• Aucun résultat trouvé

STATS REPORT FORMAT CONTROL

Dans le document Xerox Control Program-Five (CP-V) (Page 95-114)

When the user requests reports or displays, STA TS refers to snapshot fi les can be generated simultaneously. Commands are provided that enable the user to set and reset the flags.

IT

Interaction time - - - . - - - - .

TT

-1

Task turn-around time User response time

C

Intermittent CPU time due

t;, I/o

~i ~ndqueue-fo;::­

service delay

I~

Remaining

,output

T

Thinking and typing

time-User think

t-

User typing time

-t----.... -

time RT

Response time

L

First time the user types

Input complete

J

Character received at the computer ter-m i nates the ter-message input by the user.

Item Name 1

2

Program activation

L

First output character

Control of the CPU

Intermittent terminal output

is turned over to the program servicing message just received (the reading program).

Read

L

The service program issues a Read to the terminal re-questing input of the next command from the user.

Figure 13. Terminal Interaction Concepts

Table 26. STATS Flogs Description

---.

Input complete The receipt of an acti-

J

vation character initiates.

a task for the system.

A standard display that lists histogram data as for the SYSTEM display group including response time,

~ character out histograms.

interaction time, turnaround time, task time, character in ant

----_._-,._--_._---_. __

..

.. - - - 4

A standard display that includes the same information as stand grams for a preselected shared processor. The shared processo command (described later).

ard display "l' and, optionally, histo-r may be selected by the SETPROC

STATS 79

Table 26. STATS Flags (cont.) Item Name Description

3

A standard display that includes the following display groups {which are described shortly}:

I/O TASK

BATCH ONLINE

(shown in order of appearance in the display)

USERS QUEUE

RESOURCE

4 A standard display that includes the SUMMARY, CPU, and I/O display groups and the single items TIC, UC, OUC, BUC, GUC, and RT90. This is the default display except that the SUMMARY display is omitted.

5

A standard display of the SWAP display group.

6 A standard display that includes the I/O and RAHD groups.

7 A standard display that includes the multiprocessing specific groups, SCPU and EVNT.

ALL A display which includes the single user items

SUMMARY RAHD

CPU EVNT

TASK ONLINE

SCPU QUEUE

(shown in order of appearance in the display)

BATCH SYSTEM

USERS PROC

RESOURCE SWAP

I/O

BATCH A display that lists the percent of CPU time used by each shared processor for batch programs only.

The CPU percentage includes batch execution and service time. (Figure 17 describes this display in detail.)

BUC A display of the number of batch users currently in the system.

CIT A display of the total number of interactions received since start-up.

- - _ ...

_---"---CPU A display group that includes the percent of CPU time for the monitor and for on-line, batch, and ghost users broken down into execution and service time. Monitor service, idle, and swap wait time is also listed for a complete view of CPU uti lization. (Figure 16 describes this display in detai I.)

EVNT A display group of the number of CALs and number of jobs scheduled per minute as measured by the primary CPU and each of the secondary CPUs (SCPU). (Figure 26 describes this display in detai I.)

GUC A display of the number of ghost users currently in the system.

---I/O A display of the I/O rates for the system including symbiont accesses, CALs, and number of inter-actions per sample minute. (Figure 20 describes this display in detai I.)

ONU[NE] A display of the percent of CPU time used by on-line users for each of the shared processors. The CPU percentage includes user execution and service time. (Figure 18 describes this display in detail.) OUC A display of the number of on-line users currently in the system.

PARA[M] A display that includes some control parameters and some other statistics. This print flag adds these control parameters to all subsequent reports. The CONTROL I command may be used to display this group. (Figure 14 describes this display in detai I.)

80 STATS

Item Name PROC

Table 26. STATS Flags (cont.) Description

A display of five histograms for a preselected processor including think-type time, turn-around time, CPU task time, terminal input characters per line, and terminal output characters per line. The processor is preselected by setting C:PROCN via the SETPROC command. (Figure 24 describes this display in detai I.)

1 + . .

-QUEU[E] A display that summarizes the user state queues at the end of e,ach sample interval. (Figure 22 de-scribes this display in detail.)

1 - - - + - - - --- --- --- --- ----

---.----RAHD A display group that describes the read ahead and AIR usage in the system. (Figure 27 describes this display in detail.)

1 - - - f - - -

---REPO[RT] A nondisplay flag that controls whether or not a report is to be generated whi Ie executing the FILE command.

1 - - - 1 - - -

--RESO[URCE] A display group that contains the current usage of certain resources including MPOOLs, CFUs, COCBUFs, and granules (pack, symbiont, and RAD). (Figure 28 describes this display in detail.)

