设置进程亲和 - numactl 或 taskset - retrieve or set a process’s CPU affinity (affect SYSTEMTAP TIME)
背景
在使用systemtap监控进程或者内核的运行状况时, 我们会发现使用systemtap和不使用systemtap时, 某些操作的运行时间差别会比较大. 这是因为systemtap本身带来的开销导致的, 那么如何减少这部分开销呢?
可选的方法较多, 例如精简systemtap, 减少systemtap的触发事件的范围, 简化handler的逻辑等等.
除此之外, 还有其他的方法, 例如设置CPU亲和, Linux进程使用哪个CPU资源是由内核进行调度的.
被跟踪进程的亲和与stap运行的进程亲和分开, 可以有效的减少stap的影响吗. 实际上影响甚微, 但是有效果.
其次发现CentOS 6.4 x64系统下面, 使用CPU 0的话, 性能会远远不如使用其他任何CPU. (可能与OS调度有关，软中断集中在CPU 0了)
所以改亲和的话, 只要不使用0号CPU就可以了.
例如 :
服务器上有8个CPU
pg93@db-172-16-3-150-> numactl --show
policy: default
preferred node: current
physcpubind: 0 1 2 3 4 5 6 7
cpubind: 0
nodebind: 0
membind: 0
pg93@db-172-16-3-150-> cat /proc/cpuinfo
processor : 0
vendor_id : GenuineIntel
cpu family : 6
model : 26
model name : Intel(R) Xeon(R) CPU E5504 @ 2.00GHz
stepping : 5
cpu MHz : 1595.992
cache size : 4096 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 bts 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 dts tpr_shadow vnmi flexpriority ept vpid
bogomips : 3191.98
clflush size : 64
cache_alignment : 64
address sizes : 40 bits physical, 48 bits virtual
power management:
processor : 1
vendor_id : GenuineIntel
cpu family : 6
model : 26
model name : Intel(R) Xeon(R) CPU E5504 @ 2.00GHz
stepping : 5
cpu MHz : 1595.992
cache size : 4096 KB
physical id : 0
siblings : 4
core id : 0
cpu cores : 4
apicid : 0
initial apicid : 0
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 bts 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 dts tpr_shadow vnmi flexpriority ept vpid
bogomips : 3191.50
clflush size : 64
cache_alignment : 64
address sizes : 40 bits physical, 48 bits virtual
power management:
processor : 2
vendor_id : GenuineIntel
cpu family : 6
model : 26
model name : Intel(R) Xeon(R) CPU E5504 @ 2.00GHz
stepping : 5
cpu MHz : 1595.992
cache size : 4096 KB
physical id : 1
siblings : 4
core id : 1
cpu cores : 4
apicid : 18
initial apicid : 18
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 bts 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 dts tpr_shadow vnmi flexpriority ept vpid
bogomips : 3191.98
clflush size : 64
cache_alignment : 64
address sizes : 40 bits physical, 48 bits virtual
power management:
processor : 3
vendor_id : GenuineIntel
cpu family : 6
model : 26
model name : Intel(R) Xeon(R) CPU E5504 @ 2.00GHz
stepping : 5
cpu MHz : 1595.992
cache size : 4096 KB
physical id : 0
siblings : 4
core id : 1
cpu cores : 4
apicid : 2
initial apicid : 2
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 bts 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 dts tpr_shadow vnmi flexpriority ept vpid
bogomips : 3191.50
clflush size : 64
cache_alignment : 64
address sizes : 40 bits physical, 48 bits virtual
power management:
processor : 4
vendor_id : GenuineIntel
cpu family : 6
model : 26
model name : Intel(R) Xeon(R) CPU E5504 @ 2.00GHz
stepping : 5
cpu MHz : 1595.992
cache size : 4096 KB
physical id : 1
siblings : 4
core id : 2
cpu cores : 4
apicid : 20
initial apicid : 20
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 bts 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 dts tpr_shadow vnmi flexpriority ept vpid
bogomips : 3191.98
clflush size : 64
cache_alignment : 64
address sizes : 40 bits physical, 48 bits virtual
power management:
processor : 5
vendor_id : GenuineIntel
cpu family : 6
model : 26
model name : Intel(R) Xeon(R) CPU E5504 @ 2.00GHz
stepping : 5
cpu MHz : 1595.992
cache size : 4096 KB
physical id : 0
siblings : 4
core id : 2
cpu cores : 4
apicid : 4
initial apicid : 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 pbe syscall nx rdtscp lm constant_tsc arch_perfmon pebs bts 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 dts tpr_shadow vnmi flexpriority ept vpid
bogomips : 3191.50
clflush size : 64
cache_alignment : 64
address sizes : 40 bits physical, 48 bits virtual
