VoIP Mailing List Archives
Mailing list archives for the VoIP community |
|
View previous topic :: View next topic |
Author |
Message |
gmaruzz at celliax.org Guest
|
Posted: Tue Aug 25, 2009 4:05 am Post subject: [Freeswitch-users] Freeswitch performance as a redirecting s |
|
|
Maybe your load comes from disk access?
Try putting the sql and log directories on a ramdisk.
OTH,
-giovanni
On Tue, Aug 25, 2009 at 10:54 AM, Tihomir Culjaga<tculjaga@gmail.com> wrote:
Quote: | Hello,
i'm trying to use freeswitch as a redirecting server meaning FS has to
receive an INVITE and according to some rules it will redirect calls to
other destinations.
CALLING_USER FREESWITCH SOMEWHERE
INVITE ------------------------------->
<------------------------------ 100 Trying
<------------------------------ 302 Moved Temporary
ACK ------------------------------->
INVITE--------------------------------------------------------------------------------->
Well, wverything works well except i have perfromance issues .... on my HW
FS cannot do more than 40 CPS (INVITE answered by 302 Moved Temporary). When
i increase the rate, FS starts delaying 302 response. Right at 50 CPS i see
"calls" being build up in FS and the delay begining to grow.
When i observe the machine, load average is almost nothing (load average:
1.41, 0.61, 0.60) CPU never goes to 100%, and i see only one thread taking
most load... all others are just sitting there with 1-5 % CPU time.
This looks to me as FS handles 302 messages in a single thread?!?!
tculjaga@FS:/usr/local/freeswitch/conf/dialplan$ top -H
top - 10:41:37 up 167 days, 20:42, 3 users, load average: 1.41, 0.61, 0.60
Tasks: 83 total, 2 running, 81 sleeping, 0 stopped, 0 zombie
Cpu(s): 25.3%us, 1.5%sy, 0.0%ni, 30.3%id, 42.7%wa, 0.0%hi, 0.2%si,
0.0%st
Mem: 2074520k total, 571244k used, 1503276k free, 259604k buffers
Swap: 2650684k total, 3020k used, 2647664k free, 153868k cached
PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+
COMMAND
4814 root 20 0 34188 20m 3780 S 38 1.0 3:10.29
freeswitch
4800 root 20 0 34188 20m 3780 S 6 1.0 0:08.26
freeswitch
4798 root 20 0 34188 20m 3780 R 5 1.0 0:24.46
freeswitch
4787 root 20 0 34188 20m 3780 S 2 1.0 0:11.24
freeswitch
4794 root 20 0 34188 20m 3780 S 1 1.0 0:11.42
freeswitch
4803 root 20 0 34188 20m 3780 S 1 1.0 0:11.74
freeswitch
4788 root 20 0 34188 20m 3780 S 1 1.0 0:02.96
freeswitch
4804 root 20 0 34188 20m 3780 S 1 1.0 0:01.64
freeswitch
4807 root 20 0 34188 20m 3780 S 1 1.0 0:01.68
freeswitch
4811 root 20 0 34188 20m 3780 S 1 1.0 0:02.50 freeswitch
cat /proc/cpuinfo
processor : 0
vendor_id : GenuineIntel
cpu family : 6
model : 15
model name : Intel(R) Xeon(R) CPU 5140 @ 2.33GHz
stepping : 6
cpu MHz : 2333.560
cache size : 4096 KB
physical id : 0
siblings : 2
core id : 0
cpu cores : 2
apicid : 0
initial apicid : 0
fdiv_bug : no
hlt_bug : no
f00f_bug : no
coma_bug : no
fpu : yes
fpu_exception : yes
cpuid level : 10
wp : yes
flags : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca
cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe lm
constant_tsc arch_perfmon pebs bts pni monitor ds_cpl vmx est tm2 ssse3 cx16
xtpr dca lahf_lm
bogomips : 4670.78
clflush size : 64
power management:
processor : 1
vendor_id : GenuineIntel
cpu family : 6
model : 15
model name : Intel(R) Xeon(R) CPU 5140 @ 2.33GHz
stepping : 6
cpu MHz : 2333.560
cache size : 4096 KB
physical id : 0
siblings : 2
core id : 1
cpu cores : 2
apicid : 1
initial apicid : 1
fdiv_bug : no
hlt_bug : no
f00f_bug : no
coma_bug : no
fpu : yes
fpu_exception : yes
cpuid level : 10
wp : yes
flags : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca
cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe lm
constant_tsc arch_perfmon pebs bts pni monitor ds_cpl vmx est tm2 ssse3 cx16
xtpr dca lahf_lm
bogomips : 4666.82
clflush size : 64
power management:
uname -a
Linux l01sipindir1 2.6.26-1-686 #1 SMP Sat Jan 10 18:29:31 UTC 2009 i686
GNU/Linux
Of course, i've tuned the machine up
ulimit -c unlimited
ulimit -d unlimited
ulimit -f unlimited
ulimit -i unlimited
ulimit -n 999999
ulimit -q unlimited
ulimit -u unlimited
ulimit -v unlimited
ulimit -x unlimited
ulimit -s 240
ulimit -l unlimited
ulimit -a
Started FS with minimum modules but still 40 CPS seems to be the limit.
So, is there any way to improve performance?
Tihomir.
_______________________________________________
FreeSWITCH-users mailing list
FreeSWITCH-users@lists.freeswitch.org
http://lists.freeswitch.org/mailman/listinfo/freeswitch-users
UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users
http://www.freeswitch.org
|
_______________________________________________
FreeSWITCH-users mailing list
FreeSWITCH-users@lists.freeswitch.org
http://lists.freeswitch.org/mailman/listinfo/freeswitch-users
UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users
http://www.freeswitch.org |
|
Back to top |
|
|
tculjaga at gmail.com Guest
|
Posted: Tue Aug 25, 2009 8:27 am Post subject: [Freeswitch-users] Freeswitch performance as a redirecting s |
|
|
Hey Giovanni,
thanks for the tip... indeed the db files were heavily used regardless if i started freeswitch with nosql option (freeswitch -nosql)... FS was not writing anything into that files ... instead it was just accessing it....
This behaviour leads to a waste of 40% CPU time... waiting for other processes (mainly disk access) to finish!!!
I moved freeswitch/db/ to a ramdisk and the performance got a boost to 140 CPS with a CPU load of 80%. I was keeping the machine for a while (20 - 30 minutes) on that rate when i sow CPU suddenly went to 100% and FS becoming irresponsive .
What can be wrong?
What are the limits in CPU usage (50%, 60%, 70%, 80%...) we should not cross?
What fine tuning do we need in order to asure a long high load run?
Also, I'm running 32-bit OS (debian 5) on a 64 bit CPU... does it have sense to move my OS to 64 bit? ... will FS gain more preformance ?... I mean will FS perofomr drastically better 20%+ ?
Tihomir.
On Tue, Aug 25, 2009 at 11:00 AM, Giovanni Maruzzelli <gmaruzz@celliax.org (gmaruzz@celliax.org)> wrote:
Quote: | Maybe your load comes from disk access?
Try putting the sql and log directories on a ramdisk.
OTH,
-giovanni
On Tue, Aug 25, 2009 at 10:54 AM, Tihomir Culjaga<tculjaga@gmail.com (tculjaga@gmail.com)> wrote:
Quote: | Hello,
i'm trying to use freeswitch as a redirecting server meaning FS has to
receive an INVITE and according to some rules it will redirect calls to
other destinations.
CALLING_USER FREESWITCH SOMEWHERE
INVITE ------------------------------->
<------------------------------ 100 Trying
<------------------------------ 302 Moved Temporary
ACK ------------------------------->
INVITE--------------------------------------------------------------------------------->
Well, wverything works well except i have perfromance issues .... on my HW
FS cannot do more than 40 CPS (INVITE answered by 302 Moved Temporary). When
i increase the rate, FS starts delaying 302 response. Right at 50 CPS i see
"calls" being build up in FS and the delay begining to grow.
When i observe the machine, load average is almost nothing (load average:
1.41, 0.61, 0.60) CPU never goes to 100%, and i see only one thread taking
most load... all others are just sitting there with 1-5 % CPU time.
