In some instances, rather than using WLM recommendations, weighted active connections (WEIGHTEDActive) can provide a more appropriate solution to control workload distribution:
Target systems vary significantly in terms of capacity (small systems and larger systems). WLM recommendations might favor the larger systems significantly. However, a target application might not scale well to larger systems; because of its design, it might not be able to take full advantage of the additional CPU capacity on the larger systems. This can result in these types of servers getting inflated WLM recommendations when running on larger systems and getting overloaded with work.
SHAREPORT is being used, but not all systems have the same number of SHAREPORT server instances (for example, one system has two instances and the other has three). The current round-robin or WLM recommendations do not change distribution based on the number of server instances on each target. Round-robin distribution distributes one connection per target stack regardless of the number of SHAREPORT server instances on that stack. WLM server-specific weights from a target stack with multiple server instances reflect the average weight.
For example, you might need to reserve some capacity on certain systems for batch workloads that are added into selected systems during specific time periods and have specific time window completion requirements. If those systems are also a target for long-running distributed DVIPA connections, WLM recommendations allow that available capacity to be consumed. This can potentially impact the completion times of the batch jobs when they begin to run, if they are not able to displace the existing non-batch workloads. Similarly, the existing connections on that system can experience performance problems if the batch jobs displace those workloads.
Weighted active connections provide granular control over workload distribution based on predetermined active connection count proportions for each target (fixed weights). Distribution of incoming TCP connection requests is balanced across the targets so that the number of active connections on each target is proportionally equivalent to a configured active connection weight for each target (specified on the DESTIP parameter for each target). Control is gained at the expense of losing the dynamic benefits of WLM recommendations; however, server-specific abnormal completion information, the general health indicator, and the TSR value are used to reduce the active connection weight when these indicators are not optimal. If weighted active connections are used, study and determine the comparative workload that you want on each system so that you can configure appropriate connection weights.
To enable the distributing stack to use server-specific abnormal completion and health information to affect the active connection weight, specify SYSPLEXROUTING on the IPCONFIG statement for all participating stacks.
You can select this type of distribution for DVIPA and port targets by specifying the WEIGHTEDActive option on the DISTM parameter of the VIPADISTRIBUTE statement, and specifying the proportion of active connections that you want on each target using the WEIGHT option on the DESTIP dynxcfip parameter of the VIPADISTRIBUTE statement.