For some time now we've been running an older kernel, 2.6.16.60. When
we tried to upgrade, first going to 2.6.27.19 and then to 2.6.32.1 and
2.6.33.1 we noticed that latencies increased. At first we noticed it
by doing network tests via udpping, netperf, etc. We made some
tweaks, and were able to get network latency to within 1 to 2
microseconds of where we were previously on 2.6.16.60. Then we did
some more testing, and noticed that system latency also seems higher.
We've done our tests on identical hardware servers, same NICs,
connected through same network gear. Basically, we've tried to keep
everything identical except the kernel versions, and we are unable to
achieve the same performance for system latency on the newer kernels,
despite adjusting various kernel settings and recompiling.
The latency differences are about 15 microseconds per transaction.
At this point, I don't know what else to try. I haven't played around
with the /proc/sys/kernel/sched_* paramaters under the newer kernels
yet. Have tried changing pre-emption modes with little effect, in
fact, voluntary preemption seems to be peforming the best for us.
At this time the realtime patch isn't really an option for us to
consider, at least not yet.
Any suggestions? Is this a known issue when upgrading to more recent
kernel versions?
Thanks,
Taylor
Le jeudi 01 avril 2010 à 14:12 -0500, Taylor Lewick a écrit :
> For some time now we've been running an older kernel, 2.6.16.60. When
> we tried to upgrade, first going to 2.6.27.19 and then to 2.6.32.1 and
> 2.6.33.1 we noticed that latencies increased. At first we noticed it
> by doing network tests via udpping, netperf, etc. We made some
> tweaks, and were able to get network latency to within 1 to 2
> microseconds of where we were previously on 2.6.16.60. Then we did
> some more testing, and noticed that system latency also seems higher.
>
> We've done our tests on identical hardware servers, same NICs,
> connected through same network gear. Basically, we've tried to keep
> everything identical except the kernel versions, and we are unable to
> achieve the same performance for system latency on the newer kernels,
> despite adjusting various kernel settings and recompiling.
>
> The latency differences are about 15 microseconds per transaction.
>
> At this point, I don't know what else to try. I haven't played around
> with the /proc/sys/kernel/sched_* paramaters under the newer kernels
> yet. Have tried changing pre-emption modes with little effect, in
> fact, voluntary preemption seems to be peforming the best for us.
>
> At this time the realtime patch isn't really an option for us to
> consider, at least not yet.
>
> Any suggestions? Is this a known issue when upgrading to more recent
> kernel versions?
>
Hi Taylor
Well, this is bit difficult to generically answer to your generic
question. 15 us more latency per transaction seems pretty bad.
Some inputs would be nice, describing your workload and
software/hardware architecture.
lspci
cat /proc/cpuinfo
cat /proc/interrupts
dmesg
ethtool -S eth0
ethtool -c eth0
Okay. I will get this info out to the list Monday. Briefly, I'm
using identical hardware (server), identical NICs, same drivers,
connected to same switch, and using udpping, hackbench, and an
internall written app to test latency. Without exception the
evolution has looked like the following.
2.6.16.60 latencies for system and network are fast. Meaning
hackbench and udpping win, and win by quite a bit.
2.6.27.19 was awful. 2.6.32.1 and 2.6.331. were better for networking
(with some tweaks, i.e. disable netfilter, etc), and I was able to get
networking latencies to within 1-3 microseconds of 2.6.16.60
latencies, but the hackbench results are still pretty bad.
Again, I'll post numbers and more detailed hardware info on Monday
when I'm back at office...
On Thu, Apr 1, 2010 at 4:19 PM, Eric Dumazet <[email protected]> wrote:
> Le jeudi 01 avril 2010 ? 14:12 -0500, Taylor Lewick a ?crit :
>> For some time now we've been running an older kernel, 2.6.16.60. ?When
>> we tried to upgrade, first going to 2.6.27.19 and then to 2.6.32.1 and
>> 2.6.33.1 we noticed that latencies increased. ?At first we noticed it
>> by doing network tests via udpping, netperf, etc. ?We made some
>> tweaks, and were able to get network latency to within 1 to 2
>> microseconds of where we were previously on 2.6.16.60. ?Then we did
>> some more testing, and noticed that system latency also seems higher.
>>
>> We've done our tests on identical hardware servers, same NICs,
>> connected through same network gear. ?Basically, we've tried to keep
>> everything identical except the kernel versions, and we are unable to
>> achieve the same performance for system latency on the newer kernels,
>> despite adjusting various kernel settings and recompiling.
>>
>> The latency differences are about 15 microseconds per transaction.
>>
>> At this point, I don't know what else to try. ?I haven't played around
>> with the /proc/sys/kernel/sched_* paramaters under the newer kernels
>> yet. ?Have tried changing pre-emption modes with little effect, in
>> fact, voluntary preemption seems to be peforming the best for us.
>>
>> At this time the realtime patch isn't really an option for us to
>> consider, at least not yet.
>>
>> Any suggestions? ?Is this a known issue when upgrading to more recent
>> kernel versions?
>>
>
> Hi Taylor
>
> Well, this is bit difficult to generically answer to your generic
> question. 15 us more latency per transaction seems pretty bad.
>
> Some inputs would be nice, describing your workload and
> software/hardware architecture.
>
> lspci
> cat /proc/cpuinfo
> cat /proc/interrupts
> dmesg
> ethtool -S eth0
> ethtool -c eth0
>
>
>
>
Okay, don't know what to officially file this under, as a regression
with regards to performance or what, but here is the data. Again,
I've noticed system and network latency appear to have worsened with
later kernel versions.
I was turned onto this problem via the following links:
http://www.kernel.org/pub/linux/kernel/people/christoph/ols2009/ols-2009-paper.pdf
and http://kerneltrap.org/mailarchive/linux-netdev/2009/4/16/5491284
So I set up a test on two servers with Identical hardware, servers,
nics, etc, and used hackbench, udpping, and an internally written app
to compare latency.
