4 对操作系统内核参数作出调整
• 对vmtune的参数进行调整如下：
1
-p -P -r -R -f -F -N -W
2 minperm maxperm minpgahead maxpgahead minfree maxfree pd_npages maxrandwrt
3 396414 792828 0 0 960 1024 65536 1
4 -M -w -k -c -b -B -u -l -d
5 maxpin npswarn npskill numclust numfsbufs hd_pbuf_cnt lvm_bufcnt lrubucket defps
6 3250555 65536 16384 16 2048 2656 9 131072 1
7
8 -s -n -S -L -g -h
9 sync_release_ilock nokilluid v_pinshm lgpg_regions lgpg_size strict_maxperm
10
11 1 0 0 0 0 0
12 -t -j -J -z
13 maxclient j2_nPagesPer j2_maxRandomWrite j2_nRandomCluster
14
15 792828 32 0 0
16 -Z -q -Q -y
17
18 j2_nBufferPer j2_minPageReadAhead j2_maxPageReadAhead memory_affinity
19
20 512 2 8 0
21 -V -i
22
23 num_spec_dataseg spec_dataseg_int
24
25 0 512
26
27 PTA balance threshold percentage = 50.0%
28
29 number of valid memory pages = 4063193 maxperm=20.0% of real memory
30
31 maximum pinable=80.0% of real memory minperm=10.0% of real memory
32
33 number of file memory pages = 1955960 numperm=49.3% of real memory
34
35 number of compressed memory pages = 0 compressed=0.0% of real memory
36
37 number of client memory pages = 0 numclient=0.0% of real memory
38
39 # of remote pgs sched-pageout = 0 maxclient=20.0% of real memory
40
41 • 对schedtune参数的调整如下：
42
43 THRASH SUSP FORK SCHED
44 -h -p -m -w -e -f -d -r -t -s
45
46 SYS PROC MULTI WAIT GRACE TICKS SCHED_D SCHED_R TIMESLICE MAXSPIN
47 0 4 2 1 2 10 16 16 1 8192
48 CLOCK SCHED_FIFO2 IDLE MIGRATION FIXED_PRI
49
50 -c -a -b -F
51 %usDELTA AFFINITY_LIM BARRIER/16 GLOBAL(1)
52
53 100 7 4 0
-p -P -r -R -f -F -N -W2
minperm maxperm minpgahead maxpgahead minfree maxfree pd_npages maxrandwrt3
396414 792828 0 0 960 1024 65536 14
-M -w -k -c -b -B -u -l -d5
maxpin npswarn npskill numclust numfsbufs hd_pbuf_cnt lvm_bufcnt lrubucket defps6
3250555 65536 16384 16 2048 2656 9 131072 17
8
-s -n -S -L -g -h9
sync_release_ilock nokilluid v_pinshm lgpg_regions lgpg_size strict_maxperm10
11
1 0 0 0 0 012
-t -j -J -z13
maxclient j2_nPagesPer j2_maxRandomWrite j2_nRandomCluster14
15
792828 32 0 016
-Z -q -Q -y17
18
j2_nBufferPer j2_minPageReadAhead j2_maxPageReadAhead memory_affinity19
20
512 2 8 021
-V -i22
23
num_spec_dataseg spec_dataseg_int24
25
0 51226
27
PTA balance threshold percentage = 50.0%28
29
number of valid memory pages = 4063193 maxperm=20.0% of real memory30
31
maximum pinable=80.0% of real memory minperm=10.0% of real memory32
33
number of file memory pages = 1955960 numperm=49.3% of real memory34
35
number of compressed memory pages = 0 compressed=0.0% of real memory36
37
number of client memory pages = 0 numclient=0.0% of real memory38
39
# of remote pgs sched-pageout = 0 maxclient=20.0% of real memory40
41
• 对schedtune参数的调整如下：42
43
THRASH SUSP FORK SCHED44
-h -p -m -w -e -f -d -r -t -s45
46
SYS PROC MULTI WAIT GRACE TICKS SCHED_D SCHED_R TIMESLICE MAXSPIN47
0 4 2 1 2 10 16 16 1 819248
CLOCK SCHED_FIFO2 IDLE MIGRATION FIXED_PRI49
50
-c -a -b -F51
%usDELTA AFFINITY_LIM BARRIER/16 GLOBAL(1)52
53
100 7 4 05 检查资源消耗最大语句的逻辑设计。
建立合适的分区索引
将排名靠前语句的表数据与索引分别存储。
提高资源消耗靠前语句的并行度。
Oracle 的参数cursor_sharing设为force减少internal lock wait
调整Latch的数量(如DB_BLOCK_LRU_LATCHES)或内存的一些参数(如SHARED_POOL_SIZE等)解决内部锁问题。
6 由于数据存储调整后，系统I/O性能状况良好，系统和存储端的性能表现已调整至非常好的，系统和存储端的性能调优工作已经完成。建议客户下一步的工作重点是解决应用系统的性能瓶颈，结合Precise对应用的监控结果，检查应用的逻辑设计，数据索引的建立和分布。协助应用开发商修改应用。
7 调整操作系统交换空间(page space)。
