...
When you create an Ultra Task, you have the option to load balance processing across nodes. The number of instances controls how many pipelines are started to process the Ultra requests. Since the pipelines are automatically started and are always running, the number of pipelines needs to be balanced. Having too many instances uses up slots, resulting in extra load on the node. Having too few could result in delays for requests if the request volume is higher than running instances.
Manual
This As a default option, this selection sets the number of pipeline instances to start, regardless of the number of JCC nodes. Use this setting in Ultra Tasks where the pipeline limits the number of concurrent calls from downstream systems.
...
For example, setting this to 2 runs two pipelines on each JCC node. If the Snaplex has 4 four JCC nodes, then 8 eight pipeline instances start.
...
This approach doesn’t guarantee
...
two instances per node, but
...
eight nodes in total. The balancing mechanism is designed to distribute them evenly.
In this configuration, you can specify manually how many instances to run per Snaplex node. This setting suits both types of Ultra Pipeline design, providing redundancy through scaling.
For example, if you run 2 two instances in a Snaplex with 4 four JCC nodes, then 8 eight instances of the pipeline are run per node. Likewise, specifying 2 two instances at 6 six JCC nodes in the Snaplex creates 12 twelve instances of the pipeline.
...
Scale the number of pipeline instances based on the number of requests being processed. This option can be used when the request rate is expected to be variable, with low and peak volume volumes being significantly different.
| Info |
|---|
Always-on Ultra Pipelines do not support AutoScaling because they do not require a FeedMaster. |
Prerequisites
A Snaplex with one or more Feedmaster nodes.
Write permissions for the Project Space.
A valid Pipeline.
...