Resolution:
When serializing and sending strings on the wire, SDK converts all outbound strings to a single encoding (to Latin-1 or UTF-8), as specified in the Repository.

The slowdown is due to the fact that SDK has to convert all non-wide-char encoded strings to UTF-8 when attached to a UTF-8 repository.

The recommendation is to set your repository to ISO8859-1 encoding, unless you intend to store international characters outside of Latin-1 characters in Repository.

Note that SDK stores all non-Latin-1 and non-ASCII strings as UTF-16 in-memory. This means that if the source string's encoding is other than ASCII or Latin-1, the string will be converted to UTF-16 (platform endianness) for in-memory representation. This can also add latency.

NOTE that the time added by character conversion should be minimal but may be noticeable in volume. If you experience a very significant slowdown you will need to further investigate for other possible causes.

When my repository encoding is set to UTF-8, the serialization of data seems to take longer time than setting it to ISO88859-1. Why is that?