power management:
processor : 6
vendor_id : GenuineIntel
cpu family : 6
model : 26
model name : Intel(R) Xeon(R) CPU E5504 @ 2.00GHz
stepping : 5
cpu MHz : 1595.992
cache size : 4096 KB
physical id : 1
siblings : 4
core id : 3
cpu cores : 4
apicid : 22
initial apicid : 22
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 bts 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 dts tpr_shadow vnmi flexpriority ept vpid
bogomips : 3191.98
clflush size : 64
cache_alignment : 64
address sizes : 40 bits physical, 48 bits virtual
power management:
processor : 7
vendor_id : GenuineIntel
cpu family : 6
model : 26
model name : Intel(R) Xeon(R) CPU E5504 @ 2.00GHz
stepping : 5
cpu MHz : 1595.992
cache size : 4096 KB
physical id : 0
siblings : 4
core id : 3
cpu cores : 4
apicid : 6
initial apicid : 6
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 bts 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 dts tpr_shadow vnmi flexpriority ept vpid
bogomips : 3191.50
clflush size : 64
cache_alignment : 64
address sizes : 40 bits physical, 48 bits virtual
power management:
从以上CPUINFO可以看出, 这里有2个物理CPU, 0和1.
使用dmidecode也可以看到这部分信息
[root@db-172-16-3-150 ~]# dmidecode -t processor
# dmidecode 2.11
SMBIOS 2.6 present.
Handle 0x0400, DMI type 4, 40 bytes
Processor Information
Socket Designation: CPU1
Type: Central Processor
Family: Xeon
Manufacturer: Intel
ID: A5 06 01 00 FF FB EB BF
Signature: Type 0, Family 6, Model 26, Stepping 5
Flags:
FPU (Floating-point unit on-chip)
VME (Virtual mode extension)
DE (Debugging extension)
PSE (Page size extension)
TSC (Time stamp counter)
MSR (Model specific registers)
PAE (Physical address extension)
MCE (Machine check exception)
CX8 (CMPXCHG8 instruction supported)
APIC (On-chip APIC hardware supported)
SEP (Fast system call)
MTRR (Memory type range registers)
PGE (Page global enable)
MCA (Machine check architecture)
CMOV (Conditional move instruction supported)
PAT (Page attribute table)
PSE-36 (36-bit page size extension)
CLFSH (CLFLUSH instruction supported)
DS (Debug store)
ACPI (ACPI supported)
MMX (MMX technology supported)
FXSR (FXSAVE and FXSTOR instructions supported)
SSE (Streaming SIMD extensions)
SSE2 (Streaming SIMD extensions 2)
SS (Self-snoop)
HTT (Multi-threading)
TM (Thermal monitor supported)
PBE (Pending break enabled)
Version: Intel(R) Xeon(R) CPU E5504 @ 2.00GHz
Voltage: 1.2 V
External Clock: 4800 MHz
Max Speed: 3600 MHz
Current Speed: 1600 MHz
Status: Populated, Enabled
Upgrade: Socket LGA1366
L1 Cache Handle: 0x0700
L2 Cache Handle: 0x0701
L3 Cache Handle: 0x0702
Serial Number: Not Specified
Asset Tag: Not Specified
Part Number: Not Specified
Core Count: 4
Core Enabled: 4
Thread Count: 4
Characteristics:
64-bit capable
Handle 0x0401, DMI type 4, 40 bytes
Processor Information
Socket Designation: CPU2
Type: Central Processor
Family: Xeon
Manufacturer: Intel
ID: A5 06 01 00 FF FB EB BF
Signature: Type 0, Family 6, Model 26, Stepping 5
Flags:
FPU (Floating-point unit on-chip)
VME (Virtual mode extension)
DE (Debugging extension)
PSE (Page size extension)
TSC (Time stamp counter)
MSR (Model specific registers)
PAE (Physical address extension)
MCE (Machine check exception)
CX8 (CMPXCHG8 instruction supported)
APIC (On-chip APIC hardware supported)
SEP (Fast system call)
MTRR (Memory type range registers)
PGE (Page global enable)
MCA (Machine check architecture)
CMOV (Conditional move instruction supported)
PAT (Page attribute table)
PSE-36 (36-bit page size extension)
CLFSH (CLFLUSH instruction supported)
DS (Debug store)
ACPI (ACPI supported)
MMX (MMX technology supported)
FXSR (FXSAVE and FXSTOR instructions supported)
SSE (Streaming SIMD extensions)
SSE2 (Streaming SIMD extensions 2)
SS (Self-snoop)
HTT (Multi-threading)
TM (Thermal monitor supported)
PBE (Pending break enabled)
Version: Intel(R) Xeon(R) CPU E5504 @ 2.00GHz
