The world of storage is tiered, and it will become more distinctly tiered in the years ahead as the ability to manage hot data will evolve onto the PCIe bus and away from the sub-optimal SAS and SATA buses designed for traditional disk platter based storage. Using the right bus for your hot tier is very important. Up till 2014 most implementations of SSD storage have been on SAS and SATA buses which are not designed for fast non-volatile memory (NVM). What’s been required is more standardization around the actual processor’s host bus, called PCIe. A generational shift towards PCIe from Intel is now evolving.
Intel is evolving the world of PCIe and its extensibility across the necessary layers, so that PCIe can truly be a more appropriate storage bus for going wide with more devices. Thus blending network, storage, co-processors and graphics devices all via this host bus. The classic storage need for adaption from SAS or SATA back up towards PCIe and the processor will slowly fade as the server generations evolve in the years ahead. We’ll see Intel-backed standards, platform additions and PCIe bridges and switches, which will start the unstoppable evolutions of putting storage closer with much less storage latency to the processor.
Tomorrow’s super and parallel computing can only be made a reality with denser, more efficient compute power. NVM storage will play its part by being on the CPU’s host bus. This storage will be more power efficient and parallel. The future has started with Intel SSD for PCIe, come check us out at the Intel booth #1315 at SC 2014 and talk to Andrey Kudryavtsev, John Ryan or myself. We’ll be there live to show you demo’s, samples and explain the system shifts and the benefits of the new storage protocol standard, NVMe as well.