1 - - - ---/---.--.. - - - ---

---RT90 A display that indicates the 90 percent response time value.

- - - + - - -

---SCPU A display group that contains the percent of secondary CPU US<lge for each of the secondary CPUs on the system. These values are broken down into execution and idle time. (Figure 29 describes this display in detail.)

1 - - - + - - - - - - - 1

SNAP[SHOT] A nondisplay flag that controls the generation of snapshot records whi Ie executing the DISPLAY command.

, - - - + - - - - - - - 1

SUMM[ARY] A display that contains key measures of system performance (such as the number of logged users and the measures of response). (Figure 15 describes this display in detai I.)

1 +

-SWAP A display of five histograms including distribution of:

• The number of users per outswap.

• The outswap time.

• The inswap time for user J ITs, shared processors, and overlap.

• The inswap time for the entire user excluding the J IT and shared processors.

• The inswap time for the entire user excluding shared processors and overlap.

(Figure 25 describes this display in detail.)

1 - - - --- ---

---SYS[TEM] A display of seven histograms including distributions for response time, intE~raction time, think-type time, turn-around time, CPU task time, terminal input characters per line, and terminal input char-acters per line. (Figure 23 describes this display in detail.)

1 - - - + - - - ---

---.---TASK A display of task statistics including interaction time, response time, and "ask time. (Figure 21 de-scribes this display in detail.)

1 + . .

-TIC A display of the minutes since system start-up.

1 + .

-UC A display of the number of current users in the system. This number includes batch users, on-line users, ghost jobs, and user's with lines who have not yet logged on.

1 + -USER[S] A display that I ists the number of users for severa I shared processors. The numbers are the counts

existing at the end of the sample interval. (Figure 18 describes this display in detail.)

~---~---.----.---.---STATS 81

i i

82

The PARAM display is printed in the following format:

where PARAM HOUR:MINUTES MAX # BATCH USERS AVERAGE BATCH SIZE K if CHAR TERM BLOCK MSEC

wlo

INTERRUPT AVE BATCH QUANTUM

# CPU ACTIVE RAHD TIMEOUT SEC

INTERVAL IN MINUTES MAX if ONLINE USERS AVERAGE ONLINE SIZE K

# CHAR TERM UNBLOCK MSEC SWAP QUANTUM MSEC ONLINE COMPUTE AVE SCPU QUANTUM MSEC AIR TIMEOUT SEC

HOURS:MINUTES specifies the hour followed (without a separator) by the minutes. This item is useful in se lect i n9 a fi I ter for the Summary processor.

INTERVAL IN MINUTES is the number of minutes in the snapshot interval. This item is useful in selecting a filter for the Summary processor.

MAX # BATCH USERS is the maximum number of concurrent batch users and is the control parameter BUM whi ch may be set by the system manager via the Control processor.

MAX # ONLINE USERS is the maximum number of concurrent on-line users and is the control parameter OUM which may be set by the system manager via the Control processor.

AVERAGE BATCH SIZE K is the current average size of all active batch partitions. This number includes context and excludes shared processors and overlap. (Reference: UB: PCT table.)

AVERAGE ONLINE SIZE K is the current average size of all on-line users. This number includes context and excludes shared processors and overlap. (Reference: UB:PCT table.)

# CHAR TERM BLOCK is the number of characters at which to block terminal output. This is the control parameter TB which may be set by the system manager via the Control processor. This parameter should be set high enough to prevent terminal output-bound users from being swapped too frequently. The faster the terminal transmission, the higher this parameter should be set. If the value is too high, however, the al-located COC buffers may be insufficient to handle the load.

# CHAR TERM UNBLOCK is the number of characters at which terminal output is unblocked and is the con-trol parameter UB which may be set by the system manager via the Concon-trol processor. This parameter causes output-bound users to be queued for CPU service before the COC output buffer for that user is empty. The faster the terminal, the higher this parameter should be set.

MSEC

wlO

INTERRUPT is the amount of uninterrupted compute time guaranteed a user and is the current value of the control parameter QMIN which may be set by the system manager via Control. A low value increases the burst I/O rate. A high value decreases the rate at which the system services different users.

MSEC SWAP QUANTUM is the amount of time a user is guaranteed core residency before swap out. This parameter is the current value of the control parameter SQUAN which may be set by the system manager via Control. Increasing this value decreases the swapping rate. This parameter can be used to reduce the swapper load, but it will be at the expense of user response time.

