smartassist - smartio caching profiler tool.
# /etc/vx/diag.d/smartassist start [-t <time>] -o <output dir> [--cluster] -d <device1 device2 ...> # /etc/vx/diag.d/smartassist start [-t <time>] -o <output dir> [--cluster] -d <volume1_path volume2_path ...> # /etc/vx/diag.d/smartassist start [-t <time>] -o <output dir> [--cluster] -g <disk-group1 disk-group2 ...> # /etc/vx/diag.d/smartassist start [-t <time>] -o <output dir> [--cluster] -m <mount-point1 mount-point2 ...> # /etc/vx/diag.d/smartassist start [-t <time>] -o <output dir> --app oracle <oracle_user_name> <ORACLE_HOME> <ORACLE_SID> [<ORACLE_ASMDG>] # /etc/vx/diag.d/smartassist start [-t <time>] -o <output dir> --app sybase <DB_user_name> <DB_server> <DB_name> # /etc/vx/diag.d/smartassist start [-t <time>] -o <output dir> --app db2 <DB_user_name> <DB_user_home> <DB_name> # /etc/vx/diag.d/smartassist analyze [--cluster] [-c <size>] [--cache-latency <val1>] [--data-latency <val2>] -i <dir>
The Smartassist tool (version 2.0) analyzes the I/O on the specified target in the system for specified time, and calculates the optimal cache size for the workload.The Smartassist tool supports the following targets:
1. List of device paths 2. List of VxVM/LVM volumes paths 3. List of VxVM/LVM disk groups names 4. List of filesystem mount points 5. Oracle/Sybase/DB2 database instance 6. List of shared volumes/diskgroups of VxVM 7. List of CFS mount pointsThis tool works in two phases:
1. Start phase:
In start phase, the tool traces the I/O on the specified target for specified time and stores the trace output in a specified directory. In case of disk devices which have multiple paths, it is advisable to to specify all the paths to get correct results. Using VxVM/LVM volume devices in such case is more convenient. By default, the start phase runs for 3600 seconds.
2. Analayze phase:
In analyze phase, the tool parses the output generated in the start phase and uses the SmartIO algorithms to calculate optimal cache size. The optimal cache size, read hits, latency gain, performance forecast and share in cache for each target are displayed on the terminal.
The Smartassist tool runs on the following platforms:
Linux RHEL 5, RHEL 6, RHEL 7, SLES10, SLES11, SLES12 Solaris Solaris 9, Solaris 10, Solaris 11 Aix Aix 6.1, 1 7.1
Latest Smartassist tool for required platform can be downloaded from the following link:
https://sort.veritas.com/data_collectors
On the system, where the Smartassist tool runs, enter the following commands to unzip or untar the tool files:
# mkdir <Smartassist_path> # cd <Smartassist_path> # gunzip smartassist.tar.gz # tar xvf smartassist.tar The tool is installed in the <Smartassist_path> directory once the files are extracted.
For all platforms -
a. Veritas DMP device as a target is not supported.
b. Configuration involving Veritas volume sets (vset) is not supported.
c. Mixed target types is not supported. e.g. ZFS and VxFS mount
points, VxVM and LVM diskgroups etc.
You need the root permission to run the tool.
1. 1
a. Veritas Volume Manager (VxVM) is a pre-requisite.
b. Smartassist uses vxtrace to monitor I/O on the specified targets.
c. Supported filesystems: VxFS
d. Supported volumes/diskgroups: VxVM
1. ORACLE
a. It is mandatory to specify ASM diskgroup name if Oracle database
instance is configured on ASM diskgroup. 2. SYBASE
a. Following major versions of Sybase ASE are supported -
15.x, 16.0 or later.
b. Before running Smartassist start operation for Sybase database,
export SYBASE environment variable. 3. DB2
a. PATH variable of specified user should be set to include path
of db2 command. 4. Oracle, Sybase and DB2 are supported on all platforms. 5. Only single instance of databases are supported.
NOTE: Smartassist assumes that in case of database all the arguments (eg. database name, database user etc) specified are correct.
CLUSTER OPTION:
a. Pre-requisite: Veritas Cluster Server (VCS) package.
b. Supported targets: Shared VxVM volumes/diskgroups, CFS mount
points.
c. Any node in the cluster can act as initiator node for the tool.
d. Start phase starts the IO monitoring on all nodes of the cluster.
e. Analyze phase analyzes monitored IO on respective cluster nodes and
copies the per node results on the initiator node.
1. Go to the <Smartassist_path> directory. 2. Specify the targets that you want to trace the I/O for, the time span3. Enter the following command to analyze the output:
in seconds, and the directory to save the output.
