Performance Measurements
The Coprocessor operates as a server in a client-server relationship with your host-system applications. The application within the Coprocessor, such as the IBM CCA Support Program or PKCS #11 Support Program, or an application that you create or purchase from another vendor, employs the Coprocessor CP/Q++ multi-tasking control program. The two multi-tasking systems make understanding the response time for a single request in typical operating environments, or the throughput of a mixed class of work, a difficult analysis.
The information in these pages can only serve as a very rough guide to the capabilities available to your applications. The throughput of the IBM 4758 PCI Cryptographic Coprocessor subsystem depends on many variables, including your system environment, application usage, PCI bus contention, and other factors. As with all complex subsystems, performance in your environment will vary from the information presented here. You should plan on performing your own analysis to ensure that this product is suitable for your requirements.
See these sections for summary performance information:
- Processing internal to the Coprocessor: timings for calls made by an application performing within the Coprocessor at the API offered by CP/Q++
- CCA cryptography: single- and multi-threaded performance information for calls made to the CCA API in host systems
- PKCS #11 cryptography: PKCS #11 information.
- A CCA Performance Demonstration Program is available as sample code that you can use for benchmarking.
Often raw performance is secondary to other Coprocessor advantages...
- Protect keys and other secret values within the Coprocessor subsystem so that a compromise of the host system need not result in the loss of critical information
- Run critical processes in an environment whose authenticity can be remotely demonstrated
- Offload DES, RSA, and SHA-1 processing to the less expensive Coprocessor electronics.