This looks to me as FS handles 302 messages in a single thread?!?!
tculjaga@FS:/usr/local/freeswitch/conf/dialplan$ top -H
top - 10:41:37 up 167 days, 20:42, 3 users, load average: 1.41, 0.61, 0.60
Tasks: 83 total, 2 running, 81 sleeping, 0 stopped, 0 zombie
Cpu(s): 25.3%us, 1.5%sy, 0.0%ni, 30.3%id, 42.7%wa, 0.0%hi, 0.2%si,
0.0%st
Mem: 2074520k total, 571244k used, 1503276k free, 259604k buffers
Swap: 2650684k total, 3020k used, 2647664k free, 153868k cached
PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+
COMMAND
4814 root 20 0 34188 20m 3780 S 38 1.0 3:10.29
freeswitch
4800 root 20 0 34188 20m 3780 S 6 1.0 0:08.26
freeswitch
4798 root 20 0 34188 20m 3780 R 5 1.0 0:24.46
freeswitch
4787 root 20 0 34188 20m 3780 S 2 1.0 0:11.24
freeswitch
4794 root 20 0 34188 20m 3780 S 1 1.0 0:11.42
freeswitch
4803 root 20 0 34188 20m 3780 S 1 1.0 0:11.74
freeswitch
4788 root 20 0 34188 20m 3780 S 1 1.0 0:02.96
freeswitch
4804 root 20 0 34188 20m 3780 S 1 1.0 0:01.64
freeswitch
4807 root 20 0 34188 20m 3780 S 1 1.0 0:01.68
freeswitch
4811 root 20 0 34188 20m 3780 S 1 1.0 0:02.50 freeswitch
cat /proc/cpuinfo
processor : 0
vendor_id : GenuineIntel
cpu family : 6
model : 15
model name : Intel(R) Xeon(R) CPU 5140 @ 2.33GHz
stepping : 6
cpu MHz : 2333.560
cache size : 4096 KB
physical id : 0
siblings : 2
core id : 0
cpu cores : 2
apicid : 0
initial apicid : 0
fdiv_bug : no
hlt_bug : no
f00f_bug : no
coma_bug : no
fpu : yes
fpu_exception : yes
cpuid level : 10
wp : yes
flags : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca
cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe lm
constant_tsc arch_perfmon pebs bts pni monitor ds_cpl vmx est tm2 ssse3 cx16
xtpr dca lahf_lm
bogomips : 4670.78
clflush size : 64
power management:
processor : 1
vendor_id : GenuineIntel
cpu family : 6
model : 15
model name : Intel(R) Xeon(R) CPU 5140 @ 2.33GHz
stepping : 6
cpu MHz : 2333.560
cache size : 4096 KB
physical id : 0
siblings : 2
core id : 1
cpu cores : 2
apicid : 1
initial apicid : 1
fdiv_bug : no
hlt_bug : no
f00f_bug : no
coma_bug : no
fpu : yes
fpu_exception : yes
cpuid level : 10
wp : yes
flags : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca
cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe lm
constant_tsc arch_perfmon pebs bts pni monitor ds_cpl vmx est tm2 ssse3 cx16
xtpr dca lahf_lm
bogomips : 4666.82
clflush size : 64
power management:
uname -a
Linux l01sipindir1 2.6.26-1-686 #1 SMP Sat Jan 10 18:29:31 UTC 2009 i686
GNU/Linux
Of course, i've tuned the machine up
ulimit -c unlimited
ulimit -d unlimited
ulimit -f unlimited
ulimit -i unlimited
ulimit -n 999999
ulimit -q unlimited
ulimit -u unlimited
ulimit -v unlimited
ulimit -x unlimited
ulimit -s 240
ulimit -l unlimited
ulimit -a
Started FS with minimum modules but still 40 CPS seems to be the limit.
So, is there any way to improve performance?
Tihomir.
|
_______________________________________________
FreeSWITCH-users mailing list
FreeSWITCH-users@lists.freeswitch.org (FreeSWITCH-users@lists.freeswitch.org)
http://lists.freeswitch.org/mailman/listinfo/freeswitch-users
UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users
http://www.freeswitch.org
|
|
|
Back to top |
|
|
tculjaga at gmail.com Guest
|
Posted: Tue Aug 25, 2009 8:46 am Post subject: [Freeswitch-users] Freeswitch performance as a redirecting s |
|
|
thanks for the feedback... this is something im going to do tomorrow...
what about other things?
On Tue, Aug 25, 2009 at 3:39 PM, Jay Binks <jaybinks@gmail.com (jaybinks@gmail.com)> wrote:
Quote: | Everytime someone asks this , the resounding answer is use a 64bit os..
No question
Jay
On 25/08/2009, at 23:19, Tihomir Culjaga <tculjaga@gmail.com (tculjaga@gmail.com)> wrote:
Quote: | Hey Giovanni,
thanks for the tip... indeed the db files were heavily used regardless if i started freeswitch with nosql option (freeswitch -nosql)... FS was not writing anything into that files ... instead it was just accessing it....
This behaviour leads to a waste of 40% CPU time... waiting for other processes (mainly disk access) to finish!!!
I moved freeswitch/db/ to a ramdisk and the performance got a boost to 140 CPS with a CPU load of 80%. I was keeping the machine for a while (20 - 30 minutes) on that rate when i sow CPU suddenly went to 100% and FS becoming irresponsive .
What can be wrong?
What are the limits in CPU usage (50%, 60%, 70%, 80%...) we should not cross?
What fine tuning do we need in order to asure a long high load run?
Also, I'm running 32-bit OS (debian 5) on a 64 bit CPU... does it have sense to move my OS to 64 bit? ... will FS gain more preformance ?... I mean will FS perofomr drastically better 20%+ ?
Tihomir.
On Tue, Aug 25, 2009 at 11:00 AM, Giovanni Maruzzelli < (gmaruzz@celliax.org)gmaruzz@celliax.org (gmaruzz@celliax.org)> wrote:
Quote: | Maybe your load comes from disk access?
Try putting the sql and log directories on a ramdisk.
OTH,
-giovanni
On Tue, Aug 25, 2009 at 10:54 AM, Tihomir Culjaga< (tculjaga@gmail.com)tculjaga@gmail.com (tculjaga@gmail.com)> wrote:
Quote: | Hello,
i'm trying to use freeswitch as a redirecting server meaning FS has to
receive an INVITE and according to some rules it will redirect calls to
other destinations.
CALLING_USER FREESWITCH SOMEWHERE
INVITE ------------------------------->
<------------------------------ 100 Trying
<------------------------------ 302 Moved Temporary
ACK ------------------------------->
INVITE--------------------------------------------------------------------------------->
Well, wverything works well except i have perfromance issues .... on my HW
FS cannot do more than 40 CPS (INVITE answered by 302 Moved Temporary). When
i increase the rate, FS starts delaying 302 response. Right at 50 CPS i see
"calls" being build up in FS and the delay begining to grow.
When i observe the machine, load average is almost nothing (load average:
1.41, 0.61, 0.60) CPU never goes to 100%, and i see only one thread taking
most load... all others are just sitting there with 1-5 % CPU time.
