FCX110, I/O Device Details Screen – DEVICE

Performance Toolkit Subcommand: DEVICE.

The I/O Device Details Screen (FCX110) will be shown if you specify a devno on the DEVICE subcommand, or by pressing ENTER after placing the cursor on the device number field on the General I/O Device Screen (FCX108), CP-Owned Device Screen (FCX109), or CU-Cached Disks Screen (FCX111).

This screen gives detailed information on the load of a single I/O device. Although intended primarily for analyzing the I/O load to different minidisks on SYSTEM-attached disks (online SEEKS analysis), it can also give some additional information for other devices.

Figure 1. Layout of I/O Device Details Screen (FCX110)

 FCX110      CPU nnnn  SER nnnnn  Interval HH:MM:SS - HH:MM:SS    Perf. Monitor

 Detailed Analysis for Device 0145 ( CP OWNED )
 Device type :  3380-A      Function pend.:    4.8ms      Device busy   :   23%
 VOLSER      :  VMPG1C      Disconnected  :   38.5ms      I/O contention:    7%
 Nr. of LINKs:      41      Connected     :    7.3ms      Reserved      :    0%
 Last SEEK   :     524      Service time  :   50.6ms      SENSE SSCH    :    0
 SSCH rate/s :     4.6      Response time :   54.1ms      Recovery SSCH :    0
 Avoided/s   :      .0      CU queue time :     .0ms      Throttle del/s:  ...
 Status: SHARABLE

 System Page/Spool I/O Details
 Page reads/s  :   6.2      Total pages/s   :   10.9      PG serv. time:  4.8ms
 Page writes/s :   4.7                                    PG resp. time:  4.8ms
 Spool reads/s :    .0      System I/Os /s  :    3.4      PG queue len.:  .00
 Spool writes/s:    .0      User interfer./s:     .0      Avail. bsize :   14

 Path(s) to device 0145:    01    41
 Channel path status   :    ON    ON

 Device            Overall CU-Cache Performance           Split
 DIR ADDR VOLSER   IO/S %READ %RDHIT %WRHIT ICL/S BYP/S   IO/S %READ %RDHIT
 8A  0145 VMPG1C    5.5    54      0      0    .0    .0   'NORMAL' I/O only

         MDISK Extent       Userid   Addr IO/s VSEEK Status    LINK MDIO/s
 +-------------------------------------------------------------------------+
 |       0071 -  0092       SNAMAINT 03CF   .0     0 WR           1     .0 |
 |       0189 -  0204       SNAMAINT 06C3   .0     0 WR           1     .0 |
 |       0205 -  0228       SNAMAINT 06C4   .0     0 WR           2     .0 |
 |       0289 -  0298       SNAMAINT 07C2   .0     0 WR           1     .0 |
 |       0299 -  0299       CHGCS    0191   .0     0 RR           6     .0 |
 +-------------------------------------------------------------------------+
 |       0300 -  0599       SYSTEM PAGE     RD/s  WR/s  MLOAD  Used   IO/s |
 |                          LOAD   ====>     6.2   4.7    6.1   47%   .... |
 +-------------------------------------------------------------------------+
 |       0600 -  0619       SNAMAINT 07C1   .0     0 WR           1     .0 |
 |       0620 -  0664       CHVSCS   0193   .0     6 RR           4     .0 |
 |       0821 -  0884       MAINT    0A95   .0       owner                 |
 |                          XAACNT   0195  1.2     0 WR           1    1.2 |
 +-------------------------------------------------------------------------+

 Command ===> _
 F1=Help  F4=Top  F5=Bot  F7=Bkwd  F8=Fwd  F12=Return

The screen in the above figure is a sample for the type of display you can expect when analyzing a disk containing active minidisks. It shows performance data for a 3380 disk which is connected to 3880-23 controllers, but for which caching has been disabled. Basically, the information shown is grouped into up to five blocks:

A first block containing general device performance data. This block will always be shown, but its layout depends on the device type being analyzed.
A second block with details on system I/O (to paging and spooling areas on the device, for example). Shown only when some system I/O activity has ever taken place.
A third block containing information on the data path(s) available to the device. Always shown, always same layout regardless of device type.
A fourth block containing information on the use made of the control unit cache. This information is actually extracted from FCX111, CU-Cached Disks Screen - CACHDBSE, and it can be shown only when the necessary prerequisites are met: I/O domain of CP MONITOR active, and data extraction from CP MONITOR enabled for the Performance Toolkit. See CACHDBSE for a more detailed description of these requirements.
A fifth block containing data for minidisk load analysis. This block will be shown only when SYSTEM-attached or -owned disks are analyzed which either contain minidisks that have been linked by a logged-on user, or which contain system areas (paging, spooling, or dump areas, for example). The line No active MDISKs on this volume - no SEEKs analysis possible will be displayed instead for all other disks.
Note that SEEKs analysis is possible, of course, but only based on CP monitor data: you will have to select one of the specific DASD seeks displays to obtain information on the distribution of I/O activity for such a disk (see SEEKDET for more information).

