New on the Power Development Platform:
The IBM Data Engine for NoSQL - Power Systems Edition creates a new tier of memory by attaching up to 57 Terabytes of auxiliary flash memory to the processor without the latency issues of traditional I/O storage. This system is configured with software in a development EMULATION mode, and does not contain real CAPI Flash accelerators. It is intended as a development platform and is suitable for software development, unit testing, and code integration tasks.
For more information, refer to the following resources:
- http://ibm.biz/capiflash - Technical Whitepaper about the IBM Data Engine for NoSQL, including best practices
- https://github.com/open-power/capiflash/ - Complete source for Data Engine for NoSQL enabling software, including test cases
- /opt/ibm/capikv - Software installation on this system, which includes example code, sample applications, headers, and shared libraries, compiled in a development "FILE MODE."
To get started with key-value layer APIs (libarkdb) pass a NULL "file" string to ark_create(...). The database will automatically allocate key/value pairs in system RAM instead of flash.
To get started with block-layer APIs (libcflsh_block) create a temporary block file on the local file system (this is for development purposes only, e.g. "fallocate -l 2G ~/blockfile"), then pass that file path to cblk_open(...). The block APIs will read / write data to this test file instead of a real flash accelerator.
Once software is ready for testing on real systems, dynamically or statically-link to production CAPI Flash shared libraries which are available on the system with real accelerators, or available on IBM FixCentral (search for the EJ16 feature code).