This looks to me as FS handles 302 messages in a single thread?!?!
tculjaga@FS:/usr/local/freeswitch/conf/dialplan$ top -H
top - 10:41:37 up 167 days, 20:42, 3 users, load average: 1.41, 0.61, 0.60
Tasks: 83 total, 2 running, 81 sleeping, 0 stopped, 0 zombie
Cpu(s): 25.3%us, 1.5%sy, 0.0%ni, 30.3%id, 42.7%wa, 0.0%hi, 0.2%si,
0.0%st
Mem: 2074520k total, 571244k used, 1503276k free, 259604k buffers
Swap: 2650684k total, 3020k used, 2647664k free, 153868k cached
PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+
COMMAND
4814 root 20 0 34188 20m 3780 S 38 1.0 3:10.29
freeswitch
4800 root 20 0 34188 20m 3780 S 6 1.0 0:08.26
freeswitch
4798 root 20 0 34188 20m 3780 R 5 1.0 0:24.46
freeswitch
4787 root 20 0 34188 20m 3780 S 2 1.0 0:11.24
freeswitch
4794 root 20 0 34188 20m 3780 S 1 1.0 0:11.42
freeswitch
4803 root 20 0 34188 20m 3780 S 1 1.0 0:11.74
freeswitch
4788 root 20 0 34188 20m 3780 S 1 1.0 0:02.96
freeswitch
4804 root 20 0 34188 20m 3780 S 1 1.0 0:01.64
freeswitch
4807 root 20 0 34188 20m 3780 S 1 1.0 0:01.68
freeswitch
4811 root 20 0 34188 20m 3780 S 1 1.0 0:02.50 freeswitch
cat /proc/cpuinfo
processor : 0
vendor_id : GenuineIntel
cpu family : 6
model : 15
model name : Intel(R) Xeon(R) CPU 5140 @ 2.33GHz
stepping : 6
cpu MHz : 2333.560
cache size : 4096 KB
physical id : 0
siblings : 2
core id : 0
cpu cores : 2
apicid : 0
initial apicid : 0
fdiv_bug : no
hlt_bug : no
f00f_bug : no
coma_bug : no
fpu : yes
fpu_exception : yes
cpuid level : 10
wp : yes
flags : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca
cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe lm
constant_tsc arch_perfmon pebs bts pni monitor ds_cpl vmx est tm2 ssse3 cx16
xtpr dca lahf_lm
bogomips : 4670.78
clflush size : 64
power management:
processor : 1
vendor_id : GenuineIntel
cpu family : 6
model : 15
model name : Intel(R) Xeon(R) CPU 5140 @ 2.33GHz
stepping : 6
cpu MHz : 2333.560
cache size : 4096 KB
physical id : 0
siblings : 2
core id : 1
cpu cores : 2
apicid : 1
initial apicid : 1
fdiv_bug : no
hlt_bug : no
f00f_bug : no
coma_bug : no
fpu : yes
fpu_exception : yes
cpuid level : 10
wp : yes
flags : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca
cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe lm
constant_tsc arch_perfmon pebs bts pni monitor ds_cpl vmx est tm2 ssse3 cx16
xtpr dca lahf_lm
bogomips : 4666.82
clflush size : 64
power management:
uname -a
Linux l01sipindir1 2.6.26-1-686 #1 SMP Sat Jan 10 18:29:31 UTC 2009 i686
GNU/Linux
Of course, i've tuned the machine up
ulimit -c unlimited
ulimit -d unlimited
ulimit -f unlimited
ulimit -i unlimited
ulimit -n 999999
ulimit -q unlimited
ulimit -u unlimited
ulimit -v unlimited
ulimit -x unlimited
ulimit -s 240
ulimit -l unlimited
ulimit -a
Started FS with minimum modules but still 40 CPS seems to be the limit.
So, is there any way to improve performance?
Tihomir.
|
_______________________________________________
FreeSWITCH-users mailing list
(FreeSWITCH-users@lists.freeswitch.org)FreeSWITCH-users@lists.freeswitch.org (FreeSWITCH-users@lists.freeswitch.org)
[/url][url=http://lists.freeswitch.org/mailman/listinfo/freeswitch-users]http://lists.freeswitch.org/mailman/listinfo/freeswitch-users
UNSUBSCRIBE:[/url][url=http://lists.freeswitch.org/mailman/options/freeswitch-users]http://lists.freeswitch.org/mailman/options/freeswitch-users
[/url][url=http://www.freeswitch.org]http://www.freeswitch.org
|
_______________________________________________
FreeSWITCH-users mailing list
FreeSWITCH-users@lists.freeswitch.org (FreeSWITCH-users@lists.freeswitch.org)
http://lists.freeswitch.org/mailman/listinfo/freeswitch-users
UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users
http://www.freeswitch.org
|
_______________________________________________
FreeSWITCH-users mailing list
FreeSWITCH-users@lists.freeswitch.org (FreeSWITCH-users@lists.freeswitch.org)
http://lists.freeswitch.org/mailman/listinfo/freeswitch-users
UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users
http://www.freeswitch.org
|
|
|
Back to top |
|
|
jaybinks at gmail.com Guest
|
Posted: Tue Aug 25, 2009 8:46 am Post subject: [Freeswitch-users] Freeswitch performance as a redirecting s |
|
|
Everytime someone asks this , the resounding answer is use a 64bit os..
No question
Jay
On 25/08/2009, at 23:19, Tihomir Culjaga <tculjaga@gmail.com (tculjaga@gmail.com)> wrote:
Quote: | Hey Giovanni,
thanks for the tip... indeed the db files were heavily used regardless if i started freeswitch with nosql option (freeswitch -nosql)... FS was not writing anything into that files ... instead it was just accessing it....
This behaviour leads to a waste of 40% CPU time... waiting for other processes (mainly disk access) to finish!!!
I moved freeswitch/db/ to a ramdisk and the performance got a boost to 140 CPS with a CPU load of 80%. I was keeping the machine for a while (20 - 30 minutes) on that rate when i sow CPU suddenly went to 100% and FS becoming irresponsive .
What can be wrong?
What are the limits in CPU usage (50%, 60%, 70%, 80%...) we should not cross?
What fine tuning do we need in order to asure a long high load run?
Also, I'm running 32-bit OS (debian 5) on a 64 bit CPU... does it have sense to move my OS to 64 bit? ... will FS gain more preformance ?... I mean will FS perofomr drastically better 20%+ ?
Tihomir.
On Tue, Aug 25, 2009 at 11:00 AM, Giovanni Maruzzelli <[url=mailto:gmaruzz@celliax.org]gmaruzz@celliax.org (gmaruzz@celliax.org)[/url]> wrote:
Quote: | Maybe your load comes from disk access?
Try putting the sql and log directories on a ramdisk.
OTH,
-giovanni
On Tue, Aug 25, 2009 at 10:54 AM, Tihomir Culjaga<[url=mailto:tculjaga@gmail.com]tculjaga@gmail.com (tculjaga@gmail.com)[/url]> wrote:
Quote: | Hello,
i'm trying to use freeswitch as a redirecting server meaning FS has to
receive an INVITE and according to some rules it will redirect calls to
other destinations.
CALLING_USER FREESWITCH SOMEWHERE
INVITE ------------------------------->
<------------------------------ 100 Trying
<------------------------------ 302 Moved Temporary
ACK ------------------------------->
INVITE--------------------------------------------------------------------------------->
Well, wverything works well except i have perfromance issues .... on my HW
FS cannot do more than 40 CPS (INVITE answered by 302 Moved Temporary). When
i increase the rate, FS starts delaying 302 response. Right at 50 CPS i see
"calls" being build up in FS and the delay begining to grow.
When i observe the machine, load average is almost nothing (load average:
1.41, 0.61, 0.60) CPU never goes to 100%, and i see only one thread taking
most load... all others are just sitting there with 1-5 % CPU time.
This looks to me as FS handles 302 messages in a single thread?!?!
