Bulk-loader performance

[Tulasi Paradarami <tp...@23andme.com>]

Hi,

Here are the details of our environment:
Phoenix 4.3
HBase 0.98.6

I'm loading data to a Phoenix table using the csv bulk-loader and it is processing about 16,000 - 20,000 rows/sec. I noticed that the bulk-loader spends upto 40% of the execution time in the following steps. 
//...
csvRecord = csvLineParser.parse(value.toString());
csvUpsertExecutor.execute(ImmutableList.of(csvRecord));
Iterator<Pair<byte[], List<KeyValue>>> uncommittedDataIterator = PhoenixRuntime.getUncommittedDataIterator(conn, true);
//...

To load TBs of data, the overall performance of the bulk-loader is not satisfactory. Could someone comment on the following:
- Is there a way to perform bulk-loading without creating a PhoenixConnection and performing an upsert + conn.rollback?
- Some additional details around why bulk-loading is designed this way? A reference to JIRA with details will help too.
- If I want to bypass csv parsing and uncommittedDataIterator, are there any Phoenix APIs that can be used for creating the output key-values.

Thanks