Here are just the hackbench results with just the averages across a 5
runs for two different hackbench tests. The 2.6.16 and 2.6.27 kernels
as set up were configured with voluntary preemption, and 250 HZ, so I
just repeated that initially for 2.6.33.1 test. I also tested no
preemption at same HZ setting of 250.
I ran 2.6.16.60 on one server, and the other kernel versions on
another server. These tests are repeatable across different servers,
as in I verified I
don't have a bad server.
Kernel Version HB1 (25 process 300) HB2 (100 process 300)
2.6.16.60 .5402 1.8946
2.6.27.19 .619 2.6268
2.6.32.3-voluntary .5636 2.3484
2.6.33.1-voluntary .5404 2.2872
2.6.33.1-nopreempt .5606 2.3466
So 2.6.16.60 is fast, 2.6.27.19 is slow, and 2.6.33.1 with voluntary
preemption is the next best, but results didn't hold up well as
Hackbench tests used larger numbers of groups., for example, 2.6.16.60
and 2.6.33.1-voluntary were basically the same for HB1, but that
didn't hold when hackebnch tests used more groups.
At this point, I'm looking for ideas in kernel build to tweak, but I'm
not a developer. So SLAB vs SLUB, sparse vs dense IRQ numbering, etc.
Running a -rt kernel isn't an option at this time. I did test that as
well, and latencies were quite a bit worse, but I wasn't adjusting
code to take advantage of a real time OS.
I can make some changes or repeat tests.
Below is some hardware comparisons betweent the two machines.
Differences I noticed was more interrupts and CPU flags on later
kernel version.
HostA 2.6.16.60
cat /proc/interrupts
CPU0 CPU1 CPU2 CPU3 CPU4 CPU5
CPU6 CPU7
0: 108509762 0 0 0 0 0
0 0 IO-APIC-edge timer
8: 1 0 0 0 0 0
0 0 IO-APIC-edge rtc
9: 0 0 0 0 0 0
0 0 IO-APIC-level acpi
58: 305 0 5157735 220 2980100 5927
1187 0 IO-APIC-level libata
162: 0 0 0 0 0 0
0 0 IO-APIC-level uhci_hcd:usb1
170: 0 0 0 0 0 0
0 0 IO-APIC-level uhci_hcd:usb2
177: 6326 0 229018 0 283720 35597
367 0 IO-APIC-level megasas
178: 122 0 1784 1103 3531 20
1457 0 IO-APIC-level uhci_hcd:usb3, ehci_hcd:usb6
186: 0 0 0 0 0 0
0 0 IO-APIC-level uhci_hcd:usb4
194: 22 0 0 0 0 0
0 0 IO-APIC-level ehci_hcd:usb5
210: 1790109 577 0 0 0 0
0 0 PCI-MSI-X eth4-0
218: 233811 93 0 0 0 0
0 0 PCI-MSI-X eth4-1
NMI: 0 0 0 0 0 0
0 0
LOC: 108509683 108509662 108509637 108509614 108509588 108509566
108509541 108509516
ERR: 7
MIS: 0
lspci
00:00.0 Host bridge: Intel Corporation QuickPath Architecture I/O Hub
to ESI Port (rev 13)
00:01.0 PCI bridge: Intel Corporation QuickPath Architecture I/O Hub
PCI Express Root Port 1 (rev 13)
00:03.0 PCI bridge: Intel Corporation QuickPath Architecture I/O Hub
PCI Express Root Port 3 (rev 13)
00:07.0 PCI bridge: Intel Corporation QuickPath Architecture I/O Hub
PCI Express Root Port 7 (rev 13)
00:09.0 PCI bridge: Intel Corporation QuickPath Architecture I/O Hub
PCI Express Root Port 9 (rev 13)
00:14.0 PIC: Intel Corporation QuickPath Architecture I/O Hub System
Management Registers (rev 13)
00:14.1 PIC: Intel Corporation QuickPath Architecture I/O Hub GPIO and
Scratch Pad Registers (rev 13)
00:14.2 PIC: Intel Corporation QuickPath Architecture I/O Hub Control
Status and RAS Registers (rev 13)
00:16.0 System peripheral: Intel Corporation DMA Engine (rev 13)
00:16.1 System peripheral: Intel Corporation DMA Engine (rev 13)
00:16.2 System peripheral: Intel Corporation DMA Engine (rev 13)
00:16.3 System peripheral: Intel Corporation DMA Engine (rev 13)
00:16.4 System peripheral: Intel Corporation DMA Engine (rev 13)
00:16.5 System peripheral: Intel Corporation DMA Engine (rev 13)
00:16.6 System peripheral: Intel Corporation DMA Engine (rev 13)
00:16.7 System peripheral: Intel Corporation DMA Engine (rev 13)
00:1a.0 USB Controller: Intel Corporation 82801I (ICH9 Family) USB
UHCI Controller #4 (rev 02)
00:1a.1 USB Controller: Intel Corporation 82801I (ICH9 Family) USB
UHCI Controller #5 (rev 02)
00:1a.7 USB Controller: Intel Corporation 82801I (ICH9 Family) USB2
EHCI Controller #2 (rev 02)
00:1c.0 PCI bridge: Intel Corporation 82801I (ICH9 Family) PCI Express
Port 1 (rev 02)
00:1d.0 USB Controller: Intel Corporation 82801I (ICH9 Family) USB