The minidisk load analysis part of the data may extend beyond the number of lines that can be shown simultaneously on the screen, in which case the FORWARD and BACKWARD subcommands (or the corresponding PF-keys) can be used to view the remaining data.

The example shown is for a disk which is shared with another VM system; the minidisk cache cannot be used for I/O to minidisks on this pack.

Where use of the minidisk cache is possible, the minidisk load analysis part of the screen has a slightly different layout to allow displaying information on minidisk cache efficiency:

         
MDISK Extent       Userid   Addr IO/s VSEEK Status    LINK  VIO/s %MDC
+------------------------------------------------------------------------------
|       0051 -  0060       BGF      0191   .0     0 WR           1     .0  ...
|       0121 -  0170       CAR      0319  6.2     0 RR 
|                          KSC      0319  3.7     0 RR 
|                          BERS     0319  1.4     0 RR 
|                          ALO      0319   .8     0 RR 
|                          FBL      0319   .5     0 RR         814   24.6   99
|       .... -  ....       ....

Note that in this case, the information extends beyond the 80 columns of an ordinary screen. Use the LEFT/RIGHT subcommands, or the corresponding PF-keys, to shift the display window to view the remaining data.

Field Description for General Device Performance Data:

Device type

Real device type and model of the device found on the address being analyzed

SSCH rate/s

Average number of SSCH and RSCH operations per second during the last measuring interval. Please note that the total number of SSCHs executed per second on a disk will usually not match exactly the total of the I/O rates displayed for all minidisks on the pack. This inconsistency is to be expected because:

The values come from different sources and cannot be extracted at exactly the same time
Some of the minidisks which contributed to the total may have been detached, or their users logged off, in the mean time, and
I/O activity on CP areas (paging, spooling, etc.) cannot be shown.

Status

General device status information. The status indicated can be

RESERVE HELD: A RESERVE to the real DASD is held from this system
RESERVE PEND: A RESERVE to the real DASD is pending from this system
RELEASE PEND: A RELEASE to the real DASD is pending
MDCACHE USED: Minidisk caching is used for at least one minidisk on this device
SHARABLE: The disk has been defined as 'SHARED' in the HCPRIO or configuration file, or by command
HOT I/O DEV.: Hot I/O device
INTERV. REQD.: Operator intervention is required on the real device
RECORDING: Intensive recording mode is active for the device
DEV. OFFLINE: The device is OFFLINE
PATH OFFLINE: All path(s) to the device are OFFLINE
ONLINE: Device ONLINE - no special status to be indicated. This status will be inserted only when no other status applies.

for DASD only:

VOLSER: Disk label
Nr. of LINKs: The total number of LINKs to minidisks found on the disk pack. As for the total I/O rate above, this number may differ slightly from the LINK total found when adding up all the LINKs to each minidisk.
Last SEEK: The address of the last real SEEK executed for the disk. For FBA or SCSI disks, this is the block or address on which that last emulated I/O operation was performed.
Avoided/s: The rate at which I/O requests could be satisfied via the minidisk cache, without execution of real I/O operations.

for GRAF devices only:

Features

Indicates extended data stream features available on the device. Features can be

C: Device has extended color support
H: Device has extended highlighting support
PS: Device has programmed symbols support

Screen stat.

Special GRAF status information. The status indicated can be

DISA: Disabled
DISASQ: Disabling sequence
DROP: DROP processing
LOGON: User logged on
ENA: Enabled

for high speed devices only:

Function pend.

The average time, in milliseconds, during which an I/O operation remained pending in the I/O subsystem during the last measuring interval due to path busy conditions.

This value, and also the next five values, is calculated based on data from the 'host measurement facility'. Dots will be inserted for devices where this information is not available.

Disconnected

The average time, in milliseconds, during which the device remained disconnected while handling an I/O operation. Abnormally high 'disconnected' times (as in the example shown) could be caused by an overloaded channel path which did lead to many re-connect misses, or, for DASD, by long SEEKs due to I/O to data at opposite ends of the device.

Connected

The average time, in milliseconds, during which the device was connected to the channel path while handling an I/O operation (receiving commands or transferring data). This time also includes SEARCH time for DASD.

Service time

The average service time of the device, i.e. the average time required by the device for executing an I/O operation (in milliseconds). The service time is an indication of device performance only, it does not include queueing delays.