Voltage: 1.2 V
External Clock: 4800 MHz
Max Speed: 3600 MHz
Current Speed: 1600 MHz
Status: Populated, Idle
Upgrade: Socket LGA1366
L1 Cache Handle: 0x0703
L2 Cache Handle: 0x0704
L3 Cache Handle: 0x0705
Serial Number: Not Specified
Asset Tag: Not Specified
Part Number: Not Specified
Core Count: 4
Core Enabled: 4
Thread Count: 4
Characteristics:
64-bit capable
[root@db-172-16-3-150 ~]# cat /proc/cpuinfo |grep "core id"
core id : 0
core id : 0
core id : 1
core id : 1
core id : 2
core id : 2
core id : 3
core id : 3
[root@db-172-16-3-150 ~]# cat /proc/cpuinfo |grep "physical id"
physical id : 1
physical id : 0
physical id : 1
physical id : 0
physical id : 1
physical id : 0
physical id : 1
physical id : 0
在没有运行业务的情况下, 我们要测试PostgreSQL某SQL的IO请求次数和请求时间.
启动PostgreSQL数据库时, 把CPU亲和设置为0. (也就是文章开头提到的性能最烂的做法.)
pg93@db-172-16-3-150-> taskset -c 0 /home/pg93/pgsql9.3.1/bin/postgres >/dev/null 2>&1
使用亲和1启动psql.
pg93@db-172-16-3-150-> taskset -c 1 psql
psql (9.3.1)
Type "help" for help.
digoal=# select pg_backend_pid();
pg_backend_pid
----------------
24781
(1 row)
然后在启动stap时把亲和设置为和PostgreSQL不一样的CPU id. 例如3(这样也确保了不在一个物理core上.).
[root@db-172-16-3-150 ~]# taskset -c 3 stap -e '
global a
probe process("/home/pg93/pgsql9.3.1/bin/postgres").mark("query__start") {
delete a
println("query__start ", user_string($arg1), "pid:", pid())
}
probe vfs.read.return {
t = gettimeofday_ns() - @entry(gettimeofday_ns())
# if (execname() == "postgres" && devname != "N/A") # 这句判断也有一定的影响, 所以不用. 我们使用target模式.
a[pid()] <<< t
}
probe process("/home/pg93/pgsql9.3.1/bin/postgres").mark("query__done") {
printdln("**", pid(), @count(a[pid()]), @avg(a[pid()]))
println("query__done ", user_string($arg1), "pid:", pid())
delete a
}' -x 24781
在psql中连续快速的执行SQL
digoal=# explain (analyze,verbose,costs,buffers,timing) select * from tbl_cost_align;
QUERY PLAN
------------------------------------------------------------------------------------------------------------------------------------
----
Seq Scan on postgres.tbl_cost_align (cost=0.00..195393.00 rows=10100000 width=45) (actual time=0.023..3154.782 rows=10100000 loops
=1)
Output: id, info, crt_time
Buffers: shared read=94393
Total runtime: 4710.788 ms
(4 rows)
digoal=# explain (analyze,verbose,costs,buffers,timing) select * from tbl_cost_align;
QUERY PLAN
------------------------------------------------------------------------------------------------------------------------------------
----
Seq Scan on postgres.tbl_cost_align (cost=0.00..195393.00 rows=10100000 width=45) (actual time=0.040..3151.680 rows=10100000 loops
=1)
Output: id, info, crt_time
Buffers: shared hit=32 read=94361
Total runtime: 4696.888 ms
(4 rows)
digoal=# explain (analyze,verbose,costs,buffers,timing) select * from tbl_cost_align;
QUERY PLAN
------------------------------------------------------------------------------------------------------------------------------------
----
Seq Scan on postgres.tbl_cost_align (cost=0.00..195393.00 rows=10100000 width=45) (actual time=0.035..3158.134 rows=10100000 loops
=1)
Output: id, info, crt_time
Buffers: shared hit=64 read=94329
Total runtime: 4706.135 ms
(4 rows)
stap输出
query__start explain (analyze,verbose,costs,buffers,timing) select * from tbl_cost_align;pid:24781
24781**94417**5461
query__done explain (analyze,verbose,costs,buffers,timing) select * from tbl_cost_align;pid:24781
query__start explain (analyze,verbose,costs,buffers,timing) select * from tbl_cost_align;pid:24781
24781**94361**5429
query__done explain (analyze,verbose,costs,buffers,timing) select * from tbl_cost_align;pid:24781
query__start explain (analyze,verbose,costs,buffers,timing) select * from tbl_cost_align;pid:24781
24781**94329**5529
query__done explain (analyze,verbose,costs,buffers,timing) select * from tbl_cost_align;pid:24781
我们看到每次read IO的时间为5461纳秒左右.