UHCI Controller #1 (rev 02)
00:1d.1 USB Controller: Intel Corporation 82801I (ICH9 Family) USB
UHCI Controller #2 (rev 02)
00:1d.7 USB Controller: Intel Corporation 82801I (ICH9 Family) USB2
EHCI Controller #1 (rev 02)
00:1e.0 PCI bridge: Intel Corporation 82801 PCI Bridge (rev 92)
00:1f.0 ISA bridge: Intel Corporation 82801IB (ICH9) LPC Interface
Controller (rev 02)
00:1f.2 IDE interface: Intel Corporation 82801IB (ICH9) 2 port SATA
IDE Controller (rev 02)
03:00.0 RAID bus controller: LSI Logic / Symbios Logic MegaRAID SAS
1078 (rev 04)
04:00.0 PCI bridge: Integrated Device Technology, Inc. Unknown device
8018 (rev 0e)
05:02.0 PCI bridge: Integrated Device Technology, Inc. Unknown device
8018 (rev 0e)
05:04.0 PCI bridge: Integrated Device Technology, Inc. Unknown device
8018 (rev 0e)
06:00.0 Ethernet controller: Intel Corporation 82575GB Gigabit Network
Connection (rev 02)
06:00.1 Ethernet controller: Intel Corporation 82575GB Gigabit Network
Connection (rev 02)
07:00.0 Ethernet controller: Intel Corporation 82575GB Gigabit Network
Connection (rev 02)
07:00.1 Ethernet controller: Intel Corporation 82575GB Gigabit Network
Connection (rev 02)
08:00.0 Ethernet controller: Solarflare Communications Unknown device
0710 (rev 02)
09:03.0 VGA compatible controller: Matrox Graphics, Inc. Unknown
device 0532 (rev 0a)
cat /proc/cpuinfo (just showing first CPU for brevity)
processor : 0
vendor_id : GenuineIntel
cpu family : 6
model : 26
model name : Intel(R) Xeon(R) CPU X5570 @ 2.93GHz
stepping : 5
cpu MHz : 2926.090
cache size : 8192 KB
physical id : 1
siblings : 4
core id : 0
cpu cores : 4
fpu : yes
fpu_exception : yes
cpuid level : 11
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 syscall
nx rdtscp lm constant_tsc pni monitor d
s_cpl vmx est tm2 cx16 xtpr dca popcnt lahf_lm
bogomips : 5857.34
clflush size : 64
cache_alignment : 64
address sizes : 40 bits physical, 48 bits virtual
power management:
ethtool -c eth4
Coalesce parameters for eth4:
Adaptive RX: on TX: off
stats-block-usecs: 0
sample-interval: 0
pkt-rate-low: 0
pkt-rate-high: 0
rx-usecs: 0
rx-frames: 0
rx-usecs-irq: 60
rx-frames-irq: 0
tx-usecs: 0
tx-frames: 0
tx-usecs-irq: 0
tx-frames-irq: 0
rx-usecs-low: 0
rx-frame-low: 0
tx-usecs-low: 0
tx-frame-low: 0
rx-usecs-high: 0
rx-frame-high: 0
tx-usecs-high: 0
tx-frame-high: 0
HostB 2.6.33.1
CPU0 CPU1 CPU2 CPU3 CPU4 CPU5
CPU6 CPU7
0: 8637 0 0 0 0
0 0 0 IO-APIC-edge timer
1: 2 0 0 0 0
0 0 0 IO-APIC-edge i8042
3: 2 0 0 0 0
0 0 0 IO-APIC-edge
4: 2 0 0 0 0
0 0 0 IO-APIC-edge
8: 1 0 0 0 0
0 0 0 IO-APIC-edge rtc0
9: 0 0 0 0 0
0 0 0 IO-APIC-fasteoi acpi
12: 4 0 0 0 0
0 0 0 IO-APIC-edge i8042
16: 7434 683 0 0 0
0 0 0 IO-APIC-fasteoi megasas
17: 0 0 0 0 0
0 0 0 IO-APIC-fasteoi uhci_hcd:usb3
18: 0 0 0 0 0
0 0 0 IO-APIC-fasteoi uhci_hcd:usb4
19: 23 0 0 0 0
0 0 0 IO-APIC-fasteoi ehci_hcd:usb1
20: 0 0 0 0 0
0 0 0 IO-APIC-fasteoi uhci_hcd:usb6
21: 129 0 15 0 0
0 0 0 IO-APIC-fasteoi ehci_hcd:usb2,
uhci_hcd:usb5
23: 369 0 0 0 0
0 0 0 IO-APIC-fasteoi ata_piix
67: 2346 731 0 0 0
0 0 0 PCI-MSI-edge eth4-0
68: 1809 404 0 0 0
0 0 0 PCI-MSI-edge eth4-1
NMI: 0 0 0 0 0
0 0 0 Non-maskable interrupts
LOC: 33071 38348 47397 23246 15715
11065 9004 10391 Local timer interrupts
SPU: 0 0 0 0 0
0 0 0 Spurious interrupts
PMI: 0 0 0 0 0
0 0 0 Performance monitoring interrupts
PND: 0 0 0 0 0
0 0 0 Performance pending work
RES: 2490 2124 4187 4974 1724
5548 1892 2871 Rescheduling interrupts
CAL: 497 2166 141 115 133
144 140 144 Function call interrupts
TLB: 243 244 928 945 289
187 134 93 TLB shootdowns
TRM: 0 0 0 0 0
0 0 0 Thermal event interrupts
THR: 0 0 0 0 0
0 0 0 Threshold APIC interrupts
MCE: 0 0 0 0 0
0 0 0 Machine check exceptions
MCP: 2 2 2 2 2
2 2 2 Machine check polls
ERR: 7
MIS: 0
lspci
00:00.0 Host bridge: Intel Corporation X58 I/O Hub to ESI Port (rev 13)
00:01.0 PCI bridge: Intel Corporation X58 I/O Hub PCI Express Root
Port 1 (rev 13)
00:03.0 PCI bridge: Intel Corporation X58 I/O Hub PCI Express Root
Port 3 (rev 13)
00:07.0 PCI bridge: Intel Corporation X58 I/O Hub PCI Express Root
Port 7 (rev 13)
00:09.0 PCI bridge: Intel Corporation X58 I/O Hub PCI Express Root
