Last week, the Kaminario team presented at Tech Field Day 10 in San Jose. This was a unique chance to talk to industry experts about our vision for the all-flash space and provide a preview of how our technology will extend into the future.
We believe we offer the most flexible, scalable, cost-effective all-flash array platform on the market today. Our fifth-generation, software-defined architecture provides unique advantages for customers and will enable us to take advantage of exciting technology trends that are rapidly coming into focus.
We focused on seven technology trends that will transform flash storage and their implications.
- Slowdown of Moore’s Law. Moore’s Law is no longer delivering 2x yearly performance increases. Standard hardware acceleration will compensate for some, but not all, of the demand for cost-effective performance increases.
Implication for AFAs: Scale-out approaches will become even more attractive.
- Networking advancement. Infiniband EDR, 100Gb Ethernet and Omni-Path are coming and there is a clear line of sight to 400Gb network technologies.
Implication for AFAs: Further strengthens the prospects for scale-out and lowers the motivation to keep storage right next to compute.
- 3D NAND (and possibly QLC) remain dominant. NAND density (cost) will continue to improve in the next five years while endurance declines.
Implication for AFAs: NAND will remain the leading storage media and will be the commodity curve that replaces HDDs in the datacenter.
- The rise of non-volatile memory (3DXPoint, ReRAM and others). NVM price is much closer to DRAM than Flash. 0. to 2 microsec performance is much faster than flash but still far from DRAM
Implication for AFAs: NVM is too expensive to replace NAND and should be used wisely. It’s not going to replace NAND for five-plus years.
- NVMe. NVMe is much more efficient than SCSI and will soon replace SATA as direct attach.
Implication for AFAs: Lower benefit for AFA since these systems already aggregate the performance of several SSDs. The controller is the bottleneck so decreasing latency from disk to controller is not that interesting.
- NVMe fabric. NVMe fabric extends the efficiency of local NVMe interface over the fabric.
Implication for AFAs: History teaches us to exected slow market adoption for external storage connectivity.
- NVMe network shelves: Shelves with NVMe SSDs and RDMA network connectivity will be available in the next several months.
Implication for AFAs: NVMe network shelves create an interesting opportunity to further improve decoupling of capacity from compute.
In my talk, I described how, despite lots of buzz in the industry, these technologies are not yet enterprise-ready. And even when they do reach that level, for the foreseeable future they will be so expensive that they will only fit into niche, high-performance, “tier 0” applications.
So, how can we use these new technologies?
Kaminario’s agile software architecture opens up some possibilities to use the technologies in unique ways to further our goal of delivering the most flexible, scalable, cost-effective all-flash platform on the market. I talked through a few such forward-looking scenarios.
First, there is definitely the opportunity to adopt denser flash faster. Kaminario has endurance optimization IP that would help us adopt new, higher density flash sooner.
We also discussed the idea of replacing DRAM with NVM technologies. While it is clear that NVM is too expensive to replace NAND, NVM is more cost efficient than DRAM. This creates a real opportunity to replace DRAM for data and MetaData caching leveraging Kaminario’s highly flexible MetaData management Paradigm.
Finally, we discussed the potential to completely decouple storage controllers from capacity leveraging advances in NVMe, NVMe fabric, and NVMe network shelves. This is a direct extension of Kaminario’s scale-up and scale-out architecture available today. This vision would be the ultimate in cost-efficient, agile storage infrastructure. IT organizations would be able to optimize for cost, performance and capacity, maintaining a highly predictable infrastructure as the data center scales indefinitely.
I really enjoyed the opportunity to present our vision and get the feedback from the Storage Field Day panel. Make sure you check out full set of videos here.