Currently in BPM Connect, when a user schedules feeds for the exact same date and time, it batchs them together in a process called a "daisy chain". This is a classic and highly used feature of BPM Connect. However, classically, when the feeds are batched and run together, the daisy chain would process the cube and rebuild the indexes after ever feed. This added redundancy, performance overhead, and was largely inefficient, especially on large models.
Now available in DBObjects 3.3.0 and higher, the daisy chain batch is intelligent, where it will only process the cube and build the indexes a single time after the daisy chain is complete rather than after each feed.
So for example, if there are 10 fact feeds in a single daisy chain process that load to the same cube, the feed will process the cube and rebuild the indexes on the completion of the 10th feed only. This significantly reduces overhead, increases performance and saves time on the overall daisy chain process.
It is important to note, this solution only works when the feeds daisy chained together share a common destination cube. If the feeds that are daisy chained together share mixed destinations, do not use this functionality.
There is is switch that enabled this capability and it spans across the whole instance. The switch is called "DaisyChainProcessOnce" and is located in Config Builder Global Configurations. Set the switch to "1" to enable behavior , "0" to disable behavior. See below example:
No comments:
Post a Comment