如果把stap的亲和设置为非0, 即使与stap的亲和一致也没关系. 那么会发生什么情况呢?
pg93@db-172-16-3-150-> taskset -c 3 /home/pg93/pgsql9.3.1/bin/postgres >/dev/null 2>&1
pg93@db-172-16-3-150-> taskset -c 3 psql
psql (9.3.1)
Type "help" for help.
digoal=# select pg_backend_pid();
pg_backend_pid
----------------
24829
(1 row)
[root@db-172-16-3-150 ~]# taskset -c 3 stap -e '
global a
probe process("/home/pg93/pgsql9.3.1/bin/postgres").mark("query__start") {
delete a
println("query__start ", user_string($arg1), "pid:", pid())
}
probe vfs.read.return {
t = gettimeofday_ns() - @entry(gettimeofday_ns())
# if (execname() == "postgres" && devname != "N/A") # 这句判断也有一定的影响, 所以不用. 建议使用target模式.
a[pid()] <<< t
}
probe process("/home/pg93/pgsql9.3.1/bin/postgres").mark("query__done") {
printdln("**", pid(), @count(a[pid()]), @avg(a[pid()]))
println("query__done ", user_string($arg1), "pid:", pid())
delete a
}' -x 24829
digoal=# explain (analyze,verbose,costs,buffers,timing) select * from tbl_cost_align;
QUERY PLAN
------------------------------------------------------------------------------------------------------------------------------------
----
Seq Scan on postgres.tbl_cost_align (cost=0.00..195393.00 rows=10100000 width=45) (actual time=0.023..2039.248 rows=10100000 loops
=1)
Output: id, info, crt_time
Buffers: shared read=94393
Total runtime: 3065.989 ms
(4 rows)
digoal=# explain (analyze,verbose,costs,buffers,timing) select * from tbl_cost_align;
QUERY PLAN
------------------------------------------------------------------------------------------------------------------------------------
----
Seq Scan on postgres.tbl_cost_align (cost=0.00..195393.00 rows=10100000 width=45) (actual time=0.036..2039.643 rows=10100000 loops
=1)
Output: id, info, crt_time
Buffers: shared hit=32 read=94361
Total runtime: 3066.841 ms
(4 rows)
digoal=# explain (analyze,verbose,costs,buffers,timing) select * from tbl_cost_align;
QUERY PLAN
------------------------------------------------------------------------------------------------------------------------------------
----
Seq Scan on postgres.tbl_cost_align (cost=0.00..195393.00 rows=10100000 width=45) (actual time=0.034..2039.005 rows=10100000 loops
=1)
Output: id, info, crt_time
Buffers: shared hit=64 read=94329
Total runtime: 3066.746 ms
(4 rows)
query__start explain (analyze,verbose,costs,buffers,timing) select * from tbl_cost_align;pid:24829
24829**94417**3277
query__done explain (analyze,verbose,costs,buffers,timing) select * from tbl_cost_align;pid:24829
query__start explain (analyze,verbose,costs,buffers,timing) select * from tbl_cost_align;pid:24829
24829**94361**3272
query__done explain (analyze,verbose,costs,buffers,timing) select * from tbl_cost_align;pid:24829
query__start explain (analyze,verbose,costs,buffers,timing) select * from tbl_cost_align;pid:24829
24829**94329**3273
query__done explain (analyze,verbose,costs,buffers,timing) select * from tbl_cost_align;pid:24829
每次IO响应时间降到了3273纳秒左右, 几乎下降了一半.
