Hi Gabe,

The internal header is one per chunk, so I think
(4096 * n) - 16 is correct.

Tom

Gabriel Leung wrote:

> I was going through the JMS Performance Guide, Section 4.2.3 about Tuning Chunk
> Size. The section provides a guideline for estimating the optimal chunk size:
> 4096 * n -16. The 16 bytes are for internal headers. Base on the text, the chunk
> size should be 4080 (also the default). My question is that wouldn't the suggested
> calculation 4096 * n - 16 double counting the 16 bytes? Shouldn't 4080 * n + 16
> be the correct calculation for the final size and the chunk size should simply
> be 4080 * n?
>
> Thanks for your help.
>
> Gabe
>