Port 9 (rev 13)
00:14.0 PIC: Intel Corporation X58 I/O Hub System Management Registers (rev 13)
00:14.1 PIC: Intel Corporation X58 I/O Hub GPIO and Scratch Pad
Registers (rev 13)
00:14.2 PIC: Intel Corporation X58 I/O Hub Control Status and RAS
Registers (rev 13)
00:1a.0 USB Controller: Intel Corporation 82801I (ICH9 Family) USB
UHCI Controller #4 (rev 02)
00:1a.1 USB Controller: Intel Corporation 82801I (ICH9 Family) USB
UHCI Controller #5 (rev 02)
00:1a.7 USB Controller: Intel Corporation 82801I (ICH9 Family) USB2
EHCI Controller #2 (rev 02)
00:1c.0 PCI bridge: Intel Corporation 82801I (ICH9 Family) PCI Express
Port 1 (rev 02)
00:1d.0 USB Controller: Intel Corporation 82801I (ICH9 Family) USB
UHCI Controller #1 (rev 02)
00:1d.1 USB Controller: Intel Corporation 82801I (ICH9 Family) USB
UHCI Controller #2 (rev 02)
00:1d.7 USB Controller: Intel Corporation 82801I (ICH9 Family) USB2
EHCI Controller #1 (rev 02)
00:1e.0 PCI bridge: Intel Corporation 82801 PCI Bridge (rev 92)
00:1f.0 ISA bridge: Intel Corporation 82801IB (ICH9) LPC Interface
Controller (rev 02)
00:1f.2 IDE interface: Intel Corporation 82801IB (ICH9) 2 port SATA
IDE Controller (rev 02)
03:00.0 RAID bus controller: LSI Logic / Symbios Logic MegaRAID SAS
1078 (rev 04)
04:00.0 PCI bridge: Integrated Device Technology, Inc. PES12N3A PCI
Express Switch (rev 0e)
05:02.0 PCI bridge: Integrated Device Technology, Inc. PES12N3A PCI
Express Switch (rev 0e)
05:04.0 PCI bridge: Integrated Device Technology, Inc. PES12N3A PCI
Express Switch (rev 0e)
06:00.0 Ethernet controller: Intel Corporation 82575GB Gigabit Network
Connection (rev 02)
06:00.1 Ethernet controller: Intel Corporation 82575GB Gigabit Network
Connection (rev 02)
07:00.0 Ethernet controller: Intel Corporation 82575GB Gigabit Network
Connection (rev 02)
07:00.1 Ethernet controller: Intel Corporation 82575GB Gigabit Network
Connection (rev 02)
08:00.0 Ethernet controller: Solarflare Communications SFC4000 rev B
[Solarstorm] (rev 02)
09:03.0 VGA compatible controller: Matrox Graphics, Inc. MGA G200eW
WPCM450 (rev 0a)
cat /proc/cpuinfo (just showing first CPU for brevity)
processor : 0
vendor_id : GenuineIntel
cpu family : 6
model : 26
model name : Intel(R) Xeon(R) CPU X5570 @ 2.93GHz
stepping : 5
cpu MHz : 2925.888
cache size : 8192 KB
physical id : 1
siblings : 4
core id : 0
cpu cores : 4
apicid : 16
initial apicid : 16
fpu : yes
fpu_exception : yes
cpuid level : 11
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
syscall nx rdtscp lm constant_tsc arch_perfmon pebs bt
s rep_good xtopology nonstop_tsc aperfmperf pni dtes64 monitor ds_cpl
vmx est tm2 ssse3 cx16 xtpr pdcm dca sse4_1 sse4_2 popcnt lahf_lm ida
tpr_shadow vnmi flexpriority ept vpid
bogomips : 5851.77
clflush size : 64
cache_alignment : 64
address sizes : 40 bits physical, 48 bits virtual
power management:
ethtool -c eth4
Coalesce parameters for eth4:
Adaptive RX: on TX: off
stats-block-usecs: 0
sample-interval: 0
pkt-rate-low: 0
pkt-rate-high: 0
rx-usecs: 0
rx-frames: 0
rx-usecs-irq: 60
rx-frames-irq: 0
tx-usecs: 0
tx-frames: 0
tx-usecs-irq: 0
tx-frames-irq: 0
rx-usecs-low: 0
rx-frame-low: 0
tx-usecs-low: 0
tx-frame-low: 0
rx-usecs-high: 0
rx-frame-high: 0
tx-usecs-high: 0
tx-frame-high: 0
On Thu, Apr 1, 2010 at 8:53 PM, Taylor Lewick <[email protected]> wrote:
> Okay. ?I will get this info out to the list Monday. ?Briefly, I'm
> using identical hardware (server), identical NICs, same drivers,
> connected to same switch, and using udpping, hackbench, and an
> internall written app to test latency. ?Without exception the
> evolution has looked like the following.
>
> 2.6.16.60 latencies for system and network are fast. ?Meaning
> hackbench and udpping win, and win by quite a bit.
>
> 2.6.27.19 was awful. ?2.6.32.1 and 2.6.331. were better for networking
> (with some tweaks, i.e. disable netfilter, etc), and I was able to get
> networking latencies to within 1-3 microseconds of 2.6.16.60
> latencies, but the hackbench results are still pretty bad.
>
> Again, I'll post numbers and more detailed hardware info on Monday
> when I'm back at office...
