Override work factor estimates
Overview
Use this task to override the values that are computed by the work profiler.
If we have already specified a speed factor override, we do not need to configure a work factor override. For more information about speed factors, read about configuring work factors in multiple tier configurations.
The work profiler computes a work factor for each transaction class module (TCM). Specifying the override depends on the version of the product that we are using. Use work classes to classify work into transaction classes.
When using work classes, an XML element does not exist for each TCM. In this case, the work factor overrides related to a given deployment target (cluster, or unclustered server) can be specified in a custom property that is attached to that deployment target.
Use the workFactorOverrideSpec custom property. The value for this custom property is a string containing a set of matching rules, or cases, that define the work factor override, either a value or none, for every TCM of that deployment target. The syntax of that string is the same as the syntax of the speed factor overrides, except that we cannot specify a tier for a work factor override case. For more information about speed factors, read about configuring work factors in multiple tier configurations. A work factor is in units of millions of CPU clock cycles on a standard kind of machine, and describes the average amount of work that must be done on the managed tier for a single request of that TCM.
Override work factor estimates
- Reduce the number of back end nodes in the node group to one back end node in the node group panel in the administrative console.
- Use a workload generator, send traffic to one TCM only. Generate enough traffic to load the back end node to more than 70% utilization.
- Record the average values for the steady state throughput and corresponding CPU utilization. We can get those values from the runtime topology view of the admin console. To view the runtime topology, click...
Runtime operations | Runtime topology
- Compute the work factor value for the transaction class module. Use the value from the following equation:
work_factor = (CPU_Utilization * node_speed) / Throughput
In this case, Node_speed equals the value of the node.speed node custom property. For more information about setting the node.speed custom property, read about configuring node computing power. It is assumed that all the nodes in the node group are homogeneous, with equal node speeds.
- Compose a string to override the work factor for the deployment target. The pattern defines the set of service classes, transaction classes, applications, or modules that we can override. The pattern is :
service-class:transaction-class:application:module = value
We can separate each work factor with a comma, or specify a wild card for any of the variables using an asterisk. Set the value variable to the computed work factor value. For more information about the grammar for work factor override specifications, read about autonomic request flow manager advanced custom properties. In the deployment target, go to...
Custom properties | New
...and enter the Name of the custom property as workFactorOverrideSpec. The Value of the custom property is equal to the string composed in the previous step to override the work factor.
What to do next
Configure the speed of the external nodes. For more information about setting the node.speed custom property, read about configuring node computing power.
Configure the autonomic request flow manager Configure work factors in multiple tier configurations Routing requests to nodes that are not running Intelligent Management Configure node computing power Intelligent Management: autonomic request flow manager advanced custom properties Intelligent Management: overview