Massimo Biancucci wrote: >we've been installing a new application using crypto facility calling services >like 'CSNBKEX' or 'CSNBKIM'. >We've done all the possible (in our understanding) to optimize the >application, so now we can measure more than 150 msg/second per single >application server. >What does sound really strange to me is the elapsed time of the two routines >previsouly mentioned. >Every call costs more than one millisecond. >Cause the CEC is a z12 and the crypto is in HW, I believe those elapsed should >have been microseconds. >Anybody got a similar experience ?
It sounds like you think the Crypto Express is for performance. That's a common misconception: CEX is for *security*. Calling CEX is essentially doing I/O, so it's not cheap, where "not cheap" means "a couple of orders of magnitude slower than CPACF, and slower even than just software for most operations". This is why I like to say, "CEX is not always good" (yeah, say it out loud...). A millisecond for a CEX call doesn't strike me as "long" at all. HTH, -- ...phsiii Phil Smith III [email protected]<mailto:[email protected]> Voltage Security, Inc. www.voltage.com<http://www.voltage.com/> (703) 476-4511 (home office) (703) 568-6662 (cell) ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [email protected] with the message: INFO IBM-MAIN
