Operating Systems

Test Your z/OS Knowledge: WLM

6 Pages

8 ____

10 ____

14 ____

15 ____

20 ____

30 ____

31 ____

32 ____

64 ____

72 ____

95 ____

99 ____

100 ____

191 ____

208 ____

250 ____

253 ____

254 ____

999 ____

a. Seconds between Resource Adjustment intervals

b. Smallest response time goal allowed (in milliseconds)

c. Number of time slices in one SRM second

d. Minimum response time goal seconds for individual storage isolation

e. Minimum velocity goal ignored for discretionary goal management

f. Possible dispatching priority for small consumers of CPU

g. Smallest possible resource group maximum possible

h. Number of response time distribution buckets

i. Lowest CPU dispatching priority possible

j. WLM state sampling interval in milliseconds

k. Maximum number of service classes allowed

l. Default dispatching priority for JES initiators

m. Seconds between Policy Adjustment intervals

n. Seconds a period may be “ignored” if WLM can’t help it

o. Maximum number of resource groups allowed.

p. CPU busy percentage WLM considered “constrained” during WLM initiator management

q. Lowest policy adjustment CPU dispatching priority possible for Importance level 1 work

r. The medium importance level

s. WLM ISPF Application option to edit service class definitions

t. Version of MVS that first introduced the Workload Manager

u. Considered to be the importance level for discretionary periods

v. SMF record number containing WLM workload activity data

w. Maximum length of a service class name

x. Maximum number of report classes allowed

y. Highest velocity goal possible

z. Smallest resource group minimum allowed

9. Match the formulas at the top (lines a-d) with the descriptions on the bottom (lines 1-4):

____ a. Goal/Actual ____ b. Actual/Goal

____ c. (Using/(Using + Delay)) * 100 ____ d. Queue + Execution 

1. Actual Velocity achieved formula

2. Average Response Time Goal PI formula

3. Response Time formula

4. Velocity Goal PI formula

10. Although the answer “it depends” could be applicable to all of these questions, indicate the expected effect on achieved velocities when the following indicated change is made. Your choices for answers are (a) Velocity likely to increase, (b) Velocity likely to decrease, (c) Velocity likely not affected.

a. Increase the number of logical and physical processors assigned to the partition

b. Velocity of Started task periods when WLM starts batch initiators but before a job is selected

c. In the velocity goal period, increase the number of active CICS regions being managed toward region goal.

11. What are the standard CPU dispatching priority ranges for the following types of work?

Answers should be expressed in base 10:

_____ to _____ - SYSTEM service class

_____ to _____ - SYSSTC service class

_____ to _____ - Work assigned Importance level 2

_____ to _____ - Work assigned velocity goals

_____ to _____ - Work assigned discretionary goals

12. For each subsystem type listed, indicate the type of work unit classified by the classification rules of each subsystem type. Your choices for answers include: (a) Address spaceoriented, (b) Enclave-oriented, (c) Neither

a. TSO ______ b. JES ______ c. CB ______ d. STC ______ e. CICS ______ f. DDF ______ g. IMS ______ h. DB2 ______

13. Which of the following does not pertain to when WLM stops WLM-managed INITs?

a. CPU-constrained

b. Real storage-constrained

c. More that one hour of inactivity

d. Many more INITs exist than are needed

e. Operator issued a Hold Queue Command ($HQ)

Answers

1. c. PI = 0.81

A PI of zero means there’s no work running in the period or there are no ended transactions. A PI of 0.50 just means the work is doing twice as good as its goal. A PI of 1.00 means the work is meeting its goal exactly.

2. b. During enforcement of a resource minimum

The standard CPU dispatching priority range for discretionary work is CPU dispatching priorities of 192 to 201. If a period assigned a discretionary goal is running below its resource group minimum, WLM may decide to move the discretionary work equal to or above some importance 1 through 5 work if doing so won’t jeopardize the goals of the importance 1 through 5 work.

3. b. Acceptable delay for work   

In the context of WLM, velocity isn’t a speed objective. Velocity also isn’t an object of regular progress. Work in a service class period with a few using samples, a smaller number of delay samples, and lots of unknown samples could have a high velocity but not make any progress, since it’s mired in some unknown delay. It also isn’t a user satisfaction objective, since it’s only a measure of a portion of the transaction on z/OS. Work could be achieving a high velocity but the users may be very unhappy, since they’re delayed someplace outside z/OS. Velocity is the ratio of using samples to using and delay samples. A high velocity goal indicates few delays (known by WLM) are to be tolerated.

4. b. False

A resource group minimum is enforced when work is running below its minimum and its goal is being missed. The work could have CPU delays but still be meeting its goals.

5. b. False

The answer would be true if we were referring to the same period with the same workload on the same z/OS image. However, the question doesn’t qualify that we are comparing two different goals for the same service class period. Assuming two different service class periods and two different workloads, the answer is false. For example, on a five-way LPAR, a service class period with 20 CPU-intensive jobs and a velocity goal of 20 is probably more difficult to achieve than a service class period assigned a velocity 50 and running three jobs requiring less CPU.

6. b. Disconnect WLM can’t do anything to directly influence disconnect time, so it isn’t counted as either a using or a delay state. Connect time is a productive time and considered a using state. Pend is considered a delay state that can be helped by WLM with an action such as dynamic channel path management. IOSQ is considered delay state that can be helped by WLM with actions such as changing the I/O dispatching priority or parallel access volumes.

7. c. The goal for the address space is being ignored

SERVER=YES means the assigned goal for the address space is being ignored and WLM is managing the address space to help meet the goals of the transactions this address space is serving.

A CICS address space managed toward its region goal will have its velocity goal honored. Only when it’s managed toward the transaction goals is the region goal ignored and SERVER will be YES. Not all address spaces that run enclaves have their goals ignored. Only enclave servers do. An example of an address space running enclaves that has its goal still honored is the xxxxDIST DB2 address space. Application environment queue manager address usually has its goal honored. The application environment queue server’s goals are ignored, but if the queue server hasn’t been processing transactions for some period of time, WLM will start managing that address space toward its own goal again.

8. a) 2 b) 15 c) 64 d) 20 e) 31 f) 253 g) 1 h) 14 i) 191 j) 250 k) 100 l) 254 m) 10 n) 30 o) 32 p) 95 q) 208 r) 3 s) 4 t) 5 u) 6 v) 72 w) 8 x) 999 y) 99 z) 0

9. a) 4 b) 2 c) 1 d) 3

10. a) Velocity will likely increase, since there’s more opportunity to use the CPU and less likeliness for CPU delays. b) Velocity likely won’t be affected, since enabling WLM batch initiator support shouldn’t have direct effect on started tasks. c) Velocity will likely decrease, since the additional CICS regions in the period will most likely increase demand for CPU and increase the probability for CPU delays.

11.

255 to 255 - SYSTEM service class

254 to 254 - SYSSTC service class

208 to 253 - Work assigned Importance level 2

208 to 253 - Work assigned velocity goals

192 to 201 - Work assigned discretionary goals

12. a) Address space-oriented, b) Address space-oriented, c) Enclave-oriented, d) Address space-oriented e) Neither, f) Enclave-oriented, g) Neither, h) Enclave-oriented.

13. e. Operator issued a Hold Queue Command ($HQ)

The stopping of WLM-managed INITs isn’t influenced if an operator decides to issue a hold queue command. Resource constraints and unneeded INITs may cause WLM to stop WLM-managed INITs. Z

6 Pages