Introducing a "prepare-binary-format" initializing phase before actual indexing, where the binary files for each batch is generated. During indexing phase, the binary files are uploaded instead of uploading the docs one by one. Since batching during indexing is deterministic, the batches initialized line up exactly with the batches indexed.
For https://issues.apache.org/jira/browse/SOLR-16812, we needed a way to test performance of indexing formats like CBOR versus JSON and Javabin.
Introducing a "prepare-binary-format" initializing phase before actual indexing, where the binary files for each batch is generated. During indexing phase, the binary files are uploaded instead of uploading the docs one by one. Since batching during indexing is deterministic, the batches initialized line up exactly with the batches indexed.