tculjaga@FS:/usr/local/freeswitch/conf/dialplan$ top -H
top - 10:41:37 up 167 days, 20:42, 3 users, load average: 1.41, 0.61, 0.60
Tasks: 83 total, 2 running, 81 sleeping, 0 stopped, 0 zombie
Cpu(s): 25.3%us, 1.5%sy, 0.0%ni, 30.3%id, 42.7%wa, 0.0%hi, 0.2%si,
0.0%st
Mem: 2074520k total, 571244k used, 1503276k free, 259604k buffers
Swap: 2650684k total, 3020k used, 2647664k free, 153868k cached
PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+
COMMAND
4814 root 20 0 34188 20m 3780 S 38 1.0 3:10.29
freeswitch
4800 root 20 0 34188 20m 3780 S 6 1.0 0:08.26
freeswitch
4798 root 20 0 34188 20m 3780 R 5 1.0 0:24.46
freeswitch
4787 root 20 0 34188 20m 3780 S 2 1.0 0:11.24
freeswitch
4794 root 20 0 34188 20m 3780 S 1 1.0 0:11.42
freeswitch
4803 root 20 0 34188 20m 3780 S 1 1.0 0:11.74
freeswitch
4788 root 20 0 34188 20m 3780 S 1 1.0 0:02.96
freeswitch
4804 root 20 0 34188 20m 3780 S 1 1.0 0:01.64
freeswitch
4807 root 20 0 34188 20m 3780 S 1 1.0 0:01.68
freeswitch
4811 root 20 0 34188 20m 3780 S 1 1.0 0:02.50 freeswitch
cat /proc/cpuinfo
processor : 0
vendor_id : GenuineIntel
cpu family : 6
model : 15
model name : Intel(R) Xeon(R) CPU 5140 @ 2.33GHz
stepping : 6
cpu MHz : 2333.560
cache size : 4096 KB
physical id : 0
siblings : 2
core id : 0
cpu cores : 2
apicid : 0
initial apicid : 0
fdiv_bug : no
hlt_bug : no
f00f_bug : no
coma_bug : no
fpu : yes
fpu_exception : yes
cpuid level : 10
wp : yes
flags : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca
cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe lm
constant_tsc arch_perfmon pebs bts pni monitor ds_cpl vmx est tm2 ssse3 cx16
xtpr dca lahf_lm
bogomips : 4670.78
clflush size : 64
power management:
processor : 1
vendor_id : GenuineIntel
cpu family : 6
model : 15
model name : Intel(R) Xeon(R) CPU 5140 @ 2.33GHz
stepping : 6
cpu MHz : 2333.560
cache size : 4096 KB
physical id : 0
siblings : 2
core id : 1
cpu cores : 2
apicid : 1
initial apicid : 1
fdiv_bug : no
hlt_bug : no
f00f_bug : no
coma_bug : no
fpu : yes
fpu_exception : yes
cpuid level : 10
wp : yes
flags : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca
cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe lm
constant_tsc arch_perfmon pebs bts pni monitor ds_cpl vmx est tm2 ssse3 cx16
xtpr dca lahf_lm
bogomips : 4666.82
clflush size : 64
power management:
uname -a
Linux l01sipindir1 2.6.26-1-686 #1 SMP Sat Jan 10 18:29:31 UTC 2009 i686
GNU/Linux
Of course, i've tuned the machine up
ulimit -c unlimited
ulimit -d unlimited
ulimit -f unlimited
ulimit -i unlimited
ulimit -n 999999
ulimit -q unlimited
ulimit -u unlimited
ulimit -v unlimited
ulimit -x unlimited
ulimit -s 240
ulimit -l unlimited
ulimit -a
Started FS with minimum modules but still 40 CPS seems to be the limit.
So, is there any way to improve performance?
Tihomir.
|
_______________________________________________
FreeSWITCH-users mailing list
[url=mailto:FreeSWITCH-users@lists.freeswitch.org]FreeSWITCH-users@lists.freeswitch.org (FreeSWITCH-users@lists.freeswitch.org)[/url]
http://lists.freeswitch.org/mailman/listinfo/freeswitch-users
UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users
http://www.freeswitch.org
|
_______________________________________________
FreeSWITCH-users mailing list
FreeSWITCH-users@lists.freeswitch.org (FreeSWITCH-users@lists.freeswitch.org)
http://lists.freeswitch.org/mailman/listinfo/freeswitch-users
UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users
http://www.freeswitch.org
|
|
|
Back to top |
|
|
gmaruzz at celliax.org Guest
|
Posted: Tue Aug 25, 2009 8:47 am Post subject: [Freeswitch-users] Freeswitch performance as a redirecting s |
|
|
Definitely go for 64 bit OS.
If you want to be safe and sure, go for CentOS 5.2 64bit. Is the one
used both for development and for heavy duty production.
Also Ubuntu 8.04 is good.
Other versions/distros are less used by the community.
Adding RAM and CPUs helps to scale up.
-gm
Sincerely,
Giovanni Maruzzelli
Cell : +39-347-2665618
On Tue, Aug 25, 2009 at 3:19 PM, Tihomir Culjaga<tculjaga@gmail.com> wrote:
Quote: | Hey Giovanni,
thanks for the tip... indeed the db files were heavily used regardless if i
started freeswitch with nosql option (freeswitch -nosql)... FS was not
writing anything into that files ... instead it was just accessing it....
This behaviour leads to a waste of 40% CPU time... waiting for other
processes (mainly disk access) to finish!!!
I moved freeswitch/db/ to a ramdisk and the performance got a boost to 140
CPS with a CPU load of 80%. I was keeping the machine for a while (20 - 30
minutes) on that rate when i sow CPU suddenly went to 100% and FS becoming
irresponsive .
What can be wrong?
What are the limits in CPU usage (50%, 60%, 70%, 80%...) we should not
cross?
What fine tuning do we need in order to asure a long high load run?
Also, I'm running 32-bit OS (debian 5) on a 64 bit CPU... does it have sense
to move my OS to 64 bit? ... will FS gain more preformance ?... I mean will
FS perofomr drastically better 20%+ ?
Tihomir.
On Tue, Aug 25, 2009 at 11:00 AM, Giovanni Maruzzelli <gmaruzz@celliax.org>
wrote:
Quote: |
Maybe your load comes from disk access?
Try putting the sql and log directories on a ramdisk.
OTH,
-giovanni
On Tue, Aug 25, 2009 at 10:54 AM, Tihomir Culjaga<tculjaga@gmail.com>
wrote:
Quote: | Hello,
i'm trying to use freeswitch as a redirecting server meaning FS has to
receive an INVITE and according to some rules it will redirect calls to
other destinations.
CALLING_USER FREESWITCH SOMEWHERE
INVITE ------------------------------->
<------------------------------ 100 Trying
<------------------------------ 302 Moved Temporary
ACK ------------------------------->
INVITE--------------------------------------------------------------------------------->
Well, wverything works well except i have perfromance issues .... on my
HW
FS cannot do more than 40 CPS (INVITE answered by 302 Moved Temporary).
When
i increase the rate, FS starts delaying 302 response. Right at 50 CPS i
see
"calls" being build up in FS and the delay begining to grow.
When i observe the machine, load average is almost nothing (load
average:
1.41, 0.61, 0.60) CPU never goes to 100%, and i see only one thread
taking
most load... all others are just sitting there with 1-5 % CPU time.
This looks to me as FS handles 302 messages in a single thread?!?!
tculjaga@FS:/usr/local/freeswitch/conf/dialplan$ top -H
top - 10:41:37 up 167 days, 20:42, 3 users, load average: 1.41, 0.61,
0.60
Tasks: 83 total, 2 running, 81 sleeping, 0 stopped, 0 zombie
Cpu(s): 25.3%us, 1.5%sy, 0.0%ni, 30.3%id, 42.7%wa, 0.0%hi, 0.2%si,
0.0%st
Mem: 2074520k total, 571244k used, 1503276k free, 259604k buffers
Swap: 2650684k total, 3020k used, 2647664k free, 153868k cached
PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+
COMMAND
4814 root 20 0 34188 20m 3780 S 38 1.0 3:10.29
freeswitch
4800 root 20 0 34188 20m 3780 S 6 1.0 0:08.26
freeswitch
4798 root 20 0 34188 20m 3780 R 5 1.0 0:24.46
freeswitch
4787 root 20 0 34188 20m 3780 S 2 1.0 0:11.24
freeswitch
4794 root 20 0 34188 20m 3780 S 1 1.0 0:11.42
freeswitch
4803 root 20 0 34188 20m 3780 S 1 1.0 0:11.74
freeswitch
4788 root 20 0 34188 20m 3780 S 1 1.0 0:02.96
freeswitch
4804 root 20 0 34188 20m 3780 S 1 1.0 0:01.64
freeswitch
4807 root 20 0 34188 20m 3780 S 1 1.0 0:01.68
freeswitch
4811 root 20 0 34188 20m 3780 S 1 1.0 0:02.50 freeswitch
cat /proc/cpuinfo
processor : 0
vendor_id : GenuineIntel
cpu family : 6
model : 15
model name : Intel(R) Xeon(R) CPU 5140 @ 2.33GHz
stepping : 6
cpu MHz : 2333.560
cache size : 4096 KB
physical id : 0
siblings : 2
core id : 0
cpu cores : 2
apicid : 0
initial apicid : 0
fdiv_bug : no
hlt_bug : no
f00f_bug : no
coma_bug : no
fpu : yes
fpu_exception : yes
cpuid level : 10
wp : yes
flags : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge
mca
cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe lm
constant_tsc arch_perfmon pebs bts pni monitor ds_cpl vmx est tm2 ssse3
cx16
xtpr dca lahf_lm
bogomips : 4670.78
clflush size : 64
power management:
processor : 1
vendor_id : GenuineIntel
cpu family : 6
model : 15
model name : Intel(R) Xeon(R) CPU 5140 @ 2.33GHz
stepping : 6
cpu MHz : 2333.560
cache size : 4096 KB
physical id : 0
siblings : 2
core id : 1
cpu cores : 2
apicid : 1
initial apicid : 1
fdiv_bug : no
hlt_bug : no
f00f_bug : no
coma_bug : no
fpu : yes
fpu_exception : yes
cpuid level : 10
wp : yes
flags : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge
mca
cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe lm
constant_tsc arch_perfmon pebs bts pni monitor ds_cpl vmx est tm2 ssse3
cx16
xtpr dca lahf_lm
bogomips : 4666.82
clflush size : 64
power management:
uname -a
Linux l01sipindir1 2.6.26-1-686 #1 SMP Sat Jan 10 18:29:31 UTC 2009 i686
GNU/Linux
Of course, i've tuned the machine up
ulimit -c unlimited
ulimit -d unlimited
ulimit -f unlimited
ulimit -i unlimited
ulimit -n 999999
ulimit -q unlimited
ulimit -u unlimited
ulimit -v unlimited
ulimit -x unlimited
ulimit -s 240
ulimit -l unlimited
ulimit -a
Started FS with minimum modules but still 40 CPS seems to be the limit.
