Job Management Partner 1/Performance Management - Agent Option for Platform Description, User's Guide and Reference

[Contents][Glossary][Index][Back][Next]

Device Summary (PI_DEVS)

Function

The Device Summary (PI_DEVS) record stores the performance data in which the data stored in the Device Detail (PI_DEVD) record is summarized per unit time. Note that even if the Device Detail (PI_DEVD) record is not collected, performance data is stored in the Device Summary (PI_DEVS) record.

Notes:
  • To collect the Device Summary (PI_DEVS) record in AIX, execute the smitty chgsys command and set Continuously maintain DISK I/O history to true.
  • In Linux, the performance data of only the devices that have the following major numbers specified in the following file are collected from /proc/partitions and summarized in the Device Summary (PI_DEVS) record: 3, 8, 22, 33, 34, 56, 57, 65 to 71, 88 to 91, and 128 to 135
    [Figure] Linux 4: /usr/share/doc/MAKEDEV-3.15/devices.txt
    [Figure] Linux 5: /usr/share/doc/MAKEDEV-3.23/devices-2.6+.txt
    For major numbers, see the block devices in the above file that match the contents of the Device Name(DEVICE_NAME) field.
  • If the DR function of Solaris 9 or later is used to modify a system resource while a PFM - Agent for Platform service is running, some types of performance data cannot be correctly collected. The table below shows the types of system resources that are affected and the actions that you can take to correctly display reports after the modification.
    System resource type Report type Actions to be taken after modification
    Device Real-time Display the report again.
    Historical Specify the display period such that it does not include the date and time when the system resources were modified.
    Fields other than those listed below are affected by a system resource modification:
    [Figure] Interval (INTERVAL)
    [Figure] Record Time (RECORD_TIME)
    [Figure] Record Type (INPUT_RECORD_TYPE)
    [Figure] Fields that are added only when data is recorded in the Store database
    For details about the notes on modifying system resources, see Notes on performance information when modifying system resources under Notes on records in this chapter.
  • Only the local disk devices of Persistent DSF are collected in HP-UX 11i v3 (IPF).

Default values and values that can be specified

Item Default value Modifiable
Collection Interval 60 Yes
Collection Offset 0 Yes
Log No Yes
LOGIF Blank Yes