>
> On Thu, Apr 1, 2010 at 4:19 PM, Eric Dumazet <[email protected]> wrote:
>> Le jeudi 01 avril 2010 ? 14:12 -0500, Taylor Lewick a ?crit :
>>> For some time now we've been running an older kernel, 2.6.16.60. ?When
>>> we tried to upgrade, first going to 2.6.27.19 and then to 2.6.32.1 and
>>> 2.6.33.1 we noticed that latencies increased. ?At first we noticed it
>>> by doing network tests via udpping, netperf, etc. ?We made some
>>> tweaks, and were able to get network latency to within 1 to 2
>>> microseconds of where we were previously on 2.6.16.60. ?Then we did
>>> some more testing, and noticed that system latency also seems higher.
>>>
>>> We've done our tests on identical hardware servers, same NICs,
>>> connected through same network gear. ?Basically, we've tried to keep
>>> everything identical except the kernel versions, and we are unable to
>>> achieve the same performance for system latency on the newer kernels,
>>> despite adjusting various kernel settings and recompiling.
>>>
>>> The latency differences are about 15 microseconds per transaction.
>>>
>>> At this point, I don't know what else to try. ?I haven't played around
>>> with the /proc/sys/kernel/sched_* paramaters under the newer kernels
>>> yet. ?Have tried changing pre-emption modes with little effect, in
>>> fact, voluntary preemption seems to be peforming the best for us.
>>>
>>> At this time the realtime patch isn't really an option for us to
>>> consider, at least not yet.
>>>
>>> Any suggestions? ?Is this a known issue when upgrading to more recent
>>> kernel versions?
>>>
>>
>> Hi Taylor
>>
>> Well, this is bit difficult to generically answer to your generic
>> question. 15 us more latency per transaction seems pretty bad.
>>
>> Some inputs would be nice, describing your workload and
>> software/hardware architecture.
>>
>> lspci
>> cat /proc/cpuinfo
>> cat /proc/interrupts
>> dmesg
>> ethtool -S eth0
>> ethtool -c eth0
>>
>>
>>
>>
>
On Mon, Apr 5, 2010 at 12:34 PM, Taylor Lewick <[email protected]> wrote:
> Okay, don't know what to officially file this under, as a regression
> with regards to performance or what, but here is the data. Again,
> I've noticed system and network latency appear to have worsened with
> later kernel versions.
>
> I was turned onto this problem via the following links:
> http://www.kernel.org/pub/linux/kernel/people/christoph/ols2009/ols-2009-paper.pdf
> and http://kerneltrap.org/mailarchive/linux-netdev/2009/4/16/5491284
>
> So I set up a test on two servers with Identical hardware, servers,
> nics, etc, and used hackbench, udpping, and an internally written app
> to compare latency.
>
> Here are just the hackbench results with just the averages across a 5
> runs for two different hackbench tests. The 2.6.16 and 2.6.27 kernels
> as set up were configured with voluntary preemption, and 250 HZ, so I
> just repeated that initially for 2.6.33.1 test. I also tested no
> preemption at same HZ setting of 250.
>
> I ran 2.6.16.60 on one server, and the other kernel versions on
> another server. These tests are repeatable across different servers,
> as in I verified I
> don't have a bad server.
>
> Kernel Version HB1 (25 process 300) HB2 (100 process 300)
> 2.6.16.60 .5402 1.8946
> 2.6.27.19 .619 2.6268
> 2.6.32.3-voluntary .5636 2.3484
> 2.6.33.1-voluntary .5404 2.2872
> 2.6.33.1-nopreempt .5606 2.3466
>
> So 2.6.16.60 is fast, 2.6.27.19 is slow, and 2.6.33.1 with voluntary
> preemption is the next best, but results didn't hold up well as
> Hackbench tests used larger numbers of groups., for example, 2.6.16.60
> and 2.6.33.1-voluntary were basically the same for HB1, but that
> didn't hold when hackebnch tests used more groups.
>
> At this point, I'm looking for ideas in kernel build to tweak, but I'm
> not a developer. So SLAB vs SLUB, sparse vs dense IRQ numbering, etc.
> Running a -rt kernel isn't an option at this time. I did test that as
> well, and latencies were quite a bit worse, but I wasn't adjusting
> code to take advantage of a real time OS.
>
> I can make some changes or repeat tests.
>
> Below is some hardware comparisons betweent the two machines.
> Differences I noticed was more interrupts and CPU flags on later
> kernel version.