AVE BATCH QUANTUM is the average setting of the QUAN control parameter for all active batch parti-.

STATS

tions. (QUAN is the time-slice by which computer-bound users are shared.) The QUAN parameter is set for each partition by the system manager via the Control processor.

Figure 14. Description of the PARAM Display

90 16 74H-l(11/76)

MSEC ONLINE COMPUTE is the setting of the QUAN control parameter for on,-Iine users. (QUAN Is the time-slice by which computer-bound users are shared.) The QUAN parameter is set by the system manager via the Control processor. The values of QUAN for on-line and batch and the number of on-line and batch users in the computer-bound queue determine the mix of on-line and batc:h computing that is done.

# CPU ACTIVE is the number of CPUs currently active in the system. This includes both the primary CPU and all secondary CPUs that have been software Initialized and avclilable for sc:heduling.

AVE SCPU QUANTUM MSEC is the average setting of the quantum for all active CPUs. This number is derived from the MAXQ values for each initialized secondary CPU. The MAXQ value is set by the system manager via the Control processor.

RAHD TIMEOUT SEC is the time, in seconds, after which a read ahead operation Is aborted. This value is set by the system manager \Jslng the name RA TO as input to the Control processor.

AIR TIMEOUT SEC is the time in seconds after which an AIR operation is aborted. this value Is set by the system manager using the name AIRTO as Input to the Control processor.

Figure 14. Description of the PARAM Display (cont.)

~---'---The SUMMARY display group is printed in the following format:

where

SUMMARY BATCH STREAMS ONLINE USERS

% CPU/BATCH USER

% CPU/ONLINE USER BATCH EXEC/SERV ONLINE EXEC/SERV CPU MSEC PER I/O ONLINE TIME MIX ONLINE INTENSITY ONLINE TASKS/MIN

% INTERACTIVE 90/0 RESPONSE MSEC TURNAROUND SEC ETMF

BATCH STREAMS is the number of active batch partitions at the end of the snapshot interval. This statistic is avai lable for the SNAP column only.

ONLINr: USERS is the nurrber of on-line users at the end of the snapshot interval. This statistic is available for the SNAP column only.

% CPU/BATCH USER the percer.t of time used for batch execution and batch service divided by the number of batch users. This statistic is avC"ilable for the SNAP column only.

% CPU/ON LINE USER tile percent of time used for on-line execution and on-line service divided by the number of on-line users. This statistic is available for the SNAP column only.

Figure 15. Description of the SUMMARY Display

STATS 83

BATCH EXEC/SERV is batch execution time divided by batch service time. This is a measure of the degree to which batch fobs are compute-bound.

ONLINE EXEC/SERV is on-line execution time divided by on-line service time. This is a measure of the degree to which on-line jobs are compute-bound.

CPU MSEC PER I/O is the sum of batch and on-line execution and service time (in milliseconds) divided by the number of SIOs. This is a measure of the degree to which batch and on-line jobs are compute-bound.

ONLINE TIME MIX is a measure of the compute-bound character of the on-I ine user load. It is obtained by dividing the compute-bound milliseconds by the interactive task milliseconds. This measure is large when the interactive load is light. A task is considered interactive when less than SQUAN mi IIiseconds {see Table 16} are required for completion of a task.

ONLINE INTENSITY is an estimate of the on-line user load. It is computed by dividing the on-line CPU milliseconds by the number of think-type minutes. On-line CPU milliseconds include execution and service time for on-line users. This statistic is approximately equal to the average task time {in milliseconds} mul-tiplied by the task rate {measured in tasks per think-type minutes}.

ONLINE TASKS/MIN is a measure of the rate at which tasks are arriving from logged on-line users. It is computed by dividing the number of tasks received from on-line users by the total number of logged min-utes for on-line users.

% INTERACTIVE is another measure of the character of the user load. It is computed by dividing the number of interactive tasks that completed wi thin SQUAN mi lIiseconds of CPU time by the total number of tasks that completed in the sample time interval and converting this value to a percent by multiplying by 100.

90% RESPONSE TIME defines the point on the response time histogram (the point representing a number of milliseconds) at which 90% of all tasks receive CPU service within this number of milliseconds.

AVERAGE TURNAROUND is the average number of seconds between the receipt of an activation character from a user and the first I ine of output for that user.

EXECUTION MULTIPLICATION FACTOR is a dynamically changing estimate of the execution time multi-plication factor {ETMF}. The user may multiply the amount of CPU time required for a task by the ETMF to estimate the elapsed time required to complete that task in that user environment.

tALL I ists statistics for the period of time since system restart.

ttSNAP lists statistics for the sample interval only.

