Top-n filtering and data sharing groups

In rare situations in which the total range of the filtered column is narrow and the workload is distributed across the data sharing group members in a specific way, the top-n filtering might be slightly inaccurate.

You should consider the following when using top-n filtering in a data sharing environment:

• You cannot sort in ascending order for a data sharing group.
• Due to the nature of summarization in particular, all summary records are entered into CAE agent memory at the same time. However, top-n filtering prevents the situation where all the records must temporarily exist in raw memory and as Java™ data structures at the same time. This at least halves the memory requirements for large queries. In addition, it dramatically reduces the CAE agent Java heap requirements, as only the top-n records need to be brought into the Java heap. While a large request can cause a temporary growth in memory, almost all of that memory is guaranteed to be released to the operating system once the call is done, since it does not involve expanding the Java heap.

Since top-n filtering occurs in the CAE agent, it greatly reduces memory usage on the CAE server for large data requests. Top-n filtering operates within the memory limitations of the CAE server. This might cause incomplete results in rare circumstances.