对磁盘条带化性能的探索（zt）

1. 环境说明

主机： IBM P520

CPU： 4695MHz × 4

内存： 15744MB

操作系统： AIX 6100-06

存储： 400GB × 4（主机自带磁盘）

# lspv

hdisk0 00cbc586dda5fad1 rootvg active

hdisk1 00cbc586d6338d91 rootvg active

hdisk2 00cbc586551a042f testvg active

hdisk3 00cbc586551a04b7 testvg active

# lsvg testvg

VOLUME GROUP: testvg VG IDENTIFIER: 00cbc58600004c00000

001348260cf99

VG STATE: active PP SIZE: 512 megabyte(s)

VG PERMISSION: read/write TOTAL PPs: 1676 (858112 megaby

tes)

MAX LVs: 256 FREE PPs: 875 (448000 megabyt

es)

LVs: 8 USED PPs: 801 (410112 megabyt

es)

OPEN LVs: 8 QUORUM: 2 (Enabled)

TOTAL PVs: 2 VG DESCRIPTORS: 3

STALE PVs: 1 STALE PPs: 100

ACTIVE PVs: 2 AUTO ON: yes

MAX PPs per VG: 32512

MAX PPs per PV: 1016 MAX PVs: 32

LTG size (Dynamic): 1024 kilobyte(s) AUTO SYNC: no

HOT SPARE: no BB POLICY: relocatable

PV RESTRICTION: none

# lsvg -l testvg

testvg:

LV NAME TYPE LPs PPs PVs LV STATE MOUNT POINT

stripe256lv jfs2 100 100 2 open/syncd /stripe256lv

stripe512lv jfs2 100 100 2 open/syncd /stripe512lv

stripe1024lv jfs2 100 100 2 open/syncd /stripe1024lv

stripe4096lv jfs2 100 100 2 open/syncd /stripe4096lv

stripe8192lv jfs2 100 100 2 open/syncd /stripe8192lv

loglv00 jfs2log 1 1 1 open/syncd N/A

nostripelv jfs2 100 100 1 open/syncd /nostripelv

# lslv nostripelv

LOGICAL VOLUME: nostripelv VOLUME GROUP: testvg

LV IDENTIFIER: 00cbc58600004c00000001348260cf99.7 PERMISSION: read/writ

e

VG STATE: active/complete LV STATE: opened/syncd

TYPE: jfs2 WRITE VERIFY: off

MAX LPs: 512 PP SIZE: 512 megabyte(s)

COPIES: 1 SCHED POLICY: parallel

LPs: 100 PPs: 100

STALE PPs: 0 BB POLICY: relocatable

INTER-POLICY: minimum RELOCATABLE: yes

INTRA-POLICY: middle UPPER BOUND: 32

MOUNT POINT: /nostripelv LABEL: /nostripelv

MIRROR WRITE CONSISTENCY: on/ACTIVE

EACH LP COPY ON A SEPARATE PV ?: yes

Serialize IO ?: NO

# lslv stripe256lv

LOGICAL VOLUME: stripe256lv VOLUME GROUP: testvg

LV IDENTIFIER: 00cbc58600004c00000001348260cf99.1 PERMISSION: read/writ

e

VG STATE: active/complete LV STATE: opened/syncd

TYPE: jfs2 WRITE VERIFY: off

MAX LPs: 512 PP SIZE: 512 megabyte(s)

COPIES: 1 SCHED POLICY: striped

LPs: 100 PPs: 100

STALE PPs: 0 BB POLICY: relocatable

INTER-POLICY: maximum RELOCATABLE: no

INTRA-POLICY: middle UPPER BOUND: 2

MOUNT POINT: /stripe256lv LABEL: /stripe256lv

MIRROR WRITE CONSISTENCY: on/ACTIVE

EACH LP COPY ON A SEPARATE PV ?: yes (superstrict)

Serialize IO ?: NO

STRIPE WIDTH: 2

STRIPE SIZE: 256K

# lslv stripe512lv

LOGICAL VOLUME: stripe512lv VOLUME GROUP: testvg

LV IDENTIFIER: 00cbc58600004c00000001348260cf99.2 PERMISSION: read/writ

e

VG STATE: active/complete LV STATE: opened/syncd

TYPE: jfs2 WRITE VERIFY: off

MAX LPs: 512 PP SIZE: 512 megabyte(s)

COPIES: 1 SCHED POLICY: striped

LPs: 100 PPs: 100

STALE PPs: 0 BB POLICY: relocatable

INTER-POLICY: maximum RELOCATABLE: no

INTRA-POLICY: middle UPPER BOUND: 2

MOUNT POINT: /stripe512lv LABEL: /stripe512lv

MIRROR WRITE CONSISTENCY: on/ACTIVE

EACH LP COPY ON A SEPARATE PV ?: yes (superstrict)