For start phase, the arguments vary for different targets as follows-
a. list of device paths:
# ./smartassist start [-t <time>] -o <output dir> [--cluster] -d <device1 device2 ...>
b. list of VxVM/LVM volumes paths:
# ./smartassist start [-t <time>] -o <output dir> [--cluster] -d <volume1_path volume2_path ...>
c. list of VxVM/LVM disk group names
# ./smartassist start [-t <time>] -o <output dir> [--cluster] -g <disk-group1 disk-group2 ...>
d. file system mount point:
# ./smartassist start [-t <time>] -o <output dir> [--cluster] -m <mount-point1 mount-point2 ...>
e. Oracle/Sybase/DB2 database instance:
# ./smartassist start [-t <time>] -o <output dir> --app oracle <oracle_user_name> <ORACLE_HOME> <ORACLE_SID> [<ORACLE_ASMDG>]
# ./smartassist start [-t <time>] -o <output dir> --app sybase <DB_user_name> <DB_server> <DB_name>
# ./smartassist start [-t <time>] -o <output dir> --app db2 <DB_user_name> <DB_user_home> <DB_name>
--cluster Used to indicate that smartassist should perform its operation (start/analyze) across CVM cluster. This is optional.
-t Specifies the duration to run the Smartassist tool. The duration displayed is in seconds. The default value is 3600 seconds. This is optional. -d Specifies virtual (volumes) or physical devices (disks) as targets for Smartassist tool. -o Provides information about the directory in which the traced data is stored. --app Specifies database instance as a target for Smartassist tool.
It accepts different arguments depending upon the database type as mentioned below:1. Oracle: Oracle_user_name Specifies Oracle database username. ORACLE_HOME Specifies users home directory name. ORACLE_SID Specifies Oracle database instance name. ORACLE_ASMDG Specifies Oracle ASM diskgroup name. It should be specified when Oracle is configured on the ASM disk group. This is optional.
2. Sybase:DB_user_name Specifies Sybase database user/login name. DB_server Specifies Sybase dataserver name. DB_name Specifies Sybase database name.
3. DB2:DB_user_name Specifies DB2 database user name. DB_user_home Specifies DB2 database user home directory. DB_name Specifies DB2 database name. -g Specifies disk group or volume group as a target for Smartassist tool. -m Specifies file system mount point as a target for Smartassist tool. Note: The -d, -app, -g, and -m options are mutually exclusive.
-i Specifies input directory name. -c Specifies user specified cache size. This is optional. cache-latency Specifies cache device latency. The default value is 0.05 milliseconds. This is optional. data-latency Specifies data device latency. The default value is 5 milliseconds. This is optional.
You want run the tool on DB2 database instance (dtbs2) for 2 hours (7200 sec) and store the traced output in the /tmp/smdata directory.Enter the following command to analyze the traced output:
Enter the following command to start Smartassist: # ./smartassist start -t 7200 -o /tmp/smdata --app db2 db2inst3 /home/db2inst3/ dtbs2
# ./smartassist analyze -i /tmp/smdata
You can also specify cache size, cache-latency, disk-latency analyze as:
# ./smartassist analyze -i /tmp/smdata -c 2g --cache-latency 0.5 --data-latency 5
The output of the above command is displayed as follows:
---------------------------------------------------------
-------------------
SF SmartAssist tool
-------------------
Details of Sampled Data:
------------------------
Duration : 2.00 Hours
No. Of Reads : 146364
Caching device latency : 0.50 ms
Data device latency : 5.00 ms
DB2 Database User : db2inst3
DB2 Database Name : dtbs2
Statistics for User specified Cache Size:
-----------------------------------------
Cache Size : 2.000000 GB
Read Hits : 146071
Read Hit(%) : 99.80 %
Average Latency gain : 4.49 ms
Latency gain(%) : 89.82 %
Forecast : 9.82x times faster reads
-----------------------------------------
Optimal Cache Size suggested for the workload:
----------------------------------------------
Cache Size : 0.01 GB
Read Hits : 145533
Read Hit(%) : 99.43 %
Average Latency gain : 4.49 ms
Latency gain(%) : 89.77 %
Forecast : 9.77x times faster reads
----------------------------------------------
Per volume breakup:
-------------------
Device Share in Cache
----------------------------------------------
/dev/dm-1 0.01 GB
----------------------------------------------------------
Veritas InfoScale 7.1 SmartIO for Solid-State Drives Solutions Guide
VxVM 7.1 | smartassist(1M) |