Pending time overflow
Because all overflows for pending time are lost, RMF does not accurately report certain shared DASD delays. For example, a request delayed for 18 seconds overflows twice; 16.6 seconds are lost. To RMF, the delay appears to be only 1.4 seconds. Therefore, the AVERAGE PENDING TIME and the AVERAGE RESPONSE TIME values are extremely inaccurate.
For requests with extremely long delays, the missing interrupt handler (MIH) halts the request and reschedules it periodically. MIH estimates the amount of pending time, based on the MIH interval, and adds it to the value RMF reports. Therefore, pending time is lost only for requests that take longer than 8.3 seconds and less than 1.5 times the MIH interval. To increase the accuracy of AVERAGE PENDING TIME and AVERAGE RESPONSE TIME, decrease the MIH interval. An interval of four seconds will ensure that no pending time is lost. However, some performance penalty does occur because of the four-second interval.
For DASD devices actually used for synchronous I/O the Synchronous I/O Device Activity report shows detailed IBM zHyperLink activity data:
The reports for communication equipment, character reader devices, graphic devices, and unit record devices have the same format. The Communication Equipment Activity report is shown as example in Figure 3.
The following figure shows the Magnetic Tape Device Activity report.
|IODF = xx||The IODF number where xx is the suffix of the IODF data set name.|
|CR-DATE: mm/dd/yyyy||The creation date of the IODF.|
|CR-TIME: hh.mm.ss||The creation time of the IODF.|
|ACT: text||The configuration state where text indicates how the IODF was activated.|
|STORAGE GROUP||The name of the storage group to which the device belongs. Your storage administrator assigns the names. These names are available on the direct access device report only.|
|DEV NUM||The five-digit hexadecimal device number that identifies a physical I/O device. The first digit represents the ID of the subchannel set to which the I/O device is physically configured.|
|DEVICE TYPE||The device type on which the data set resides.|
|NUMBER OF CYL||The DASD volume capacity (in cylinders).|
|VOLUME SERIAL||The volume serial number (for direct access and magnetic tape reports) of the volume mounted on the device at the end of the reporting interval.|
|PAV||The number of parallel access volumes (base and alias) which
were available at the end of the reporting interval.
If the number has changed during the reporting interval, it is followed by an '*'.
the device is is a HyperPAV base device, the number is followed by
an 'H', for example,
|LCU||The number of the logical control unit (LCU) to which the device
An LCU is a set of devices attached to the same physical control unit (or a group of physical control units with one or more devices in common.) The IOP, which is part of the channel subsystem, manages and schedules I/O work requests.
There are two reasons that this field is blank:
|DEVICE ACTIVITY RATE||The rate at which start subchannel (SSCH) instructions to the device completed successfully.
This formula applies to the activity rate measured during asynchronous I/O processing. For devices using suspended channel programs, resume I/O requests are included in the SSCH counts.
Character ‘S’ appended to the DEVICE ACTIVITY RATE value of a device shown in the Direct Access Device Activity report indicates that the device performed synchronous I/O requests and that detailed synchronous I/O performance measurements for this device are available in the Synchronous I/O Device Activity report section.
For easy comparison the Synchronous I/O Device Activity report lists the asynchronous I/O device activity rate calculated (ASYNCH I/O) in adjacent columns showing
The synchronous I/O activity rate is calculated as
In the LCU summary line, this field contains the sum of the rates for each individual device.
If the device has been deleted during the last interval, DEVICE DYNAMICALLY DELETED appears in the field instead of the measurement data.
If the device has changed from static to dynamic, or was deleted and a new device added with the same device number, DEVICE DYNAMICALLY CHANGED appears in the field instead of the measurement data.
|AVG RESP TIME||The average number of milliseconds the device required to complete an asynchronous I/O request. This value reflects the total hardware service
time and the front end software queuing time involved for the average I/O request to the device. The
channel measures active time, which starts at the acceptance of a SSCH instruction (indicated by a
condition code 0) and ends at the acceptance of the channel end (primary status pending). It does
not, however, include the time required to process the interruption. The IOS queue length is
factored in to reflect the front end queuing time.
The active time is the sum of connect, disconnect, and pending time as described later.
In the LCU summary line, this field contains the weighted average of the individual average response times for each device.