Serialize IO ?: NO

STRIPE WIDTH: 2

STRIPE SIZE: 512K

# lslv stripe1024lv

LOGICAL VOLUME: stripe1024lv VOLUME GROUP: testvg

LV IDENTIFIER: 00cbc58600004c00000001348260cf99.3 PERMISSION: read/writ

e

VG STATE: active/complete LV STATE: opened/syncd

TYPE: jfs2 WRITE VERIFY: off

MAX LPs: 512 PP SIZE: 512 megabyte(s)

COPIES: 1 SCHED POLICY: striped

LPs: 100 PPs: 100

STALE PPs: 0 BB POLICY: relocatable

INTER-POLICY: maximum RELOCATABLE: no

INTRA-POLICY: middle UPPER BOUND: 2

MOUNT POINT: /stripe1024lv LABEL: /stripe1024lv

MIRROR WRITE CONSISTENCY: on/ACTIVE

EACH LP COPY ON A SEPARATE PV ?: yes (superstrict)

Serialize IO ?: NO

STRIPE WIDTH: 2

STRIPE SIZE: 1M

# lslv stripe4096lv

LOGICAL VOLUME: stripe4096lv VOLUME GROUP: testvg

LV IDENTIFIER: 00cbc58600004c00000001348260cf99.4 PERMISSION: read/writ

e

VG STATE: active/complete LV STATE: opened/syncd

TYPE: jfs2 WRITE VERIFY: off

MAX LPs: 512 PP SIZE: 512 megabyte(s)

COPIES: 1 SCHED POLICY: striped

LPs: 100 PPs: 100

STALE PPs: 0 BB POLICY: relocatable

INTER-POLICY: maximum RELOCATABLE: no

INTRA-POLICY: middle UPPER BOUND: 2

MOUNT POINT: /stripe4096lv LABEL: /stripe4096lv

MIRROR WRITE CONSISTENCY: on/ACTIVE

EACH LP COPY ON A SEPARATE PV ?: yes (superstrict)

Serialize IO ?: NO

STRIPE WIDTH: 2

STRIPE SIZE: 4M

# lslv stripe8192lv

LOGICAL VOLUME: stripe8192lv VOLUME GROUP: testvg

LV IDENTIFIER: 00cbc58600004c00000001348260cf99.5 PERMISSION: read/writ

e

VG STATE: active/complete LV STATE: opened/syncd

TYPE: jfs2 WRITE VERIFY: off

MAX LPs: 512 PP SIZE: 512 megabyte(s)

COPIES: 1 SCHED POLICY: striped

LPs: 100 PPs: 100

STALE PPs: 0 BB POLICY: relocatable

INTER-POLICY: maximum RELOCATABLE: no

INTRA-POLICY: middle UPPER BOUND: 2

MOUNT POINT: /stripe8192lv LABEL: /stripe8192lv

MIRROR WRITE CONSISTENCY: on/ACTIVE

EACH LP COPY ON A SEPARATE PV ?: yes (superstrict)

Serialize IO ?: NO

STRIPE WIDTH: 2

STRIPE SIZE: 8M

2. 测试方法

在对应的lv上建立文件系统，通过iozone对各文件系统的读写吞吐量进行测试。

主要针对Writer,Re-writer,Reader,Re-reader,Rondom Read,Rondom Write进行测试。

此次测试执行的命令如下：

./iozone -n 512m -g 32g -i 0 -i 1 -i 2 -Rab stripe8192lv.xls

iozone的使用说明请参考http://www.iozone.org

3. 测试结论

连续写：

1、 未做条带化的性能相当于条带化后的一半

2、 条带化为256k和512k的性能最优，1m与之差距不大，1m以后随着条带化大小的加大，性能逐步下降

3、 块大小大于1m以后，吞吐量开始明显有下降，小于1m的情况下差距不大

4、 随着文件的增大，性能逐步下降

连续读：

1、 块大小为1m时性能最佳，当块大小为16m时，性能骤降

2、 条带化对于吞吐量的影响不大

3、 当读取文件大于16g时，性能骤降

随机读结论同连续读

随机写：

1、 块小于256k时性能较好，大于256k后性能骤降

2、 写大于16g的文件时性能骤降

综合结论：

1、 文件大小需要控制，随着文件的变大，读写性能也会逐步降低，特别是当文件大于16g的时候，性能衰退更为明显。针对数据库的数据文件，单个文件需要控制在8g以内，且不能让其自主扩展。对于undo,temp表空间的数据文件，也应适时进行重建，以控制其大小。

2、 对于写频繁的业务系统，磁盘规划时必须要进行条带化以提高性能。

3、 对于条带化大小需要根据业务系统的IO性质来觉得，对于OLTP系统，256k及512k为佳。

4、 程序如可控制每次读写块的大小时，应采用与LTG大小为块大小。

[@more@]

来自 “ ITPUB博客 ” ，链接：http://blog.itpub.net/41451/viewspace-1057792/，如需转载，请注明出处，否则将追究法律责任。

转载于:http://blog.itpub.net/41451/viewspace-1057792/