Customizing IBM Tivoli Workload Scheduler for z/OS V8.2 to Improve Performance

Chapter 3. Optimizing Symphony file creation and distribution 41

The size of the network can help you decide whether to use a single domain or

multiple domains. If you have a small number of systems or a small number of

applications to manage with IBM Tivoli Workload Scheduler, there might not be a

need for multiple domains.

Geographic location and the communication speed between locations are one of

the primary reasons for choosing a multiple domain architecture. One domain for

each geographical location is a common configuration. If you choose a single

domain architecture, you will be more reliant on the network to maintain

continuous processing.

An IBM Tivoli Workload Scheduler network, with either a single domain or

multiple domains, gives you the ability to manage IBM Tivoli Workload Scheduler

from a single node–the master domain manager. If you want to manage multiple

locations separately, you can consider installing a separate IBM Tivoli Workload

Scheduler network at each location. Note that some degree of decentralized

management is possible in a stand-alone IBM Tivoli Workload Scheduler network

by mounting or sharing file systems.

There can be other reasons for choosing a multiple domain configuration, for

example, a domain for each building, department, business function, or

application.

The degree of interdependence between jobs is an important consideration when

designing your IBM Tivoli Workload Scheduler network. If you use multiple

domains, you should try to keep interdependent objects in the same domain.

This will decrease network traffic and take better advantage of the domain

architecture.

What level of fault-tolerance is required? An obvious disadvantage of the single

domain configuration is the reliance on a single domain manager. In a

multi-domain network, the loss of a single domain manager only affects the

agents in its own domain.

3.5.5 FTA tuning

For the test with 250,000 FTA jobs, the localopts values for the FTAs were not

changed. However, in some earlier testing (see 3.3.1, “FTA tuning parameters

(localopts)” on page 31) with a smaller number of jobs (10,000 and 50,000), the

effects of modifying the localopts parameters were tested, and a significant

improvement can be made by setting these parameters as indicated in the

following list.

42 Customizing IBM Tivoli Workload Scheduler for z/OS V8.2 to Improve Performance

Recommended localopts changes for FTAs are:

 wr enable compression=yes (compress Symphony/Sinfonia if it is 4 MB or

larger)

 sync level=low

 mm cache enable=yes

 mm cache size=512 (maximum value, unless you are running the FTA on a

low memory machine; see the following Tip about mm cache size)

 bm look=9 (reduce from default value of 15 and make lower than bm read)

The following section provides more insight into these parameters.

Mailman cache (mm cache mailbox and mm cache size)

Tivoli Workload Scheduler is able to read groups of messages from the mailbox

and put them into a memory cache. Access to disk through cache is much faster

than the accessing to disk directly. The advantage is even more relevant if you

realize that a traditional mailman needs at least two disk accesses for every

mailbox message.

A special mechanism ensures that messages considered essential are not put

into cache but are immediately handled. This avoids loss of vital information in

case of a mailman failure. The settings in the localopts file regulate the behavior

of mailman cache:

 mm cache mailbox: The default is no. Specify yes to enable mailman to use a

reading cache for incoming messages. Note that this option is

not used in

USS. It is

only available for UNIX and Windows systems.

 mm cache size: Specify this option only if you use the mm cache mailbox

option. The default is 32 bytes and should be a reasonable value for small and

medium-sized Tivoli Workload Scheduler installations. For larger installations,

you can use the maximum value, which is 512. Higher values are ignored.

Tip: You can use larger values than 32 bytes for machines with high memory.

But if you are using the FTA on a low memory machine, be careful for not

setting this value unnecessarily large, since this would reduce the available

memory that could be allocated to other applications or other Tivoli Workload

Scheduler processes.