>
> HostA 2.6.16.60
> cat /proc/interrupts
> CPU0 CPU1 CPU2 CPU3 CPU4 CPU5
> CPU6 CPU7
> 0: 108509762 0 0 0 0 0
> 0 0 IO-APIC-edge timer
> 8: 1 0 0 0 0 0
> 0 0 IO-APIC-edge rtc
> 9: 0 0 0 0 0 0
> 0 0 IO-APIC-level acpi
> 58: 305 0 5157735 220 2980100 5927
> 1187 0 IO-APIC-level libata
> 162: 0 0 0 0 0 0
> 0 0 IO-APIC-level uhci_hcd:usb1
> 170: 0 0 0 0 0 0
> 0 0 IO-APIC-level uhci_hcd:usb2
> 177: 6326 0 229018 0 283720 35597
> 367 0 IO-APIC-level megasas
> 178: 122 0 1784 1103 3531 20
> 1457 0 IO-APIC-level uhci_hcd:usb3, ehci_hcd:usb6
> 186: 0 0 0 0 0 0
> 0 0 IO-APIC-level uhci_hcd:usb4
> 194: 22 0 0 0 0 0
> 0 0 IO-APIC-level ehci_hcd:usb5
> 210: 1790109 577 0 0 0 0
> 0 0 PCI-MSI-X eth4-0
> 218: 233811 93 0 0 0 0
> 0 0 PCI-MSI-X eth4-1
> NMI: 0 0 0 0 0 0
> 0 0
> LOC: 108509683 108509662 108509637 108509614 108509588 108509566
> 108509541 108509516
> ERR: 7
> MIS: 0
>
> lspci
> 00:00.0 Host bridge: Intel Corporation QuickPath Architecture I/O Hub
> to ESI Port (rev 13)
> 00:01.0 PCI bridge: Intel Corporation QuickPath Architecture I/O Hub
> PCI Express Root Port 1 (rev 13)
> 00:03.0 PCI bridge: Intel Corporation QuickPath Architecture I/O Hub
> PCI Express Root Port 3 (rev 13)
> 00:07.0 PCI bridge: Intel Corporation QuickPath Architecture I/O Hub
> PCI Express Root Port 7 (rev 13)
> 00:09.0 PCI bridge: Intel Corporation QuickPath Architecture I/O Hub
> PCI Express Root Port 9 (rev 13)
> 00:14.0 PIC: Intel Corporation QuickPath Architecture I/O Hub System
> Management Registers (rev 13)
> 00:14.1 PIC: Intel Corporation QuickPath Architecture I/O Hub GPIO and
> Scratch Pad Registers (rev 13)
> 00:14.2 PIC: Intel Corporation QuickPath Architecture I/O Hub Control
> Status and RAS Registers (rev 13)
> 00:16.0 System peripheral: Intel Corporation DMA Engine (rev 13)
> 00:16.1 System peripheral: Intel Corporation DMA Engine (rev 13)
> 00:16.2 System peripheral: Intel Corporation DMA Engine (rev 13)
> 00:16.3 System peripheral: Intel Corporation DMA Engine (rev 13)
> 00:16.4 System peripheral: Intel Corporation DMA Engine (rev 13)
> 00:16.5 System peripheral: Intel Corporation DMA Engine (rev 13)
> 00:16.6 System peripheral: Intel Corporation DMA Engine (rev 13)
> 00:16.7 System peripheral: Intel Corporation DMA Engine (rev 13)
> 00:1a.0 USB Controller: Intel Corporation 82801I (ICH9 Family) USB
> UHCI Controller #4 (rev 02)
> 00:1a.1 USB Controller: Intel Corporation 82801I (ICH9 Family) USB
> UHCI Controller #5 (rev 02)
> 00:1a.7 USB Controller: Intel Corporation 82801I (ICH9 Family) USB2
> EHCI Controller #2 (rev 02)
> 00:1c.0 PCI bridge: Intel Corporation 82801I (ICH9 Family) PCI Express
> Port 1 (rev 02)
> 00:1d.0 USB Controller: Intel Corporation 82801I (ICH9 Family) USB
> UHCI Controller #1 (rev 02)
> 00:1d.1 USB Controller: Intel Corporation 82801I (ICH9 Family) USB
> UHCI Controller #2 (rev 02)
> 00:1d.7 USB Controller: Intel Corporation 82801I (ICH9 Family) USB2
> EHCI Controller #1 (rev 02)
> 00:1e.0 PCI bridge: Intel Corporation 82801 PCI Bridge (rev 92)
> 00:1f.0 ISA bridge: Intel Corporation 82801IB (ICH9) LPC Interface
> Controller (rev 02)
> 00:1f.2 IDE interface: Intel Corporation 82801IB (ICH9) 2 port SATA
> IDE Controller (rev 02)
> 03:00.0 RAID bus controller: LSI Logic / Symbios Logic MegaRAID SAS
> 1078 (rev 04)
> 04:00.0 PCI bridge: Integrated Device Technology, Inc. Unknown device
> 8018 (rev 0e)
> 05:02.0 PCI bridge: Integrated Device Technology, Inc. Unknown device
> 8018 (rev 0e)
> 05:04.0 PCI bridge: Integrated Device Technology, Inc. Unknown device
> 8018 (rev 0e)
> 06:00.0 Ethernet controller: Intel Corporation 82575GB Gigabit Network
> Connection (rev 02)
> 06:00.1 Ethernet controller: Intel Corporation 82575GB Gigabit Network
> Connection (rev 02)
> 07:00.0 Ethernet controller: Intel Corporation 82575GB Gigabit Network
> Connection (rev 02)
> 07:00.1 Ethernet controller: Intel Corporation 82575GB Gigabit Network
> Connection (rev 02)
> 08:00.0 Ethernet controller: Solarflare Communications Unknown device
> 0710 (rev 02)
> 09:03.0 VGA compatible controller: Matrox Graphics, Inc. Unknown
> device 0532 (rev 0a)
>
> cat /proc/cpuinfo (just showing first CPU for brevity)
> processor : 0
> vendor_id : GenuineIntel
> cpu family : 6
> model : 26
> model name : Intel(R) Xeon(R) CPU X5570 @ 2.93GHz
> stepping : 5
> cpu MHz : 2926.090
> cache size : 8192 KB
> physical id : 1
> siblings : 4
> core id : 0
> cpu cores : 4
> fpu : yes
> fpu_exception : yes
> cpuid level : 11
> 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 syscall
> nx rdtscp lm constant_tsc pni monitor d
> s_cpl vmx est tm2 cx16 xtpr dca popcnt lahf_lm
