Re: arp -na performance w/ many permanent entries
On Sun, Jun 6, 2010 at 4:23 PM, Nick Rogers ncrog...@gmail.com wrote: On Sat, Jun 5, 2010 at 11:54 PM, Garrett Cooper yanef...@gmail.comwrote: I agree with Jeremy. I think that the problem that you've discovered is the fact that it's using stdio-based buffered output instead of buffering more of the contents in a string and punting it out in larger chunks. HTH, -Garrett I don't think so. The performance difference when taking out the interface lookup is huge even though the data output to STDOUT is mostly the same. I'll try the other lists, thanks. FYI there is a bugfix/patch for this issue being discussed in freebsd-hackers: http://www.mail-archive.com/freebsd-hack...@freebsd.org/msg157097.html Thanks again for suggesting I try another list. ___ freebsd-stable@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-stable To unsubscribe, send any mail to freebsd-stable-unsubscr...@freebsd.org
Re: arp -na performance w/ many permanent entries
On Sat, Jun 5, 2010 at 11:54 PM, Garrett Cooper yanef...@gmail.com wrote: I agree with Jeremy. I think that the problem that you've discovered is the fact that it's using stdio-based buffered output instead of buffering more of the contents in a string and punting it out in larger chunks. HTH, -Garrett I don't think so. The performance difference when taking out the interface lookup is huge even though the data output to STDOUT is mostly the same. I'll try the other lists, thanks. ___ freebsd-stable@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-stable To unsubscribe, send any mail to freebsd-stable-unsubscr...@freebsd.org
Re: arp -na performance w/ many permanent entries
On Mon, May 31, 2010 at 10:54 PM, Nick Rogers ncrog...@gmail.com wrote: [root@ ~]# time arp -na /dev/null real 0m12.761s user 0m2.959s sys 0m9.753s [root@ ~]# Notice that arp -na takes about 13s to execute even though there is no other load. This can get a lot worse by a few orders of magnitude on a loaded machine in a production environment, and seems to scale up linearly when more aliases are added to the interface (permanent ARP entries created). Is this a reasonable problem that can be fixed/improved, or am I stuck with the slow arp -na output? Any help or comments is greatly appreciated. I tried the same scenario on 8.1-BETA1 and it still takes a very long time for arp(8) to complete. I was able to isolate the performance bottleneck to a small piece of the arp(8) code. It seems that looking up the interface for an ARP entry is a very heavy operation when that entry corresponds to an alias assigned to the interface. Permanent ARP entries that do not correspond with an interface alias do not seem to cause arp(8) to puke on the interface lookup. The following commands and code diff illustrates how arp(8) can be modified to run a lot faster in this scenario, but obviously the associated interface is no longer printed for each entry. [root@ /usr/src/usr.sbin/arp]# uname -a FreeBSD .localdomain 8.1-BETA1 FreeBSD 8.1-BETA1 #0: Thu May 27 15:03:30 UTC 2010 r...@mason.cse.buffalo.edu:/usr/obj/usr/src/sys/GENERIC amd64 [root@ /usr/src/usr.sbin/arp]# time /usr/sbin/arp -na | wc -l 4100 real 0m14.903s user 0m3.133s sys 0m11.519s [root@ /usr/src/usr.sbin/arp]# pwd /usr/src/usr.sbin/arp [root@ /usr/src/usr.sbin/arp]# !diff diff -ruN arp.c.orig arp.c --- arp.c.orig 2010-06-05 18:25:24.0 + +++ arp.c 2010-06-05 18:28:19.0 + @@ -562,7 +562,7 @@ const char *host; struct hostent *hp; struct iso88025_sockaddr_dl_data *trld; - char ifname[IF_NAMESIZE]; + //char ifname[IF_NAMESIZE]; int seg; if (nflag == 0) @@ -591,8 +591,8 @@ } } else printf((incomplete)); - if (if_indextoname(sdl-sdl_index, ifname) != NULL) - printf( on %s, ifname); + //if (if_indextoname(sdl-sdl_index, ifname) != NULL) + //printf( on %s, ifname); if (rtm-rtm_rmx.rmx_expire == 0) printf( permanent); else { [root@ /usr/src/usr.sbin/arp]# make clean make rm -f arp arp.o arp.4.gz arp.8.gz arp.4.cat.gz arp.8.cat.gz Warning: Object directory not changed from original /usr/src/usr.sbin/arp cc -O2 -pipe -std=gnu99 -fstack-protector -Wsystem-headers -Werror -Wall -Wno-format-y2k -W -Wno-unused-parameter -Wstrict-prototypes -Wmissing-prototypes -Wpointer-arith -Wno-uninitialized -Wno-pointer-sign -c arp.c cc -O2 -pipe -std=gnu99 -fstack-protector -Wsystem-headers -Werror -Wall -Wno-format-y2k -W -Wno-unused-parameter -Wstrict-prototypes -Wmissing-prototypes -Wpointer-arith -Wno-uninitialized -Wno-pointer-sign -o arp arp.o gzip -cn arp.4 arp.4.gz gzip -cn arp.8 arp.8.gz [root@ /usr/src/usr.sbin/arp]# time ./arp -na | wc -l 4099 real 0m0.036s user 0m0.015s sys 0m0.021s [root@ /usr/src/usr.sbin/arp]# Notice that 0.036s without the interface lookup is a heck of a lot faster than 14.903s when doing the interface lookup. Is there something that can be done to speedup the call to if_indextoname(), or would it be worthwhile for me to submit a patch that adds the ability to skip the interface lookup as an arp(8) option? ___ freebsd-stable@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-stable To unsubscribe, send any mail to freebsd-stable-unsubscr...@freebsd.org
Re: arp -na performance w/ many permanent entries
On Sat, Jun 05, 2010 at 09:48:01PM -0400, Nick Rogers wrote: On Mon, May 31, 2010 at 10:54 PM, Nick Rogers ncrog...@gmail.com wrote: [root@ ~]# time arp -na /dev/null real 0m12.761s user 0m2.959s sys 0m9.753s [root@ ~]# Notice that arp -na takes about 13s to execute even though there is no other load. This can get a lot worse by a few orders of magnitude on a loaded machine in a production environment, and seems to scale up linearly when more aliases are added to the interface (permanent ARP entries created). Is this a reasonable problem that can be fixed/improved, or am I stuck with the slow arp -na output? Any help or comments is greatly appreciated. I tried the same scenario on 8.1-BETA1 and it still takes a very long time for arp(8) to complete. I was able to isolate the performance bottleneck to a small piece of the arp(8) code. It seems that looking up the interface for an ARP entry is a very heavy operation when that entry corresponds to an alias assigned to the interface. Permanent ARP entries that do not correspond with an interface alias do not seem to cause arp(8) to puke on the interface lookup. The following commands and code diff illustrates how arp(8) can be modified to run a lot faster in this scenario, but obviously the associated interface is no longer printed for each entry. [root@ /usr/src/usr.sbin/arp]# uname -a FreeBSD .localdomain 8.1-BETA1 FreeBSD 8.1-BETA1 #0: Thu May 27 15:03:30 UTC 2010 r...@mason.cse.buffalo.edu:/usr/obj/usr/src/sys/GENERIC amd64 [root@ /usr/src/usr.sbin/arp]# time /usr/sbin/arp -na | wc -l 4100 real 0m14.903s user 0m3.133s sys 0m11.519s [root@ /usr/src/usr.sbin/arp]# pwd /usr/src/usr.sbin/arp [root@ /usr/src/usr.sbin/arp]# !diff diff -ruN arp.c.orig arp.c --- arp.c.orig 2010-06-05 18:25:24.0 + +++ arp.c 2010-06-05 18:28:19.0 + @@ -562,7 +562,7 @@ const char *host; struct hostent *hp; struct iso88025_sockaddr_dl_data *trld; - char ifname[IF_NAMESIZE]; + //char ifname[IF_NAMESIZE]; int seg; if (nflag == 0) @@ -591,8 +591,8 @@ } } else printf((incomplete)); - if (if_indextoname(sdl-sdl_index, ifname) != NULL) - printf( on %s, ifname); + //if (if_indextoname(sdl-sdl_index, ifname) != NULL) + //printf( on %s, ifname); if (rtm-rtm_rmx.rmx_expire == 0) printf( permanent); else { [root@ /usr/src/usr.sbin/arp]# make clean make rm -f arp arp.o arp.4.gz arp.8.gz arp.4.cat.gz arp.8.cat.gz Warning: Object directory not changed from original /usr/src/usr.sbin/arp cc -O2 -pipe -std=gnu99 -fstack-protector -Wsystem-headers -Werror -Wall -Wno-format-y2k -W -Wno-unused-parameter -Wstrict-prototypes -Wmissing-prototypes -Wpointer-arith -Wno-uninitialized -Wno-pointer-sign -c arp.c cc -O2 -pipe -std=gnu99 -fstack-protector -Wsystem-headers -Werror -Wall -Wno-format-y2k -W -Wno-unused-parameter -Wstrict-prototypes -Wmissing-prototypes -Wpointer-arith -Wno-uninitialized -Wno-pointer-sign -o arp arp.o gzip -cn arp.4 arp.4.gz gzip -cn arp.8 arp.8.gz [root@ /usr/src/usr.sbin/arp]# time ./arp -na | wc -l 4099 real 0m0.036s user 0m0.015s sys 0m0.021s [root@ /usr/src/usr.sbin/arp]# Notice that 0.036s without the interface lookup is a heck of a lot faster than 14.903s when doing the interface lookup. Is there something that can be done to speedup the call to if_indextoname(), or would it be worthwhile for me to submit a patch that adds the ability to skip the interface lookup as an arp(8) option? This might be a better question for either freebsd-net or freebsd-hackers. I should warn you in advance that you might receive a bit of flack given that you have over 4000 IP aliases assigned to an interface. Explaining your setup may also help people understand why it is you need what you do. -- | Jeremy Chadwick j...@parodius.com | | Parodius Networking http://www.parodius.com/ | | UNIX Systems Administrator Mountain View, CA, USA | | Making life hard for others since 1977. PGP: 4BD6C0CB | ___ freebsd-stable@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-stable To unsubscribe, send any mail to freebsd-stable-unsubscr...@freebsd.org
Re: arp -na performance w/ many permanent entries
On Sat, Jun 5, 2010 at 8:16 PM, Jeremy Chadwick free...@jdc.parodius.com wrote: On Sat, Jun 05, 2010 at 09:48:01PM -0400, Nick Rogers wrote: On Mon, May 31, 2010 at 10:54 PM, Nick Rogers ncrog...@gmail.com wrote: [root@ ~]# time arp -na /dev/null real 0m12.761s user 0m2.959s sys 0m9.753s [root@ ~]# Notice that arp -na takes about 13s to execute even though there is no other load. This can get a lot worse by a few orders of magnitude on a loaded machine in a production environment, and seems to scale up linearly when more aliases are added to the interface (permanent ARP entries created). Is this a reasonable problem that can be fixed/improved, or am I stuck with the slow arp -na output? Any help or comments is greatly appreciated. I tried the same scenario on 8.1-BETA1 and it still takes a very long time for arp(8) to complete. I was able to isolate the performance bottleneck to a small piece of the arp(8) code. It seems that looking up the interface for an ARP entry is a very heavy operation when that entry corresponds to an alias assigned to the interface. Permanent ARP entries that do not correspond with an interface alias do not seem to cause arp(8) to puke on the interface lookup. The following commands and code diff illustrates how arp(8) can be modified to run a lot faster in this scenario, but obviously the associated interface is no longer printed for each entry. [root@ /usr/src/usr.sbin/arp]# uname -a FreeBSD .localdomain 8.1-BETA1 FreeBSD 8.1-BETA1 #0: Thu May 27 15:03:30 UTC 2010 r...