Response time

The average I/O response time for the device. The I/O response time is the service time plus any queueing delays, it is calculated from the service time and the average I/O request queue length.

CU queue time

The time when the device was logically disconnected from the channel subsystem while it was busy with an I/O operation initiated from another system.

The field will contain valid data only if the control unit queuing measurement facility is installed.

Device busy

Percent of elapsed time during which the device was busy with an I/O operation. This value is calculated based on the I/O rate of the device and it's service time, using the data provided by the 'host measurement facility'.

I/O contention

The smoothed relative I/O queue length, in percent. The value is based on the number of IORBKs found queued during the last few measurements. (see DEVICE).

This value should be very low, it indicates an I/O performance problem because of an overloaded device (too many users having high-activity minidisks allocated on the same real device) if it is often high.

Please note that the contention is calculated based on one sample per measurement interval. The resulting contention figures should be taken as rough indicators only.

Reserved

Percentage of measurements during which the device was found to have been 'reserved' by real reserve/release logic. This value is a smoothed average, based on the last few samples, and it is inserted only for DASD.

SENSE SSCH

The number of 'SENSE' SSCH operations during the last measuring interval. This should normally be zero: SENSE operations are usually performed when a previous I/O operation has terminated with an error condition (unit check).

Recovery SSCH

The number of recovery SSCH operations during the last measuring interval.

Throttle del/s

The rate at which I/Os to the device were delayed due to throttling (applicable only when throttling set on).

Field Descriptions for System Page/Spool I/O Details:

These fields describe the CP I/O activity to system areas (e.g. paging and spooling areas), and to minidisks which have been mapped for paging I/O to VM data spaces. The fields shown and their meaning are

Page reads/s

The number of pages read per second

Page writes/s

The number of pages written per second

Spool reads/s

The number of spool pages read per second

Spool writes/s

The number of spool pages written per second

Total pages/s

The total number of pages moved per second

System I/Os /s

The number of actual SSCH or RSCH operations per second for system areas

User interf./s

The rate at which user-generated I/Os caused interference with system-generated I/O activity (system I/O program could not be resumed)

PG serv. time

The smoothed service time per PIOBK (page I/O CCW package), in milliseconds. Continuously updated by CP after each I/O.

PG resp. time

The smoothed response time per PIOBK. Based on the smoothed service time, but taking into account also the PIOBK queue length. CP uses this value when selecting suitable disks for the next paging or spooling write operation.

PG queue len.

The smoothed PIOBK queue length. This value is determined by sampling the actual current queue length, and then smoothing it according to the formula

  (15*old value + current value) / 16

Avail. bsize

The average number of consecutive free slots found during the last measuring interval. Dots will be inserted when no new slot sizes have been determined in an interval.

Field Description for Path Information:

The path information section shows the channel paths by which the I/O device is connected to the system, based on the specifications in the IOCP.

For the emulated FBA devices (EDEV), only the first path ID is shown, even when multiple channel connections to the device are available.

The format is:

Path(s) to device nnnn:    p1    p2   p3   p4
Channel path status   :   xxx   xxx  xxx  xxx

where:

nnnn: Is the real device number
p1 p2 ..: Are the path numbers of all generated paths to the device
xxx: Is the channel path status. It can be either 'ON' for 'ONLINE', or 'OFF' for 'OFFLINE'.

Field Descriptions for Cache Usage:

The output shown is an extract of the data displayed on FCX111, CU-Cached Disks Screen - CACHDBSE. Refer to CACHDBSE for the corresponding field descriptions.

Please be aware that cache performance data are asynchronously collected by the CP MONITOR I/O domain. Since the data collect interval of the CP MONITOR is different from the one used for collecting detailed device data, the I/O rates shown will usually be different too. See also the Usage notes for FCX111, CU-Cached Disks Screen - CACHDBSE for further guidelines.

Field Descriptions for Minidisk Load Analysis:

Minidisks found on the device which belong to, or are linked by currently logged on users will be displayed in ascending sequence of start cylinder or block. Minidisks for which any I/O activity was found during the last measuring interval will be highlighted.

Since both the total I/O count to each minidisk and also the individual contribution to this total by individual users are shown, this display can be used as a very powerful tool for detecting the reason for high I/O contention on a specific device.

MDISK Extent

Start and end cylinders for minidisks on CKD disks. For minidisks on FBA devices the numbers of the first and last block of the minidisk are inserted instead.

Userid

User identifications of users linked to the minidisk. The first entry shown for any minidisk is always the minidisk owner, regardless of I/O activity. Where other users are linked to the same minidisk, up to four of them will be shown if they have contributed to the minidisk's I/O activity in the preceding measuring interval, in descending order of I/O activity.

