Load profiles allow you to define the operating criteria for a logical cluster. These criteria are typically called “load score metrics,” with the value associated for each criteria rolled into a “score” for each instance in the logical cluster that uses the load profile. You can periodically compare load scores for different instances within a logical cluster to detect when the workload is undesirably skewed to one or more instances, or determine if an instance is under-utilized.
Instances included in multiple logical clusters can be impacted by multiple load profiles, so take care when associating instances with multiple logical clusters and when defining, and applying load profiles.
The Cluster Edition includes two system load profiles: sybase_profile_oltp for
OLTP environments and sybase_profile_dss for
DSS environments. You cannot modify or delete system load profiles.
However, you can duplicate them and modify the duplicates to create
your own load profiles.
The load profile status reports the:
Name – the name for the load profile configuration.
Type – the load profile type: system or user.
Metric weights – relative weight assigned to each metric in the load profile. Metrics include:
User connections – displays the weight of users connected to the particular load profile.
CPU busy – displays the weight of CPUs that are currently busy.
Run queue length – displays the weight of the run queue.
IO load – displays the weight of the I/O load.
Engine deficit – displays the weight of the engine deficit.
Users – displays the weight for a metric the user chooses to measure.
Thresholds – configured difference (as a percentage) in the load between two instances in a logical cluster at which point the following can occur:
Login redirection – used for connection-time load balancing and routing connections to a logical cluster. If necessary, an instance directs a client to stop the current login attempt and try connecting to instances it supplies as a list of available network addresses.
Dynamic migration – (also known as the hysteresis value) displays the dynamic migration configuration.
Minimum load score – load score necessary to trigger login redirection and dynamic migration.
Select Load Profiles from the Workload Management folder and select Add Load Profile.
Enter the name for your profile.
Select Next.
Adjust the load profile metric weights.
When a load profile is associated with a logical cluster, the workload manager calculates a load score for each instance in the logical cluster. This is calculated using the weight you entered for each metric, the raw value of each metric for the instance, and the workload measurement algorithm. See “Viewing workload status”.
The metrics measured by the server are:
User connections – the capacity of an instance to accept a new connection, based on resource availability.
CPU utilization – the capacity of an instance to accept additional work.
Run-queue length – the number of runnable tasks on a system. Run-queue length measures the processing backlog, and is a good indicator of relative response time.
I/O load – outstanding asynchronous I/Os.
Engine deficit – the difference in the number of online engines among instances in the cluster.
Engine deficit is measurable only when instances in
the cluster have unequal numbers of engines. Engine deficit adds
a metric for maximum relative capacity to the load score.
User metric – an optional customer-supplied metric specific to the user’s environment.
Make sure the load scores you specify add up to 100. If they do not sum to 100, the workload manager uses the scores to create proportionate values that add to 100.
Select Next.
Enter values for:
Minimum Load Score – the load score is not a percentage, but is the minimum score the workload manager requires before it redirects work to other instances. The minimum load score is meaningful when compared to the load scores of other instances in the logical cluster using the load profile.
Login Redirection (%) – the load threshold for determining how best to distributed incoming connections.
Dynamic Connection Migration (%) – the load threshold that determines whether to distribute existing connections.
The load threshold is the percentage difference between the load on the current instance and the load on the least-loaded instance currently participating in a logical cluster. That value must be met before the Cluster Edition redirects a login or migrates an existing connection.
The percentages for Login Redirection and Dynamic Connection Migration
are independent percentages, and do not need to add up 100.
Select Finish to create the load profile.
