Performance issue while grouping in HCM Extracts

Summary:

We have a requirement to build a compensation interface to send 3 months of payroll earnings data from run results to third party vendor.

The xml size getting generated from the HCM Extract is about 3-4 GB.

Interface output needs to be written by grouping xml data for a combination of 5 columns(all text). We tried multiple approaches as mentioned below but unable to generate the final output since BIP job is timing out.

Content (required):

Below are the approaches we tried

1. Grouping the records in E-text template - HCM Extract sub process are completing in 50 mins but the BIP job is running for days without completing.