So, is there any way to improve performance?
Tihomir.
_______________________________________________
FreeSWITCH-users mailing list
FreeSWITCH-users@lists.freeswitch.org
http://lists.freeswitch.org/mailman/listinfo/freeswitch-users
UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users
http://www.freeswitch.org
|
_______________________________________________
FreeSWITCH-users mailing list
FreeSWITCH-users@lists.freeswitch.org
http://lists.freeswitch.org/mailman/listinfo/freeswitch-users
UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users
http://www.freeswitch.org
|
_______________________________________________
FreeSWITCH-users mailing list
FreeSWITCH-users@lists.freeswitch.org
http://lists.freeswitch.org/mailman/listinfo/freeswitch-users
UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users
http://www.freeswitch.org
|
_______________________________________________
FreeSWITCH-users mailing list
FreeSWITCH-users@lists.freeswitch.org
http://lists.freeswitch.org/mailman/listinfo/freeswitch-users
UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users
http://www.freeswitch.org |
|
Back to top |
|
|
gmaruzz at celliax.org Guest
|
Posted: Tue Aug 25, 2009 8:52 am Post subject: [Freeswitch-users] Freeswitch performance as a redirecting s |
|
|
is a heavely multithreaded software, it benefits from number of CPUs
(or cores), RAM, and heavy duty kernel features (found in 64bit
kernels)
put all accesses on ramdisk, leave out the modules you don't use...
experiment, test, and find the best for your specific application/workload
test not only with sipp, but with real load too (often they're very different)
-gm
On Tue, Aug 25, 2009 at 3:42 PM, Tihomir Culjaga<tculjaga@gmail.com> wrote:
Quote: | thanks for the feedback... this is something im going to do tomorrow...
what about other things?
On Tue, Aug 25, 2009 at 3:39 PM, Jay Binks <jaybinks@gmail.com> wrote:
Quote: |
Everytime someone asks this , the resounding answer is use a 64bit os..
No question
Jay
On 25/08/2009, at 23:19, Tihomir Culjaga <tculjaga@gmail.com> wrote:
Hey Giovanni,
thanks for the tip... indeed the db files were heavily used regardless if
i started freeswitch with nosql option (freeswitch -nosql)... FS was not
writing anything into that files ... instead it was just accessing it....
This behaviour leads to a waste of 40% CPU time... waiting for other
processes (mainly disk access) to finish!!!
I moved freeswitch/db/ to a ramdisk and the performance got a boost to 140
CPS with a CPU load of 80%. I was keeping the machine for a while (20 - 30
minutes) on that rate when i sow CPU suddenly went to 100% and FS becoming
irresponsive .
What can be wrong?
What are the limits in CPU usage (50%, 60%, 70%, 80%...) we should not
cross?
What fine tuning do we need in order to asure a long high load run?
Also, I'm running 32-bit OS (debian 5) on a 64 bit CPU... does it have
sense to move my OS to 64 bit? ... will FS gain more preformance ?... I mean
will FS perofomr drastically better 20%+ ?
Tihomir.
On Tue, Aug 25, 2009 at 11:00 AM, Giovanni Maruzzelli
<gmaruzz@celliax.org> wrote:
Quote: |
Maybe your load comes from disk access?
Try putting the sql and log directories on a ramdisk.
OTH,
-giovanni
On Tue, Aug 25, 2009 at 10:54 AM, Tihomir Culjaga<tculjaga@gmail.com>
wrote:
Quote: | Hello,
i'm trying to use freeswitch as a redirecting server meaning FS has to
receive an INVITE and according to some rules it will redirect calls to
other destinations.
CALLING_USER FREESWITCH SOMEWHERE
INVITE ------------------------------->
<------------------------------ 100 Trying
<------------------------------ 302 Moved Temporary
ACK ------------------------------->
INVITE--------------------------------------------------------------------------------->
Well, wverything works well except i have perfromance issues .... on my
HW
FS cannot do more than 40 CPS (INVITE answered by 302 Moved Temporary).
When
i increase the rate, FS starts delaying 302 response. Right at 50 CPS i
see
"calls" being build up in FS and the delay begining to grow.
When i observe the machine, load average is almost nothing (load
average:
1.41, 0.61, 0.60) CPU never goes to 100%, and i see only one thread
taking
most load... all others are just sitting there with 1-5 % CPU time.
This looks to me as FS handles 302 messages in a single thread?!?!
tculjaga@FS:/usr/local/freeswitch/conf/dialplan$ top -H
top - 10:41:37 up 167 days, 20:42, 3 users, load average: 1.41, 0.61,
0.60
Tasks: 83 total, 2 running, 81 sleeping, 0 stopped, 0 zombie
Cpu(s): 25.3%us, 1.5%sy, 0.0%ni, 30.3%id, 42.7%wa, 0.0%hi, 0.2%si,
0.0%st
Mem: 2074520k total, 571244k used, 1503276k free, 259604k
buffers
Swap: 2650684k total, 3020k used, 2647664k free, 153868k cached
PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+
COMMAND
4814 root 20 0 34188 20m 3780 S 38 1.0 3:10.29
freeswitch
4800 root 20 0 34188 20m 3780 S 6 1.0 0:08.26
freeswitch
4798 root 20 0 34188 20m 3780 R 5 1.0 0:24.46
freeswitch
4787 root 20 0 34188 20m 3780 S 2 1.0 0:11.24
freeswitch
4794 root 20 0 34188 20m 3780 S 1 1.0 0:11.42
freeswitch
4803 root 20 0 34188 20m 3780 S 1 1.0 0:11.74
freeswitch
4788 root 20 0 34188 20m 3780 S 1 1.0 0:02.96
freeswitch
4804 root 20 0 34188 20m 3780 S 1 1.0 0:01.64
freeswitch
4807 root 20 0 34188 20m 3780 S 1 1.0 0:01.68
freeswitch
4811 root 20 0 34188 20m 3780 S 1 1.0 0:02.50 freeswitch
cat /proc/cpuinfo
processor : 0
vendor_id : GenuineIntel
cpu family : 6
model : 15
model name : Intel(R) Xeon(R) CPU 5140 @ 2.33GHz
stepping : 6
cpu MHz : 2333.560
cache size : 4096 KB
physical id : 0
siblings : 2
core id : 0
cpu cores : 2
apicid : 0
initial apicid : 0
fdiv_bug : no
hlt_bug : no
f00f_bug : no
coma_bug : no
fpu : yes
fpu_exception : yes
cpuid level : 10
wp : yes
flags : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge
mca
cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe lm
constant_tsc arch_perfmon pebs bts pni monitor ds_cpl vmx est tm2 ssse3
cx16
xtpr dca lahf_lm
bogomips : 4670.78
clflush size : 64
power management:
processor : 1
vendor_id : GenuineIntel
cpu family : 6
model : 15
model name : Intel(R) Xeon(R) CPU 5140 @ 2.33GHz
stepping : 6
cpu MHz : 2333.560
cache size : 4096 KB
physical id : 0
siblings : 2
core id : 1
cpu cores : 2
apicid : 1
initial apicid : 1
fdiv_bug : no
hlt_bug : no
f00f_bug : no
coma_bug : no
fpu : yes
fpu_exception : yes
cpuid level : 10
wp : yes
flags : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge
mca
cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe lm
constant_tsc arch_perfmon pebs bts pni monitor ds_cpl vmx est tm2 ssse3
cx16
xtpr dca lahf_lm
bogomips : 4666.82
clflush size : 64
power management:
uname -a
Linux l01sipindir1 2.6.26-1-686 #1 SMP Sat Jan 10 18:29:31 UTC 2009
i686
GNU/Linux
Of course, i've tuned the machine up
ulimit -c unlimited
ulimit -d unlimited
ulimit -f unlimited
ulimit -i unlimited
ulimit -n 999999
ulimit -q unlimited
ulimit -u unlimited
ulimit -v unlimited
ulimit -x unlimited
ulimit -s 240
ulimit -l unlimited
ulimit -a
Started FS with minimum modules but still 40 CPS seems to be the limit.