For easy comparison the Synchronous I/O Device Activity report lists the asynchronous I/O average response time calculated (ASYNCH I/O) in adjacent columns showing
|AVG SYNCH I/O TRANSFER RATE||
|% REQ SUCCESS||Percentage of synchronous I/O requests that completed successfully.|
|% LINK BUSY||Percentage of synchronous I/O requests that hit a link busy condition when trying to use a synchronous I/O link.|
|% CACHE MISS||Percentage of synchronous I/O read requests that hit a cache miss condition.|
|AVG IOSQ TIME||The average number of milliseconds an I/O request must wait on an IOS queue before a SSCH
instruction can be issued.
|AVG CMR DLY||The average number of milliseconds of delay that a successfully
initiated start or resume function needs until the first command is
indicated as accepted by the device. It allows to distinguish between
real H/W errors versus workload spikes (contention in the fabric and
at the destination port).
|AVG DB DLY||The average number of milliseconds of delay that I/O requests
to this device encountered because the device was busy. Device busy
|AVG INT DLY||The average interrupt delay time in units of
milliseconds encountered for I/O requests to this device. For each
I/O request, the time is measured from when the I/O operation is complete
to when the operating system begins to process the status.
|AVG PEND TIME||The average number of milliseconds an I/O request must wait
in the hardware. This value reflects the time between acceptance of
the SSCH function by the channel subsystem (SSCH-function pending)
and acceptance of the first command associated with the SSCH function
at the device (subchannel active). This value also includes the time
waiting for an available channel path and control unit as well as
the delay due to shared DASD contention.
If the value is high, refer to the device's LCU entry in the I/O queuing activity report for an indicator of the major cause of the delay.
|AVG DISC TIME||The average number of milliseconds the device was disconnected
while processing an SSCH instruction. This value reflects the time
when the device was in use but not transferring data. It includes
the overhead time when a device might disconnect to perform positioning
functions such as SEEK/SET SECTOR, as well as any reconnection delay.
The measurement event count is the same as the number of SSCH instructions issued, unless there has been a timer overflow error in the channel.
|AVG CONN TIME||The average number of milliseconds the device was connected
to a channel path and actually transferring data between the device
and central storage. Typically, this value, measures data transfer
time but also includes the search time needed to maintain channel
path, control unit, and device connection.
|% DEV CONN||The percentage of time during the interval when the device
was connected to a channel path.
|% DEV UTIL||The percentage of time during the interval when the device
was in use. This percentage includes both the time when the device
was involved in I/O operations (connect and disconnect time) and the
time when it was reserved but not involved in an I/O operation.
percentage reported represents the time during the interval when the
|% DEV RESV||The percentage of time during the interval when a shared device
was reserved by the processor on which RMF was started.
At each RMF cycle, RMF checks to see if a device is reserved, and a counter is kept of all such samples. At the end of the interval, the percentage is computed.
|AVG NMBR ALLOC||The average number of data control blocks (DCBs) and access
method control blocks (ACBs) concurrently allocated for each volume.
This field is reported only for direct access storage devices.
At each RMF cycle, a counter is increased to reflect the number of data sets concurrently allocated. At the end of the interval, the average is calculated by dividing the total number of allocated data sets for all samples by the total number of samples.
|% ANY ALLOC||The percentage of time during the reporting interval when the
device was allocated to one or more data sets. Permanently mounted
direct access devices show a 100% allocation, regardless of whether
or not a data set was actually allocated.
To determine the value, RMF keeps a count of whether or not the device was allocated or permanently resident at each cycle. At the end of the interval, the percentage is computed.
|% MT PEND||The percentage of time during the interval when a mount was
pending for the device. This field is reported only for direct access
devices and magnetic tape devices.
At each cycle, RMF updates a counter when it detects a mount pending condition. At the end of the interval, the percentage is computed.
|%NOT RDY||The percentage of time during the reporting interval when the
device was not ready for use. For example, when a tape has just been
mounted but is not yet ready to be used to the system. This field
is not reported for direct access devices. However, the value is recorded
in the corresponding field of the SMF record, should your installation
need the information.
At each RMF cycle, a counter is updated when the status of the device indicates that it is not ready. At the end of the interval, the percentage is computed.
|NUMBER OF MOUNTS||The number of tape mounts, shown as an integer value, detected
If the tape mount was pending at the first cycle of the interval, an asterisk is placed before the numerical value of the tape mount. If the tape mount was pending at the last cycle of the interval, an asterisk is placed immediately following the numerical value of the tape mount.
If a mount-pending condition is detected at the first cycle of the interval, the mount count for the interval increments by one.
In the LCU summary line, this field contains the sum of all mount counts.
This field is reported only for magnetic tape devices.
Note: Due to the fact that the tape mount count is a sampled value, it might happen that it does not contain all subsecond mounts of VTS devices.
|AVG MOUNT TIME||The average mount time pending for every device, expressed
in the form of HH:MM:SS.
If the mount count or the sample count is zero, the result is zero.
This field is reported only for magnetic tape devices.
|TIME DEVICE ALLOC||The total time the device was allocated during the interval,
expressed in the form of HH:MM:SS.
If the sample count is zero, the result is zero.
This field is reported only for magnetic tape devices.