Urjit Gokhale wrote:
> Hello everyone,
>
> I would like to know your view on "Would the performance impact for
> transmitting 'X' bytes of unencrypted data over SSLized channel be
> almost same irrespective of the number of SSL_* calls" ?
>
> Consider I am using a buffer of size 'buf_size' for SSL_read/write
> (SSL_read(ssl_obj,buf,buf_size) ):
> If buf_size < 'X' I would require more number of SSL_reads/writes.
> If buf_size > 'X' a single SSL_read/write should suffice.
> So would the performance hit for end to end transmission of this 'X'
> bytes of data be the same in the cases mentioned above?
> I am assuming ideal network (no network latency etc) and 'X' is not big
> enough to fill kernel buffers on either side.
>
> ~ Urjit
>


My view; it has a similar performance impact as if you'd sent X bytes
over TCP with TCP_NODELAY set each time only sending Y bytes, where X is
pretty big and Y is a small number (Y corresponds to your buf_size above).

I don't remember SSL having an option to make it wait for a certain
buffer size before sending it down to the network (some trick in the
BIO? maybe someone can correct me).

-jb
--
Real computer scientists don't comment their code. The identifiers are
so long they can't afford the disk space.
__________________________________________________ ____________________
OpenSSL Project http://www.openssl.org
User Support Mailing List openssl-users@openssl.org
Automated List Manager majordomo@openssl.org