Addr

Virtual minidisk address

IO/s

Number of I/O requests per second to this particular minidisk (to this virtual address of this user)

VSEEK

Last virtual SEEK address. The seek address has only little meaning for CMS minidisks where the user usually does not know where specific files have been allocated, and where the address may not even be updated for normal CMS file I/O. It provides an indication on the dataset being processed if the minidisk is used by a VSE or MVS system.

Status

Minidisk status information. The status information consists of two parts. The first part indicates how the minidisk has been linked by the user: the characters 'WR' are shown for minidisks linked in write mode and the characters 'RR' for minidisks in read-only mode (always shown). The string 'owner' will be inserted instead for the first entry of minidisks which are in use, but whose owner does not have them linked.

The second part of the status information is inserted only if the minidisk was found to be in a special status at the end of the measuring interval. The status indicated can be:

BUSY: Indicates that the minidisk was busy with an I/O operation when the last sample was taken
SUSP: Indicates that an I/O operation has been suspended
RPEND: Indicates that a 'resume SSCH' operation was pending
QUED: Indicates that an I/O operation was queued for the minidisk
RS/RL: Indicates that the minidisk has been defined in the directory with virtual reserve/release support
TDISK: Indicates that the minidisk has been defined as a temporary minidisk

Only the first status found (from top to bottom of the above list) will be inserted, even if more than one status applies.

LINK

Number of LINKs to the minidisk

VIO/s

Total number of virtual I/O requests per second to the minidisk. This field is shown only for disks which can use the minidisk cache.

%MDC

Minidisk cache efficiency, shown as the percentage of virtual I/O requests which did not require a real I/O operation. This field is shown only for disks which can use the minidisk cache; dots will be inserted when the value cannot be calculated because the virtual I/O rate is 0.

Although not absolutely accurate (there is no fixed 1:1 relation between virtual and real I/O operations to a disk, especially to non-CMS minidisks, even if no minidisk cache is available), this value should be a good indicator of minidisk cache efficiency for a specific CMS minidisk.

MDIO/s

Total number of real I/O requests per second to the minidisk

Cache-eligible minidisks:

If a minidisk has been defined in the directory as eligible for caching in the expanded storage minidisk cache (the default, valid unless a 'CACHOPT' statement specifying 'NOMDC' follows the minidisk entry), the character 'C' will be used as left and right delimiter instead of the standard '|'.

If the minidisk cache is used for a minidisk (less real I/Os measured than there were virtual I/Os, i.e. %MDC value > 0), these delimiters and the extent information for the minidisk will be highlighted.

Additional fields for system areas on the disk:

System area fields are separated from each other and from general user minidisk entries by a dashed line.

The layout for system areas is as follows:

+----------------------------------------------------------------+
|  0412 - 0471    SYSTEM  type    RD/s  WR/s  MLOAD  Used   IO/s |
|                 LOAD   ====>     4.1   2.3   16.5   74%    6.4 |
+----------------------------------------------------------------+

SYSTEM type

The area type can be:

DIRECT: For CP directory space (allocation DRCT)
PAGE: For a paging area (allocation PAGE)
SPOOL: For a spooling area (allocation SPOL)
TDSK: For a temporary minidisk area (allocation TDSK)

The following three fields contain meaningful data only if the system area is of the type PAGE or SPOOL:

RD/s: Page or SPOOL read rate during the last interval (number of PIOBKs read per second)
WR/s: Page or SPOOL write rate during the last interval (number of PIOBKs written per second)
MLOAD: The smoothed response time for CP I/O operations to this area, in milliseconds (msec).
The value is normalized to reflect the response time for servicing a single page. It also takes into account the PIOBK queue length for the device.
Used: Percentage of blocks (PAGE and SPOOL areas) or of cylinders (DIRECT and TDSK) currently in use in this area.
IO/s: Information for calculating I/O rates to system areas directly is not available. The I/O rate will be inserted for SPOOL areas as the sum of the read and write request rate, and for PAGE areas as the difference between the sum of all read and write requests and the number of requests which could be appended to an existing channel program without a separate SSCH/RSCH. Where more than one PAGE or SPOOL area exists on the same disk pack, the load information is inserted for the first area only (no information is available to tell how the load is split). Dots are inserted for the I/O rate in further PAGE or SPOOL areas, and for all other types of system areas.
See also the system I/O details section of this screen for information on I/O handling to system areas on the disk.

For temporary minidisk areas you will see both a system area entry, specifying the total size of the TDSK space available, and ordinary minidisk entries for each of the temporary minidisks which are currently defined.

Effect of AVERAGE and INTERIM subcommands: None. Averages cannot be shown for the detailed I/O device display.