planning to implement sounds like the right thing to do. But, right now I am using 2009.06 and it''ll be nice if the default of 20 can be changed to something higher as an immediate relief, that might be the easiest fix. > > When are these changes to the stats code going to be implement and available? > > Thanks! > -- > This message posted from opensolaris.org > _______________________________________________ > crossbow-discuss mailing list > crossbow-discuss at opensolaris.org > http://mail.opensolaris.org/mailman/listinfo/crossbow-discuss >
In relation to the previous topic, will PSARC/2009/364 make the 2010.03 release of OpenSolaris? Just wanted to see when the performance improvements in the show-usage (rather dlstat) will make it into the OS. Thanks> > Thanks, Anil. > > We don''t have any details, incl. dates, for the > changes as yet. We will > post it on this list when we do. In the meantime we > will check how we > can make incremental fixes available for this (or > else a workaround). > > -venu > > > On Mon, 13 Jul 2009, Anil wrote: > > > I doubt I am hitting into any kind of bottleneck on > the system. This is a x4150 quad core that has been > pretty idle. Usually, when people take network > measurements on the network gear, they sample at > every 5 minutes. 20 seconds is indeed too short. > > > > The new changes you guys are planning to implement > sounds like the right thing to do. But, right now I > am using 2009.06 and it''ll be nice if the default of > 20 can be changed to something higher as an immediate > relief, that might be the easiest fix. > > > > When are these changes to the stats code going to > be implement and available? > > > > Thanks! > > -- > > This message posted from opensolaris.org > > _______________________________________________ > > crossbow-discuss mailing list > > crossbow-discuss at opensolaris.org > > > http://mail.opensolaris.org/mailman/listinfo/crossbow- > discuss > > > _______________________________________________ > crossbow-discuss mailing list > crossbow-discuss at opensolaris.org > http://mail.opensolaris.org/mailman/listinfo/crossbow- > discuss-- This message posted from opensolaris.org
venugopal iyer
2010-Feb-05 18:55 UTC
[crossbow-discuss] extended accounting and dladm show-usage
Hi, Anil: PSARC/2009/364 is not going to part of 2010.03. The psarc case doesn''t address the enhancements to show-usage; that will follow subsequently. -venu On Fri, 5 Feb 2010, Anil wrote:> In relation to the previous topic, will PSARC/2009/364 make the 2010.03 release of OpenSolaris? Just wanted to see when the performance improvements in the show-usage (rather dlstat) will make it into the OS. > > Thanks > > >> >> Thanks, Anil. >> >> We don''t have any details, incl. dates, for the >> changes as yet. We will >> post it on this list when we do. In the meantime we >> will check how we >> can make incremental fixes available for this (or >> else a workaround). >> >> -venu >> >> >> On Mon, 13 Jul 2009, Anil wrote: >> >>> I doubt I am hitting into any kind of bottleneck on >> the system. This is a x4150 quad core that has been >> pretty idle. Usually, when people take network >> measurements on the network gear, they sample at >> every 5 minutes. 20 seconds is indeed too short. >>> >>> The new changes you guys are planning to implement >> sounds like the right thing to do. But, right now I >> am using 2009.06 and it''ll be nice if the default of >> 20 can be changed to something higher as an immediate >> relief, that might be the easiest fix. >>> >>> When are these changes to the stats code going to >> be implement and available? >>> >>> Thanks! >>> -- >>> This message posted from opensolaris.org >>> _______________________________________________ >>> crossbow-discuss mailing list >>> crossbow-discuss at opensolaris.org >>> >> http://mail.opensolaris.org/mailman/listinfo/crossbow- >> discuss >>> >> _______________________________________________ >> crossbow-discuss mailing list >> crossbow-discuss at opensolaris.org >> http://mail.opensolaris.org/mailman/listinfo/crossbow- >> discuss > -- > This message posted from opensolaris.org > _______________________________________________ > crossbow-discuss mailing list > crossbow-discuss at opensolaris.org > http://mail.opensolaris.org/mailman/listinfo/crossbow-discuss >