> Kind of stuck on this one. Had the CE come out and enable the Crypto
> co-processor CPACF feature code for our z9-104 yesterday, then went to
> define and use the feature in a Linux LPAR, but it doesn't work. We
have
> the libica code installed, but whether it's used or not we get the
same
> throughput from the openssl speed tests. I didn't think it took a POR
to
> get the feature recognized - is there something I'm missing here?

Enabling the crypto engine really might not help you much. How much help
you'll get depends a lot on what you're trying to do with it.

There are two components to a SSL transaction: the initial asymmetric
crypto-ignition process at connection startup, and the ongoing symmetric
process after the connection is established. Pre z9 BC/EC, depending on
how you configured the crypto engine (as coprocessor or accelerator),
you get enhancement of one or the other function. The BC and EC models
can be configured in such a way to help somewhat with both tasks.

If a majority of your transactions are short=lived connections, the SSL
offload for the asymmetric step will help a lot. If you're doing
long-lived sessions (like tn3270 wrapping), then you won't get a lot out
of it, except after a network interruption when all the clients try to
renegotiate keys at once. If you're expecting it to help with SSH
sessions, it doesn't. Most of that is symmetric, or uses algorithms that
CPACF doesn't yet know how to accelerate. 

(AFAIS, the openssl speed tests don't really do enough connection volume
to show much of a difference even when the crypto engine is known to be
working. )

----------------------------------------------------------------------
For LINUX-390 subscribe / signoff / archive access instructions,
send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or visit
http://www.marist.edu/htbin/wlvindex?LINUX-390

Reply via email to