> bogomips : 5857.34
> clflush size : 64
> cache_alignment : 64
> address sizes : 40 bits physical, 48 bits virtual
> power management:
>
> ethtool -c eth4
> Coalesce parameters for eth4:
> Adaptive RX: on TX: off
> stats-block-usecs: 0
> sample-interval: 0
> pkt-rate-low: 0
> pkt-rate-high: 0
>
> rx-usecs: 0
> rx-frames: 0
> rx-usecs-irq: 60
> rx-frames-irq: 0
>
> tx-usecs: 0
> tx-frames: 0
> tx-usecs-irq: 0
> tx-frames-irq: 0
>
> rx-usecs-low: 0
> rx-frame-low: 0
> tx-usecs-low: 0
> tx-frame-low: 0
>
> rx-usecs-high: 0
> rx-frame-high: 0
> tx-usecs-high: 0
> tx-frame-high: 0
>
>
> HostB 2.6.33.1
> CPU0 CPU1 CPU2 CPU3 CPU4 CPU5
> CPU6 CPU7
> 0: 8637 0 0 0 0
> 0 0 0 IO-APIC-edge timer
> 1: 2 0 0 0 0
> 0 0 0 IO-APIC-edge i8042
> 3: 2 0 0 0 0
> 0 0 0 IO-APIC-edge
> 4: 2 0 0 0 0
> 0 0 0 IO-APIC-edge
> 8: 1 0 0 0 0
> 0 0 0 IO-APIC-edge rtc0
> 9: 0 0 0 0 0
> 0 0 0 IO-APIC-fasteoi acpi
> 12: 4 0 0 0 0
> 0 0 0 IO-APIC-edge i8042
> 16: 7434 683 0 0 0
> 0 0 0 IO-APIC-fasteoi megasas
> 17: 0 0 0 0 0
> 0 0 0 IO-APIC-fasteoi uhci_hcd:usb3
> 18: 0 0 0 0 0
> 0 0 0 IO-APIC-fasteoi uhci_hcd:usb4
> 19: 23 0 0 0 0
> 0 0 0 IO-APIC-fasteoi ehci_hcd:usb1
> 20: 0 0 0 0 0
> 0 0 0 IO-APIC-fasteoi uhci_hcd:usb6
> 21: 129 0 15 0 0
> 0 0 0 IO-APIC-fasteoi ehci_hcd:usb2,
> uhci_hcd:usb5
> 23: 369 0 0 0 0
> 0 0 0 IO-APIC-fasteoi ata_piix
> 67: 2346 731 0 0 0
> 0 0 0 PCI-MSI-edge eth4-0
> 68: 1809 404 0 0 0
> 0 0 0 PCI-MSI-edge eth4-1
> NMI: 0 0 0 0 0
> 0 0 0 Non-maskable interrupts
> LOC: 33071 38348 47397 23246 15715
> 11065 9004 10391 Local timer interrupts
> SPU: 0 0 0 0 0
> 0 0 0 Spurious interrupts
> PMI: 0 0 0 0 0
> 0 0 0 Performance monitoring interrupts
> PND: 0 0 0 0 0
> 0 0 0 Performance pending work
> RES: 2490 2124 4187 4974 1724
> 5548 1892 2871 Rescheduling interrupts
> CAL: 497 2166 141 115 133
> 144 140 144 Function call interrupts
> TLB: 243 244 928 945 289
> 187 134 93 TLB shootdowns
> TRM: 0 0 0 0 0
> 0 0 0 Thermal event interrupts
> THR: 0 0 0 0 0
> 0 0 0 Threshold APIC interrupts
> MCE: 0 0 0 0 0
> 0 0 0 Machine check exceptions
> MCP: 2 2 2 2 2
> 2 2 2 Machine check polls
> ERR: 7
> MIS: 0
>
> lspci
> 00:00.0 Host bridge: Intel Corporation X58 I/O Hub to ESI Port (rev 13)
> 00:01.0 PCI bridge: Intel Corporation X58 I/O Hub PCI Express Root
> Port 1 (rev 13)
> 00:03.0 PCI bridge: Intel Corporation X58 I/O Hub PCI Express Root
> Port 3 (rev 13)
> 00:07.0 PCI bridge: Intel Corporation X58 I/O Hub PCI Express Root
> Port 7 (rev 13)
> 00:09.0 PCI bridge: Intel Corporation X58 I/O Hub PCI Express Root
> Port 9 (rev 13)
> 00:14.0 PIC: Intel Corporation X58 I/O Hub System Management Registers (rev 13)
> 00:14.1 PIC: Intel Corporation X58 I/O Hub GPIO and Scratch Pad
> Registers (rev 13)
> 00:14.2 PIC: Intel Corporation X58 I/O Hub Control Status and RAS
> Registers (rev 13)
> 00:1a.0 USB Controller: Intel Corporation 82801I (ICH9 Family) USB
> UHCI Controller #4 (rev 02)
> 00:1a.1 USB Controller: Intel Corporation 82801I (ICH9 Family) USB
> UHCI Controller #5 (rev 02)
> 00:1a.7 USB Controller: Intel Corporation 82801I (ICH9 Family) USB2
> EHCI Controller #2 (rev 02)
> 00:1c.0 PCI bridge: Intel Corporation 82801I (ICH9 Family) PCI Express
> Port 1 (rev 02)
> 00:1d.0 USB Controller: Intel Corporation 82801I (ICH9 Family) USB
> UHCI Controller #1 (rev 02)
> 00:1d.1 USB Controller: Intel Corporation 82801I (ICH9 Family) USB
> UHCI Controller #2 (rev 02)
> 00:1d.7 USB Controller: Intel Corporation 82801I (ICH9 Family) USB2
> EHCI Controller #1 (rev 02)
> 00:1e.0 PCI bridge: Intel Corporation 82801 PCI Bridge (rev 92)
> 00:1f.0 ISA bridge: Intel Corporation 82801IB (ICH9) LPC Interface
> Controller (rev 02)
> 00:1f.2 IDE interface: Intel Corporation 82801IB (ICH9) 2 port SATA
> IDE Controller (rev 02)
> 03:00.0 RAID bus controller: LSI Logic / Symbios Logic MegaRAID SAS
> 1078 (rev 04)
> 04:00.0 PCI bridge: Integrated Device Technology, Inc. PES12N3A PCI
> Express Switch (rev 0e)
> 05:02.0 PCI bridge: Integrated Device Technology, Inc. PES12N3A PCI
> Express Switch (rev 0e)
> 05:04.0 PCI bridge: Integrated Device Technology, Inc. PES12N3A PCI
> Express Switch (rev 0e)
> 06:00.0 Ethernet controller: Intel Corporation 82575GB Gigabit Network
> Connection (rev 02)
> 06:00.1 Ethernet controller: Intel Corporation 82575GB Gigabit Network
> Connection (rev 02)
> 07:00.0 Ethernet controller: Intel Corporation 82575GB Gigabit Network
> Connection (rev 02)
> 07:00.1 Ethernet controller: Intel Corporation 82575GB Gigabit Network
> Connection (rev 02)
> 08:00.0 Ethernet controller: Solarflare Communications SFC4000 rev B