ODBC key fields

None

Lifetime

None

Record size

Fields

PFM-View name
(PFM-Manager name)		 Description Summary Format Delta Not Supported versions Data source
Avg Service Time/device(SERVICE_TIME) Average service time per device (seconds).
In HP-UX, this value indicates the total service time of all I/Os, and therefore, it may significantly exceed the value of Interval when processes are continuously executed on a device. Even with other OSs, the value may exceed the value of Interval when processes are continuously executed on a device.		 AVG utime No -- TOTAL_SERVICE_TIME / DEVICE_COUNT
Avg Service Time/op(AVG_SERVICE_TIME) Average I/O service time for a device (seconds) AVG utime No -- TOTAL_SERVICE_TIME / TOTAL_IO_OPS
Busy %(BUSY_PERCENT) Average percentage of time a disk was busy (%).
This value may exceed 100 when processes are continuously executed on a device.		 % float No -- ((TOTAL_BUSY_TIME / INTERVAL) / DEVICE_COUNT) x 100
Devices(DEVICE_COUNT) Device count AVG ulong No -- Number of Device Detail (PI_DEVD) records of the category "disk" summarized in the Device Summary (PI_DEVS) record
I/O Mbytes(TOTAL_IO_MBYTES) Total size of data transferred through I/O operations (MB) AVG double Yes -- --
Interval(INTERVAL) Interval at which the Device Summary (PI_DEVS) record was stored (seconds) COPY ulong Yes -- RECORD_TIME - previous-record-time; for real-time, RECORD_TIME - last-boot-time
Mbytes Xferd/sec(MBYTES_TRANSFERRED_PER_SECOND) Average I/O speed (MB per second) R float Yes -- TOTAL_IO_MBYTES / INTERVAL
Queue Length(QUEUE_LENGTH) Average device queue length. The amount of data processed by an I/O in 1 second is given a value of 1. AVG ulong No AIX sum-of-queue-lengths-of-all-devices / DEVICE_COUNT
Read Ops %(READ_OPS_PERCENT) Read operations as a percentage of all I/O operations (%) % float No HP-UX,
AIX		 TOTAL_READ_OPS / TOTAL_IO_OPS x 100
Reads(TOTAL_READ_OPS) Number of read operations that occurred AVG ulong Yes HP-UX,
AIX		 --
Reads/sec(READ_OPS_PER_SECOND) Frequency at which read operations occurred (number of read operations per second) R float Yes HP-UX,
AIX		 TOTAL_READ_OPS / INTERVAL
Record Time(RECORD_TIME) Time at which the record was created (Greenwich Mean Time) COPY time_t No -- --
Record Type(INPUT_RECORD_TYPE) Record type. Always DEVS. COPY char(8) No -- --
Seek Ops(SEEK_OPS) Number of seek operations AVG ulong Yes Solaris,
AIX,
Linux		 --
Total Busy Time(TOTAL_BUSY_TIME) Total busy time for a device (seconds).
This value may exceed the value of Interval x Devices when processes are continuously executed on a device.		 AVG utime Yes -- --
Total I/O Ops(TOTAL_IO_OPS) Number of I/O operations that occurred AVG ulong Yes -- TOTAL_READ_OPS + TOTAL_WRITE_OPS
Total I/O Ops/sec(TOTAL_OPS_PER_SECOND) Frequency at which I/O operations occurred (number of operations per second) R float Yes -- TOTAL_IO_OPS / INTERVAL
Total Service Time(TOTAL_SERVICE_TIME) Total service time for a device (seconds). This value includes wait time.
In HP-UX, this value indicates the total service time of all I/Os, and therefore, it may significantly exceed the value of Interval x Devices when processes are continuously executed on a device. Even with other OSs, the value may exceed the value of Interval when processes are continuously executed on a device.		 AVG utime Yes -- --
Total Wait Length Time(TOTAL_WAIT_LEN_TIME) Total length of time I/Os waited for a device (seconds). The processing capacity of the I/Os that were waiting is divided by the I/Os' processing capacity per second, and is then integrated over the wait time. AVG utime Yes AIX --
Total Wait Time(TOTAL_WAIT_TIME) Total wait time for a device (seconds).
In HP-UX, this value indicates the total wait time of all I/Os, and therefore, it may significantly exceed the value of Interval x Devices when processes are continuously executed on a device.		 AVG utime Yes AIX --
Wait Length Time(WAIT_LEN_TIME) Average length of time an I/O waited for a device (seconds). The processing capacity of the I/Os that were waiting is divided by the I/Os' processing capacity per second, is integrated over the wait time, and is then divided by the device count. AVG utime No AIX TOTAL_WAIT_LEN_TIME / DEVICE_COUNT
Wait Time(WAIT_TIME) Average wait time for a device (seconds).
In HP-UX, this value indicates the total wait time of all I/Os, and therefore, it may significantly exceed the value of Interval when processes are continuously executed on a device.		 AVG utime No AIX TOTAL_WAIT_TIME / DEVICE_COUNT
Write Ops %(WRITE_OPS_PERCENT) Write operations as a percentage of all I/O operations (%) % float No HP-UX,
AIX		 TOTAL_WRITE_OPS / TOTAL_IO_OPS x 100
Writes(TOTAL_WRITE_OPS) Number of write operations that occurred AVG ulong Yes HP-UX,
AIX		 --
Writes/sec(WRITE_OPS_PER_SECOND) Frequency at which write operations occurred (number of write operations per second) R float Yes HP-UX,
AIX		 TOTAL_WRITE_OPS / INTERVAL

[Contents][Back][Next]

[Trademarks]

All Rights Reserved. Copyright (C) 2009, Hitachi, Ltd.