So, is there any way to improve performance?
Tihomir.
_______________________________________________
FreeSWITCH-users mailing list
FreeSWITCH-users@lists.freeswitch.org
http://lists.freeswitch.org/mailman/listinfo/freeswitch-users
UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users
http://www.freeswitch.org
|
_______________________________________________
FreeSWITCH-users mailing list
FreeSWITCH-users@lists.freeswitch.org
http://lists.freeswitch.org/mailman/listinfo/freeswitch-users
UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users
http://www.freeswitch.org
|
_______________________________________________
FreeSWITCH-users mailing list
FreeSWITCH-users@lists.freeswitch.org
http://lists.freeswitch.org/mailman/listinfo/freeswitch-users
UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users
http://www.freeswitch.org
_______________________________________________
FreeSWITCH-users mailing list
FreeSWITCH-users@lists.freeswitch.org
http://lists.freeswitch.org/mailman/listinfo/freeswitch-users
UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users
http://www.freeswitch.org
|
_______________________________________________
FreeSWITCH-users mailing list
FreeSWITCH-users@lists.freeswitch.org
http://lists.freeswitch.org/mailman/listinfo/freeswitch-users
UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users
http://www.freeswitch.org
|
_______________________________________________
FreeSWITCH-users mailing list
FreeSWITCH-users@lists.freeswitch.org
http://lists.freeswitch.org/mailman/listinfo/freeswitch-users
UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users
http://www.freeswitch.org |
|
Back to top |
|
|
mike at jerris.com Guest
|
Posted: Tue Aug 25, 2009 10:36 am Post subject: [Freeswitch-users] Freeswitch performance as a redirecting s |
|
|
Actually in this case, we are bound to one thread in sofia.
Mike
On Aug 25, 2009, at 9:47 AM, Giovanni Maruzzelli wrote:
Quote: | is a heavely multithreaded software, it benefits from number of CPUs
(or cores), RAM, and heavy duty kernel features (found in 64bit
kernels)
put all accesses on ramdisk, leave out the modules you don't use...
experiment, test, and find the best for your specific application/
workload
test not only with sipp, but with real load too (often they're very
different)
-gm
|
_______________________________________________
FreeSWITCH-users mailing list
FreeSWITCH-users@lists.freeswitch.org
http://lists.freeswitch.org/mailman/listinfo/freeswitch-users
UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users
http://www.freeswitch.org |
|
Back to top |
|
|
msc at freeswitch.org Guest
|
Posted: Tue Aug 25, 2009 11:20 am Post subject: [Freeswitch-users] Freeswitch performance as a redirecting s |
|
|
Quote: | Also, I'm running 32-bit OS (debian 5) on a 64 bit CPU... does it have sense to move my OS to 64 bit? ... will FS gain more preformance ?... I mean will FS perofomr drastically better 20%+ ?
|
If you really want to get on the same page as the developers then get the 64bit CentOS 5.3 loaded on your machine. Also, we've seen reports of really weird things happening when people run 32bit OS on 64bit hardware. That's a big no-no.
Let us know how it goes.
-MC |
|
Back to top |
|
|
tculjaga at gmail.com Guest
|
Posted: Tue Aug 25, 2009 11:42 am Post subject: [Freeswitch-users] Freeswitch performance as a redirecting s |
|
|
Exactly... the scenario i use seems operating on a single thread... why is that ? can it be changed?
T.
On Tue, Aug 25, 2009 at 5:31 PM, Michael Jerris <mike@jerris.com (mike@jerris.com)> wrote:
Quote: | Actually in this case, we are bound to one thread in sofia.
Mike
On Aug 25, 2009, at 9:47 AM, Giovanni Maruzzelli wrote:
Quote: | is a heavely multithreaded software, it benefits from number of CPUs
(or cores), RAM, and heavy duty kernel features (found in 64bit
kernels)
put all accesses on ramdisk, leave out the modules you don't use...
experiment, test, and find the best for your specific application/
workload
test not only with sipp, but with real load too (often they're very
different)
-gm
|
_______________________________________________
FreeSWITCH-users mailing list
FreeSWITCH-users@lists.freeswitch.org (FreeSWITCH-users@lists.freeswitch.org)
http://lists.freeswitch.org/mailman/listinfo/freeswitch-users
UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users
http://www.freeswitch.org
|
|
|
Back to top |
|
|
tculjaga at gmail.com Guest
|
Posted: Tue Aug 25, 2009 11:44 am Post subject: [Freeswitch-users] Freeswitch performance as a redirecting s |
|
|
well ... this is something we are going to change tomorrow.... of course will let you posted.
T.
On Tue, Aug 25, 2009 at 6:11 PM, Michael Collins <msc@freeswitch.org (msc@freeswitch.org)> wrote:
|
|
Back to top |
|
|
anthony.minessale at g... Guest
|
Posted: Tue Aug 25, 2009 12:16 pm Post subject: [Freeswitch-users] Freeswitch performance as a redirecting s |
|
|
You also should put your extension first in your dialplan ahead of the default extensions which match every call and do a lot of db access for
record keeping etc.
The single thread in sofia is part of their concurrency model.
The single thread acts as a scheduler and indicates to us that an invite is requested
we react to that and spawn a new thread for that call as soon as the data is received.
like I said above you are probably hitting all the default stuff to save that last dialed number etc
that exists in the default config.
sigh, we go through this every new guy doing load testing.
On Tue, Aug 25, 2009 at 11:34 AM, Tihomir Culjaga <tculjaga@gmail.com (tculjaga@gmail.com)> wrote:
Quote: | Exactly... the scenario i use seems operating on a single thread... why is that ? can it be changed?
T.
On Tue, Aug 25, 2009 at 5:31 PM, Michael Jerris <mike@jerris.com (mike@jerris.com)> wrote:
Quote: | Actually in this case, we are bound to one thread in sofia.