> [Solarstorm] (rev 02)
> 09:03.0 VGA compatible controller: Matrox Graphics, Inc. MGA G200eW
> WPCM450 (rev 0a)
>
> cat /proc/cpuinfo (just showing first CPU for brevity)
> processor : 0
> vendor_id : GenuineIntel
> cpu family : 6
> model : 26
> model name : Intel(R) Xeon(R) CPU X5570 @ 2.93GHz
> stepping : 5
> cpu MHz : 2925.888
> cache size : 8192 KB
> physical id : 1
> siblings : 4
> core id : 0
> cpu cores : 4
> apicid : 16
> initial apicid : 16
> fpu : yes
> fpu_exception : yes
> cpuid level : 11
> 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
> syscall nx rdtscp lm constant_tsc arch_perfmon pebs bt
> s rep_good xtopology nonstop_tsc aperfmperf pni dtes64 monitor ds_cpl
> vmx est tm2 ssse3 cx16 xtpr pdcm dca sse4_1 sse4_2 popcnt lahf_lm ida
> tpr_shadow vnmi flexpriority ept vpid
> bogomips : 5851.77
> clflush size : 64
> cache_alignment : 64
> address sizes : 40 bits physical, 48 bits virtual
> power management:
>
> ethtool -c eth4
> Coalesce parameters for eth4:
> Adaptive RX: on TX: off
> stats-block-usecs: 0
> sample-interval: 0
> pkt-rate-low: 0
> pkt-rate-high: 0
>
> rx-usecs: 0
> rx-frames: 0
> rx-usecs-irq: 60
> rx-frames-irq: 0
>
> tx-usecs: 0
> tx-frames: 0
> tx-usecs-irq: 0
> tx-frames-irq: 0
>
> rx-usecs-low: 0
> rx-frame-low: 0
> tx-usecs-low: 0
> tx-frame-low: 0
>
> rx-usecs-high: 0
> rx-frame-high: 0
> tx-usecs-high: 0
> tx-frame-high: 0
>
>
>
> On Thu, Apr 1, 2010 at 8:53 PM, Taylor Lewick <[email protected]> wrote:
>> Okay. I will get this info out to the list Monday. Briefly, I'm
>> using identical hardware (server), identical NICs, same drivers,
>> connected to same switch, and using udpping, hackbench, and an
>> internall written app to test latency. Without exception the
>> evolution has looked like the following.
>>
>> 2.6.16.60 latencies for system and network are fast. Meaning
>> hackbench and udpping win, and win by quite a bit.
>>
>> 2.6.27.19 was awful. 2.6.32.1 and 2.6.331. were better for networking
>> (with some tweaks, i.e. disable netfilter, etc), and I was able to get
>> networking latencies to within 1-3 microseconds of 2.6.16.60
>> latencies, but the hackbench results are still pretty bad.
>>
>> Again, I'll post numbers and more detailed hardware info on Monday
>> when I'm back at office...
>>
>> On Thu, Apr 1, 2010 at 4:19 PM, Eric Dumazet <[email protected]> wrote:
>>> Le jeudi 01 avril 2010 à 14:12 -0500, Taylor Lewick a écrit :
>>>> For some time now we've been running an older kernel, 2.6.16.60. When
>>>> we tried to upgrade, first going to 2.6.27.19 and then to 2.6.32.1 and
>>>> 2.6.33.1 we noticed that latencies increased. At first we noticed it
>>>> by doing network tests via udpping, netperf, etc. We made some
>>>> tweaks, and were able to get network latency to within 1 to 2
>>>> microseconds of where we were previously on 2.6.16.60. Then we did
>>>> some more testing, and noticed that system latency also seems higher.
>>>>
>>>> We've done our tests on identical hardware servers, same NICs,
>>>> connected through same network gear. Basically, we've tried to keep
>>>> everything identical except the kernel versions, and we are unable to
>>>> achieve the same performance for system latency on the newer kernels,
>>>> despite adjusting various kernel settings and recompiling.
>>>>
>>>> The latency differences are about 15 microseconds per transaction.
>>>>
>>>> At this point, I don't know what else to try. I haven't played around
>>>> with the /proc/sys/kernel/sched_* paramaters under the newer kernels
>>>> yet. Have tried changing pre-emption modes with little effect, in
>>>> fact, voluntary preemption seems to be peforming the best for us.
>>>>
>>>> At this time the realtime patch isn't really an option for us to
>>>> consider, at least not yet.
>>>>
>>>> Any suggestions? Is this a known issue when upgrading to more recent
>>>> kernel versions?
>>>>
>>>
>>> Hi Taylor
>>>
>>> Well, this is bit difficult to generically answer to your generic
>>> question. 15 us more latency per transaction seems pretty bad.
>>>
>>> Some inputs would be nice, describing your workload and
>>> software/hardware architecture.
>>>
>>> lspci
>>> cat /proc/cpuinfo
>>> cat /proc/interrupts
>>> dmesg
>>> ethtool -S eth0
>>> ethtool -c eth0
>>>
>>>
>>>
>>>
>>
>
Just want to ack you here, I upgraded a 2.6.18 kernel to 2.6.33.1 on a
shipping product and the performance(hackbench, latency, cpu
usage,etc) is a lot worse on the same hardware platform. We tried
2.6.27 before and it's also bad. I'm tring various CONFIG options and
so far nothing really helped. I'm using the RT patch.
Xianghua