最后把STAP的亲和设置为4, (与postgres进程分开使用物理CPU)那么会怎么样呢?
digoal=# explain (analyze,verbose,costs,buffers,timing) select * from tbl_cost_align;
QUERY PLAN
------------------------------------------------------------------------------------------------------------------------------------
----
Seq Scan on postgres.tbl_cost_align (cost=0.00..195393.00 rows=10100000 width=45) (actual time=0.055..2032.953 rows=10100000 loops
=1)
Output: id, info, crt_time
Buffers: shared hit=96 read=94297
Total runtime: 3059.635 ms
(4 rows)
digoal=# explain (analyze,verbose,costs,buffers,timing) select * from tbl_cost_align;
QUERY PLAN
------------------------------------------------------------------------------------------------------------------------------------
----
Seq Scan on postgres.tbl_cost_align (cost=0.00..195393.00 rows=10100000 width=45) (actual time=0.044..2034.114 rows=10100000 loops
=1)
Output: id, info, crt_time
Buffers: shared hit=128 read=94265
Total runtime: 3060.832 ms
(4 rows)
digoal=# explain (analyze,verbose,costs,buffers,timing) select * from tbl_cost_align;
QUERY PLAN
------------------------------------------------------------------------------------------------------------------------------------
----
Seq Scan on postgres.tbl_cost_align (cost=0.00..195393.00 rows=10100000 width=45) (actual time=0.043..2038.548 rows=10100000 loops
=1)
Output: id, info, crt_time
Buffers: shared hit=160 read=94233
Total runtime: 3065.674 ms
(4 rows)
测得的IO响应时间又有了一定的下降, 说明还是有一定效果的. 但是相比不使用CPU 0 这里显得效果比较弱了.
query__start explain (analyze,verbose,costs,buffers,timing) select * from tbl_cost_align;pid:24829
24829**94297**3194
query__done explain (analyze,verbose,costs,buffers,timing) select * from tbl_cost_align;pid:24829
query__start explain (analyze,verbose,costs,buffers,timing) select * from tbl_cost_align;pid:24829
24829**94265**3200
query__done explain (analyze,verbose,costs,buffers,timing) select * from tbl_cost_align;pid:24829
query__start explain (analyze,verbose,costs,buffers,timing) select * from tbl_cost_align;pid:24829
24829**94233**3195
query__done explain (analyze,verbose,costs,buffers,timing) select * from tbl_cost_align;pid:24829
最后在附一个未开启stap得到的结果.
digoal=# explain (analyze,verbose,costs,buffers,timing) select * from tbl_cost_align;
QUERY PLAN
------------------------------------------------------------------------------------------------------------------------------------
----
Seq Scan on postgres.tbl_cost_align (cost=0.00..195393.00 rows=10100000 width=45) (actual time=0.037..1812.712 rows=10100000 loops
=1)
Output: id, info, crt_time
Buffers: shared hit=192 read=94201
Total runtime: 2833.760 ms
(4 rows)
digoal=# explain (analyze,verbose,costs,buffers,timing) select * from tbl_cost_align;
QUERY PLAN
------------------------------------------------------------------------------------------------------------------------------------
----
Seq Scan on postgres.tbl_cost_align (cost=0.00..195393.00 rows=10100000 width=45) (actual time=0.038..1814.074 rows=10100000 loops
=1)
Output: id, info, crt_time
Buffers: shared hit=224 read=94169
Total runtime: 2835.363 ms
(4 rows)
digoal=# explain (analyze,verbose,costs,buffers,timing) select * from tbl_cost_align;
QUERY PLAN
------------------------------------------------------------------------------------------------------------------------------------
----
Seq Scan on postgres.tbl_cost_align (cost=0.00..195393.00 rows=10100000 width=45) (actual time=0.039..1814.641 rows=10100000 loops
=1)
Output: id, info, crt_time
Buffers: shared hit=256 read=94137
Total runtime: 2840.508 ms
(4 rows)
stap开启后, 查询时间长了220毫秒左右. 本文中每个handler带来的额外开销是 :
digoal=# select 220/94137.0;
?column?
------------------------
0.00233701945037551653
(1 row)
0.002337毫秒.
2337纳秒.
使用numactl也可以绑定进程的物理CPU核。
参考
1. man taskset , man numactl
2. https://access.redhat.com/site/documentation/en-US/Red_Hat_Enterprise_Linux/6/html/Performance_Tuning_Guide/main-cpu.html
3. http://docs.mongodb.org/manual/administration/production-notes/#mongodb-on-numa-hardware