@mason.cse.buffalo.edu:/usr/obj/usr/src/sys/GENERIC amd64 [root@ /usr/src/usr.sbin/arp]# time /usr/sbin/arp -na | wc -l 4100 real 0m14.903s user 0m3.133s sys 0m11.519s [root@ /usr/src/usr.sbin/arp]# pwd /usr/src/usr.sbin/arp [root@ /usr/src/usr.sbin/arp]# !diff diff -ruN arp.c.orig arp.c --- arp.c.orig 2010-06-05 18:25:24.0 + +++ arp.c 2010-06-05 18:28:19.0 + @@ -562,7 +562,7 @@ const char *host; struct hostent *hp; struct iso88025_sockaddr_dl_data *trld; - char ifname[IF_NAMESIZE]; + //char ifname[IF_NAMESIZE]; int seg; if (nflag == 0) @@ -591,8 +591,8 @@ } } else printf((incomplete)); - if (if_indextoname(sdl-sdl_index, ifname) != NULL) - printf( on %s, ifname); + //if (if_indextoname(sdl-sdl_index, ifname) != NULL) + //printf( on %s, ifname); if (rtm-rtm_rmx.rmx_expire == 0) printf( permanent); else { [root@ /usr/src/usr.sbin/arp]# make clean make rm -f arp arp.o arp.4.gz arp.8.gz arp.4.cat.gz arp.8.cat.gz Warning: Object directory not changed from original /usr/src/usr.sbin/arp cc -O2 -pipe -std=gnu99 -fstack-protector -Wsystem-headers -Werror -Wall -Wno-format-y2k -W -Wno-unused-parameter -Wstrict-prototypes -Wmissing-prototypes -Wpointer-arith -Wno-uninitialized -Wno-pointer-sign -c arp.c cc -O2 -pipe -std=gnu99 -fstack-protector -Wsystem-headers -Werror -Wall -Wno-format-y2k -W -Wno-unused-parameter -Wstrict-prototypes -Wmissing-prototypes -Wpointer-arith -Wno-uninitialized -Wno-pointer-sign -o arp arp.o gzip -cn arp.4 arp.4.gz gzip -cn arp.8 arp.8.gz [root@ /usr/src/usr.sbin/arp]# time ./arp -na | wc -l 4099 real 0m0.036s user 0m0.015s sys 0m0.021s [root@ /usr/src/usr.sbin/arp]# Notice that 0.036s without the interface lookup is a heck of a lot faster than 14.903s when doing the interface lookup. Is there something that can be done to speedup the call to if_indextoname(), or would it be worthwhile for me to submit a patch that adds the ability to skip the interface lookup as an arp(8) option? This might be a better question for either freebsd-net or freebsd-hackers. I should warn you in advance that you might receive a bit of flack given that you have over 4000 IP aliases assigned to an interface. Explaining your setup may also help people understand why it is you need what you do. I agree with Jeremy. I think that the problem that you've discovered is the fact that it's using stdio-based buffered output instead of buffering more of the contents in a string and punting it out in larger chunks. HTH, -Garrett ___ freebsd-stable@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-stable To unsubscribe, send any mail to freebsd-stable-unsubscr...@freebsd.org
arp -na performance w/ many permanent entries
I have an 8.0-RELEASE system with 4000 permanent ARP entries due to having a network interface (em(4)) configured with 4000 aliases. The arp -na command takes what I consider to be an extremely long time to finish (up to 30s on an otherwise unloaded system). I am able to replicate this in a test environment by installing 8.0-RELEASE-amd64 on a VMWare VM w/ 1GB of RAM and a 2GHz CPU. The 4000 aliases/entries is arbitrary, but nicely illustrates the performance problem. The performance is much worse on a real/loaded system. I realize the 4k aliases on an interface is unusual but I have been effectively using this configuration in my network to try and keep my end-users's each on his/her own broadcast domain. The box is a router and I allocate addresses to each user and put each on his/her own subnet with a netmask of /30. If you would like more info on this I can provide it, but it has worked effectively in FreeBSD 6.0-7.2. The slow performance of arp -na is an issue for me because I have a web/CGI tool that runs various reports, many of them relying on acquiring the current ARP table, and the performance of arp(8) makes the web interface extremely slow. I believe the problem was introduced between 7.2 and 8.0, when, as far as I understand, parts of the ARP subsystem were improved. In 7.2, the aliases configured on an interface were not considered ARP entries (at least according to arp(8)), but as of 8.0 they are marked as permanent ARP entries and displayed by arp(8), which seems to attribute to the performance problem. I ran the following perl script to setup my test system. This script was run after installing 8.0-RELEASE and adding the bash, perl, and p5-NetAddr-IP packages via pkg_add -r. #!