J

"

-Figure 15. Description of the SUMMARY Display (cont.)

84 STATS

The CPU display group II printed In the following format:

CPU '7.

BATCH EXEC BATCH SERV ONLINE EXEC ONLINE SERV MONITOR SERV GHOST EXEC GHOST SERV IDLE SWAP WAIT I/O WAIT I/O&SWP WAIT TOTAL

where

BATCH EXEC II the percent of CPU time Ipent for batch programs executrng In the mapped slave mode.

BA TCH SERV II the percent of CPU time lpent for the monitor for services required by batch programs. User service for batch programs Is always executed In the mapped master mode.

ON-LINE EXEC II the percent of CPU time spent for on-line usen executing Tn the mapped slave mode.

ON-LINE SERV Is the percent of CPU time spent In the monitor for services required by on-Itne users.

User service for on-line users Is always executed In the mapped master mode.

MONITOR SERV Is the percent of CPU time lpent for the monitor for scheduling, swapping, symbionts, and other monitor services executing In the unmapped master mode.

GHOST EXEC Is the percent of CPU time lpent for ghost users executing In the mapped slave mode.

GHOST SERV Is the percent of CPU time lpent In the monitor for services requIred by gholt usen. User service for ghost users Is always executed In the mapped master mode.

IDLEttt Is the percent of CPU time spent for unmapped master mode with no execution and no swapl In progress.

SWAP WAlTttt II the percent of CPU time spent In unmapped malter mode wIth a swap In progress but with no execution In progress.

I/O WAlTttt Is the percent of CPU time spent In unmapped master mode with

I/o

In progress but with no execution and no swapping In progrels.

I/O & SWP WAITttt Is the percent of CPU time lpent In unmapped master mode with a swap and

I/o

In progress but with no execution In progress.

TOTAL Is the sum of the percentages listed above. Deferred accountrng may cCluse a value of less than 100%

to be followed by a value greater than 100% on the next Interval. For monoprocessing systems, the total will approximate 100%. For dual processing systems, the total will be as high as 200% with the missing portion representing secondary CPU Idle trme. See the SCPU display for the Idle time.

tALL lists statistics for the period of time since system restart.

ttSNAP lists statistics for the sample Interval only.

ttt In a dual processing system, these values are for the primary CPU only.

Figure 16. Description of the CPU Display

STATS 85

The BATCH display group prints In the following format:

BATCH %

:POO

:Pll EDIT·

LINK·

DELTA·

PCl·

BASIC·

METASYM·

lOADER·

FORT·

USER SHARED

where the percent of the CPU time for batch users Is listed for each of the shared processors. All programs which are not shared are grouped and listed as USER PROG. All other shared processors are grouped and listed as OTHER SHARED.

t All lists statistics for the period of time since system restart.

tt SNAP lists statisti cs for the sample interval only.

·These are the first eight processors that appear on the :SPROCS command of PASS2.

Figure 17. Description of the 8A TCH Display

The ON-LINE display group prInts In the followIng format:

ONLINE % AlLt :p

:POO

:Pll LINK·

DELTA·

EDIT·

PCl*

BASIC·

METASYM*

lOADER· • FORT·

USER SHARED

where each of the shared processors (as defined by SYSGEN) Is named and the percent of the CPU time used by each processor is listed. The percent is based upon the total available CPU time in the interval. All programs that are not shored ore listed as USER PROG. All other shared processors are grouped and listed as OTHER SHARED.

t ALL lists statistics for the period of time since system restart.

tt SNAP lists statistics for the sample intervol only.

*These ore the first eight processors that appear on the :SPROCS command of PASS2.

L...-_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ . ---.---~,,-.. . .

-Figure 18. Description of the ON-LINE Display 86 STATS

90 16 74H-2(9/78)

The USERS drsplay group prints in the following format:

USERS ,

ANSFORT AP APl BASIC EDIT :POO FORTRAN :Pll METASYM PCl TEXT IN CORE

where each of the listed processors fs associated with the number of users printed at the right. The number of users is evaluated only at the end of each snapshot Interval and does not represent an average count for the Interval. The number of users In core (Item IN CORE) Is a count at the end of the Interval and Includes STATS as one of the users

where each of the listed processors fs associated with the number of users printed at the right. The number of users is evaluated only at the end of each snapshot Interval and does not represent an average count for the Interval. The number of users In core (Item IN CORE) Is a count at the end of the Interval and Includes STATS as one of the users

Dans le document Xerox Control Program-Five (CP-V) (Page 95-114)