Mike
On Aug 25, 2009, at 9:47 AM, Giovanni Maruzzelli wrote:
Quote: | is a heavely multithreaded software, it benefits from number of CPUs
(or cores), RAM, and heavy duty kernel features (found in 64bit
kernels)
put all accesses on ramdisk, leave out the modules you don't use...
experiment, test, and find the best for your specific application/
workload
test not only with sipp, but with real load too (often they're very
different)
-gm
|
_______________________________________________
FreeSWITCH-users mailing list
FreeSWITCH-users@lists.freeswitch.org (FreeSWITCH-users@lists.freeswitch.org)
http://lists.freeswitch.org/mailman/listinfo/freeswitch-users
UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users
http://www.freeswitch.org
|
_______________________________________________
FreeSWITCH-users mailing list
FreeSWITCH-users@lists.freeswitch.org (FreeSWITCH-users@lists.freeswitch.org)
http://lists.freeswitch.org/mailman/listinfo/freeswitch-users
UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users
http://www.freeswitch.org
|
--
Anthony Minessale II
FreeSWITCH http://www.freeswitch.org/
ClueCon http://www.cluecon.com/
Twitter: http://twitter.com/FreeSWITCH_wire
AIM: anthm
MSN:anthony_minessale@hotmail.com ([email]MSN%3Aanthony_minessale@hotmail.com[/email])
GTALK/JABBER/PAYPAL:anthony.minessale@gmail.com ([email]PAYPAL%3Aanthony.minessale@gmail.com[/email])
IRC: irc.freenode.net #freeswitch
FreeSWITCH Developer Conference
sip:888@conference.freeswitch.org ([email]sip%3A888@conference.freeswitch.org[/email])
iax:guest@conference.freeswitch.org/888
googletalk:conf+888@conference.freeswitch.org ([email]googletalk%3Aconf%2B888@conference.freeswitch.org[/email])
pstn:213-799-1400 |
|
Back to top |
|
|
mike at jerris.com Guest
|
Posted: Tue Aug 25, 2009 12:19 pm Post subject: [Freeswitch-users] Freeswitch performance as a redirecting s |
|
|
This requires invasive changes in the sofia-sip stack to get thread-pooling working again. I am sure they would accept patches if you can provide some that fully address any issues that may come up from adding this such as race conditions.
Mike
On Aug 25, 2009, at 12:34 PM, Tihomir Culjaga wrote:
Quote: | Exactly... the scenario i use seems operating on a single thread... why is that ? can it be changed?
T.
On Tue, Aug 25, 2009 at 5:31 PM, Michael Jerris <mike@jerris.com (mike@jerris.com)> wrote:
Quote: | Actually in this case, we are bound to one thread in sofia.
Mike
On Aug 25, 2009, at 9:47 AM, Giovanni Maruzzelli wrote:
Quote: | is a heavely multithreaded software, it benefits from number of CPUs
(or cores), RAM, and heavy duty kernel features (found in 64bit
kernels)
put all accesses on ramdisk, leave out the modules you don't use...
experiment, test, and find the best for your specific application/
workload
test not only with sipp, but with real load too (often they're very
different)
-gm
|
|
|
|
|
Back to top |
|
|
tculjaga at gmail.com Guest
|
Posted: Tue Aug 25, 2009 12:39 pm Post subject: [Freeswitch-users] Freeswitch performance as a redirecting s |
|
|
Of course i removed everytihng from teh dialplan except my extension
when exactly do you react and bring up a new thread ? ... is it on INVITE or on 1st 1xx response ?
i beleive i can have several lets call it SIP interfaces ... on different ports 5060, 5070, 5080 ... every interface will have it's own sip profile.
does it mean i will have one thread per profile?
T.
On Tue, Aug 25, 2009 at 7:10 PM, Anthony Minessale <anthony.minessale@gmail.com (anthony.minessale@gmail.com)> wrote:
Quote: | You also should put your extension first in your dialplan ahead of the default extensions which match every call and do a lot of db access for
record keeping etc.
The single thread in sofia is part of their concurrency model.
The single thread acts as a scheduler and indicates to us that an invite is requested
we react to that and spawn a new thread for that call as soon as the data is received.
like I said above you are probably hitting all the default stuff to save that last dialed number etc
that exists in the default config.
sigh, we go through this every new guy doing load testing.
On Tue, Aug 25, 2009 at 11:34 AM, Tihomir Culjaga <tculjaga@gmail.com (tculjaga@gmail.com)> wrote:
Quote: | Exactly... the scenario i use seems operating on a single thread... why is that ? can it be changed?
T.
On Tue, Aug 25, 2009 at 5:31 PM, Michael Jerris <mike@jerris.com (mike@jerris.com)> wrote:
Quote: | Actually in this case, we are bound to one thread in sofia.
Mike
On Aug 25, 2009, at 9:47 AM, Giovanni Maruzzelli wrote:
Quote: | is a heavely multithreaded software, it benefits from number of CPUs
(or cores), RAM, and heavy duty kernel features (found in 64bit
kernels)
put all accesses on ramdisk, leave out the modules you don't use...
experiment, test, and find the best for your specific application/
workload
test not only with sipp, but with real load too (often they're very
different)
-gm
|
_______________________________________________
FreeSWITCH-users mailing list
FreeSWITCH-users@lists.freeswitch.org (FreeSWITCH-users@lists.freeswitch.org)
http://lists.freeswitch.org/mailman/listinfo/freeswitch-users
UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users
http://www.freeswitch.org
|
_______________________________________________
FreeSWITCH-users mailing list
FreeSWITCH-users@lists.freeswitch.org (FreeSWITCH-users@lists.freeswitch.org)
http://lists.freeswitch.org/mailman/listinfo/freeswitch-users
UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users
http://www.freeswitch.org
|
--
Anthony Minessale II
FreeSWITCH http://www.freeswitch.org/
ClueCon http://www.cluecon.com/
Twitter: http://twitter.com/FreeSWITCH_wire
AIM: anthm
MSN:anthony_minessale@hotmail.com ([email]MSN%3Aanthony_minessale@hotmail.com[/email])
GTALK/JABBER/PAYPAL:anthony.minessale@gmail.com ([email]PAYPAL%3Aanthony.minessale@gmail.com[/email])
IRC: irc.freenode.net #freeswitch
FreeSWITCH Developer Conference
sip:888@conference.freeswitch.org ([email]sip%3A888@conference.freeswitch.org[/email])
iax:guest@conference.freeswitch.org/888
googletalk:conf+888@conference.freeswitch.org ([email]googletalk%3Aconf%2B888@conference.freeswitch.org[/email])
pstn:213-799-1400
_______________________________________________
FreeSWITCH-users mailing list
FreeSWITCH-users@lists.freeswitch.org (FreeSWITCH-users@lists.freeswitch.org)
http://lists.freeswitch.org/mailman/listinfo/freeswitch-users
UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users
http://www.freeswitch.org
|
|
|
Back to top |
|
|
msc at freeswitch.org Guest
|
Posted: Tue Aug 25, 2009 12:46 pm Post subject: [Freeswitch-users] Freeswitch performance as a redirecting s |
|
|
On Tue, Aug 25, 2009 at 10:31 AM, Tihomir Culjaga <tculjaga@gmail.com (tculjaga@gmail.com)> wrote:
Quote: | Of course i removed everytihng from teh dialplan except my extension
when exactly do you react and bring up a new thread ? ... is it on INVITE or on 1st 1xx response ?
i beleive i can have several lets call it SIP interfaces ... on different ports 5060, 5070, 5080 ... every interface will have it's own sip profile.
does it mean i will have one thread per profile? | Yes. |
|
Back to top |
|
|
tculjaga at gmail.com Guest
|
Posted: Tue Aug 25, 2009 12:56 pm Post subject: [Freeswitch-users] Freeswitch performance as a redirecting s |
|
|
Hi Giovanny,
thanks for your help,
everything that heavyly accesses the disk is on ramdisk now...
hopefully will get some real traffic pretty soon...
On Tue, Aug 25, 2009 at 3:47 PM, Giovanni Maruzzelli <gmaruzz@celliax.org (gmaruzz@celliax.org)> wrote:
Quote: | is a heavely multithreaded software, it benefits from number of CPUs
(or cores), RAM, and heavy duty kernel features (found in 64bit
kernels)
put all accesses on ramdisk, leave out the modules you don't use...
experiment, test, and find the best for your specific application/workload
test not only with sipp, but with real load too (often they're very different)
-gm
On Tue, Aug 25, 2009 at 3:42 PM, Tihomir Culjaga<tculjaga@gmail.com (tculjaga@gmail.com)> wrote:
Quote: | thanks for the feedback... this is something im going to do tomorrow...
what about other things?
On Tue, Aug 25, 2009 at 3:39 PM, Jay Binks <jaybinks@gmail.com (jaybinks@gmail.com)> wrote:
Quote: |
Everytime someone asks this , the resounding answer is use a 64bit os..