/usr/bin/perl use strict; use diagnostics; use NetAddr::IP; my $interface = 'em1'; my $cidr= '10.0.0.1/18'; # configure the interface with 4000 or so aliases foreach my $na (@{NetAddr::IP-new($cidr)-splitref(30)}) { my $ip= $na-addr(); my $mask = $na-mask(); my $bcast = $na-broadcast()-addr(); my $cmd = ifconfig $interface inet alias $ip netmask $mask broadcast $bcast; print STDERR $cmd\n; system($cmd); } The results are as follows: [root@ ~]# uname -a FreeBSD .localdomain 8.0-RELEASE FreeBSD 8.0-RELEASE #0: Sat Nov 21 15:02:08 UTC 2009 r...@mason.cse.buffalo.edu:/usr/obj/usr/src/sys/GENERIC amd64 [root@ ~]# ifconfig -a | wc -l 4113 [root@ ~]# ifconfig -a | head -15 em0: flags=8843UP,BROADCAST,RUNNING,SIMPLEX,MULTICAST metric 0 mtu 1500 options=9bRXCSUM,TXCSUM,VLAN_MTU,VLAN_HWTAGGING,VLAN_HWCSUM ether 00:0c:29:65:4d:3e inet 172.16.16.244 netmask 0xff00 broadcast 172.16.16.255 media: Ethernet autoselect (1000baseT full-duplex) status: active em1: flags=8843UP,BROADCAST,RUNNING,SIMPLEX,MULTICAST metric 0 mtu 1500 options=9bRXCSUM,TXCSUM,VLAN_MTU,VLAN_HWTAGGING,VLAN_HWCSUM ether 00:0c:29:65:4d:48 inet 10.0.0.0 netmask 0xfffc broadcast 10.0.0.3 inet 10.0.0.4 netmask 0xfffc broadcast 10.0.0.7 inet 10.0.0.8 netmask 0xfffc broadcast 10.0.0.11 inet 10.0.0.12 netmask 0xfffc broadcast 10.0.0.15 inet 10.0.0.16 netmask 0xfffc broadcast 10.0.0.19 inet 10.0.0.20 netmask 0xfffc broadcast 10.0.0.23 [root@ ~]# time ifconfig -a /dev/null real 0m0.032s user 0m0.023s sys 0m0.008s [root@ ~]# arp -na | wc -l 4100 [root@ ~]# arp -na | tail -15 ? (10.0.5.80) at 00:0c:29:65:4d:48 on em1 permanent [ethernet] ? (10.0.5.48) at 00:0c:29:65:4d:48 on em1 permanent [ethernet] ? (10.0.5.16) at 00:0c:29:65:4d:48 on em1 permanent [ethernet] ? (10.0.1.244) at 00:0c:29:65:4d:48 on em1 permanent [ethernet] ? (10.0.1.212) at 00:0c:29:65:4d:48 on em1 permanent [ethernet] ? (10.0.1.180) at 00:0c:29:65:4d:48 on em1 permanent [ethernet] ? (10.0.1.148) at 00:0c:29:65:4d:48 on em1 permanent [ethernet] ? (10.0.1.116) at 00:0c:29:65:4d:48 on em1 permanent [ethernet] ? (10.0.1.84) at 00:0c:29:65:4d:48 on em1 permanent [ethernet] ? (10.0.1.52) at 00:0c:29:65:4d:48 on em1 permanent [ethernet] ? (10.0.1.20) at 00:0c:29:65:4d:48 on em1 permanent [ethernet] ? (172.16.16.1) at 00:50:56:c0:00:08 on em0 [ethernet] ? (172.16.16.2) at 00:50:56:ea:ea:1a on em0 [ethernet] ? (172.16.16.254) at 00:50:56:f2:75:00 on em0 [ethernet] ? (172.16.16.244) at 00:0c:29:65:4d:3e on em0 permanent [ethernet] [root@ ~]# uptime 7:28PM up 42 mins, 2 users, load averages: 0.00, 0.00, 0.00 [root@ ~]# time arp -na /dev/null real 0m12.761s user 0m2.959s sys 0m9.753s [root@ ~]# Notice that arp -na takes about 13s to execute even though there is no other load. This can get a lot worse by a few orders of magnitude on a loaded machine in a production environment, and seems to scale up linearly when more aliases are added to the interface (permanent ARP entries created). Is this a reasonable problem that can be fixed/improved, or am I stuck with the slow arp -na output? Any help or comments is greatly appreciated. ___ freebsd-stable@freebsd.org mailing