No question
Jay
On 25/08/2009, at 23:19, Tihomir Culjaga <tculjaga@gmail.com (tculjaga@gmail.com)> wrote:
Hey Giovanni,
thanks for the tip... indeed the db files were heavily used regardless if
i started freeswitch with nosql option (freeswitch -nosql)... FS was not
writing anything into that files ... instead it was just accessing it....
This behaviour leads to a waste of 40% CPU time... waiting for other
processes (mainly disk access) to finish!!!
I moved freeswitch/db/ to a ramdisk and the performance got a boost to 140
CPS with a CPU load of 80%. I was keeping the machine for a while (20 - 30
minutes) on that rate when i sow CPU suddenly went to 100% and FS becoming
irresponsive .
What can be wrong?
What are the limits in CPU usage (50%, 60%, 70%, 80%...) we should not
cross?
What fine tuning do we need in order to asure a long high load run?
Also, I'm running 32-bit OS (debian 5) on a 64 bit CPU... does it have
sense to move my OS to 64 bit? ... will FS gain more preformance ?... I mean
will FS perofomr drastically better 20%+ ?
Tihomir.
On Tue, Aug 25, 2009 at 11:00 AM, Giovanni Maruzzelli
<gmaruzz@celliax.org (gmaruzz@celliax.org)> wrote:
Quote: |
Maybe your load comes from disk access?
Try putting the sql and log directories on a ramdisk.
OTH,
-giovanni
On Tue, Aug 25, 2009 at 10:54 AM, Tihomir Culjaga<tculjaga@gmail.com (tculjaga@gmail.com)>
wrote:
Quote: | Hello,
i'm trying to use freeswitch as a redirecting server meaning FS has to
receive an INVITE and according to some rules it will redirect calls to
other destinations.
CALLING_USER FREESWITCH SOMEWHERE
INVITE ------------------------------->
<------------------------------ 100 Trying
<------------------------------ 302 Moved Temporary
ACK ------------------------------->
INVITE--------------------------------------------------------------------------------->
Well, wverything works well except i have perfromance issues .... on my
HW
FS cannot do more than 40 CPS (INVITE answered by 302 Moved Temporary).
When
i increase the rate, FS starts delaying 302 response. Right at 50 CPS i
see
"calls" being build up in FS and the delay begining to grow.
When i observe the machine, load average is almost nothing (load
average:
1.41, 0.61, 0.60) CPU never goes to 100%, and i see only one thread
taking
most load... all others are just sitting there with 1-5 % CPU time.
This looks to me as FS handles 302 messages in a single thread?!?!
tculjaga@FS:/usr/local/freeswitch/conf/dialplan$ top -H
top - 10:41:37 up 167 days, 20:42, 3 users, load average: 1.41, 0.61,
0.60
Tasks: 83 total, 2 running, 81 sleeping, 0 stopped, 0 zombie
Cpu(s): 25.3%us, 1.5%sy, 0.0%ni, 30.3%id, 42.7%wa, 0.0%hi, 0.2%si,
0.0%st
Mem: 2074520k total, 571244k used, 1503276k free, 259604k
buffers
Swap: 2650684k total, 3020k used, 2647664k free, 153868k cached
PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+
COMMAND
4814 root 20 0 34188 20m 3780 S 38 1.0 3:10.29
freeswitch
4800 root 20 0 34188 20m 3780 S 6 1.0 0:08.26
freeswitch
4798 root 20 0 34188 20m 3780 R 5 1.0 0:24.46
freeswitch
4787 root 20 0 34188 20m 3780 S 2 1.0 0:11.24
freeswitch
4794 root 20 0 34188 20m 3780 S 1 1.0 0:11.42
freeswitch
4803 root 20 0 34188 20m 3780 S 1 1.0 0:11.74
freeswitch
4788 root 20 0 34188 20m 3780 S 1 1.0 0:02.96
freeswitch
4804 root 20 0 34188 20m 3780 S 1 1.0 0:01.64
freeswitch
4807 root 20 0 34188 20m 3780 S 1 1.0 0:01.68
freeswitch
4811 root 20 0 34188 20m 3780 S 1 1.0 0:02.50 freeswitch
cat /proc/cpuinfo
processor : 0
vendor_id : GenuineIntel
cpu family : 6
model : 15
model name : Intel(R) Xeon(R) CPU 5140 @ 2.33GHz
stepping : 6
cpu MHz : 2333.560
cache size : 4096 KB
physical id : 0
siblings : 2
core id : 0
cpu cores : 2
apicid : 0
initial apicid : 0
fdiv_bug : no
hlt_bug : no
f00f_bug : no
coma_bug : no
fpu : yes
fpu_exception : yes
cpuid level : 10
wp : yes
flags : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge
mca
cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe lm
constant_tsc arch_perfmon pebs bts pni monitor ds_cpl vmx est tm2 ssse3
cx16
xtpr dca lahf_lm
bogomips : 4670.78
clflush size : 64
power management:
processor : 1
vendor_id : GenuineIntel
cpu family : 6
model : 15
model name : Intel(R) Xeon(R) CPU 5140 @ 2.33GHz
stepping : 6
cpu MHz : 2333.560
cache size : 4096 KB
physical id : 0
siblings : 2
core id : 1
cpu cores : 2
apicid : 1
initial apicid : 1
fdiv_bug : no
hlt_bug : no
f00f_bug : no
coma_bug : no
fpu : yes
fpu_exception : yes
cpuid level : 10
wp : yes
flags : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge
mca
cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe lm
constant_tsc arch_perfmon pebs bts pni monitor ds_cpl vmx est tm2 ssse3
cx16
xtpr dca lahf_lm
bogomips : 4666.82
clflush size : 64
power management:
uname -a
Linux l01sipindir1 2.6.26-1-686 #1 SMP Sat Jan 10 18:29:31 UTC 2009
i686
GNU/Linux
Of course, i've tuned the machine up
ulimit -c unlimited
ulimit -d unlimited
ulimit -f unlimited
ulimit -i unlimited
ulimit -n 999999
ulimit -q unlimited
ulimit -u unlimited
ulimit -v unlimited
ulimit -x unlimited
ulimit -s 240
ulimit -l unlimited
ulimit -a
Started FS with minimum modules but still 40 CPS seems to be the limit.
So, is there any way to improve performance?
Tihomir.
_______________________________________________
FreeSWITCH-users mailing list
FreeSWITCH-users@lists.freeswitch.org (FreeSWITCH-users@lists.freeswitch.org)
http://lists.freeswitch.org/mailman/listinfo/freeswitch-users
UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users
http://www.freeswitch.org
|
_______________________________________________
FreeSWITCH-users mailing list
FreeSWITCH-users@lists.freeswitch.org (FreeSWITCH-users@lists.freeswitch.org)
http://lists.freeswitch.org/mailman/listinfo/freeswitch-users
UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users
http://www.freeswitch.org
|
_______________________________________________
FreeSWITCH-users mailing list
FreeSWITCH-users@lists.freeswitch.org (FreeSWITCH-users@lists.freeswitch.org)
http://lists.freeswitch.org/mailman/listinfo/freeswitch-users
UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users
http://www.freeswitch.org
_______________________________________________
FreeSWITCH-users mailing list
FreeSWITCH-users@lists.freeswitch.org (FreeSWITCH-users@lists.freeswitch.org)
http://lists.freeswitch.org/mailman/listinfo/freeswitch-users
UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users
http://www.freeswitch.org
|
_______________________________________________
FreeSWITCH-users mailing list
FreeSWITCH-users@lists.freeswitch.org (FreeSWITCH-users@lists.freeswitch.org)
http://lists.freeswitch.org/mailman/listinfo/freeswitch-users
UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users
http://www.freeswitch.org
|
_______________________________________________
FreeSWITCH-users mailing list
FreeSWITCH-users@lists.freeswitch.org (FreeSWITCH-users@lists.freeswitch.org)
http://lists.freeswitch.org/mailman/listinfo/freeswitch-users
UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users
http://www.freeswitch.org
|
|
|
Back to top |
|
|
|
|
|
You cannot post new topics in this forum You cannot reply to topics in this forum You cannot edit your posts in this forum You cannot delete your posts in this forum You cannot vote in polls in this forum
|
Powered by phpBB © 2001, 2005 phpBB Group
|