The latest version of the non-volatile memory express protocol is NVMe 1.2, which allows a solid-state drive controller...
to use system RAM on an as-needed basis.
NVMe is a publically available interface that provides a command and feature set for solid-state drives (SSDs) that connect through PCI Express. The standard is meant to improve SSD performance and to provide users with the flexibility needed to interconnect systems from different vendors.
Many SSDs have onboard dynamic RAM (DRAM), which is commonly used by the disk translation tables in the Flash Translation Layer. The FTL converts logical mapping to physical mapping on flash storage. When blocks are written to flash, the FTL determines where those blocks will physically reside. The FTL is associated with wear leveling and power cycle recovery. Storage manufacturers have long wanted to eliminate SSD-level DRAM -- the impetus behind NVMe 1.2 -- because it would reduce the cost of the disk, free up physical space and increase storage capacity.
While it is possible for an SSD to function without DRAM, manufacturers found that removing translation tables from DRAM and storing them in NAND negatively affected disk performance. DRAM is faster than flash memory, so storing the translation tables in DRAM results in faster physical allocation of blocks than when the tables are stored on flash. However, RAM-based tables must be protected by battery backup to guard against power loss. The NVMe 1.2 protocol provides manufacturers with a way of moving translation tables off the drive and into system memory.
Using system RAM for NVMe -- as outlined in NVMe 1.2 -- has only recently started to gain traction. Moving the tables off flash and into memory isn't a user- or administrator-level process; it is something the manufacturer does to decrease the cost and physical footprint of the drive without incurring the performance hit normally associated with DRAM removal.
Why NVMe and SCSI Express are important for SSDs
The importance of decreasing SDD sizes
Compare DRAM SSD and NAND SSD
Dig Deeper on Flash storage for applications
Related Q&A from Brien Posey
The reasons for going hyper-converged vary. Often, however, organizations deploy HCI technology to effectively address one or more of the five issues... Continue Reading
Adhering to service-level agreements, keeping up with performance demands and planning for future workloads are just a few of the goals you should ... Continue Reading
Is hyper-convergence a good fit for disaster recovery? More organizations appear to believe so and are increasingly using hyper-converged ... Continue Reading
Have a question for an expert?
Please add a title for your question
Get answers from a TechTarget expert on whatever's puzzling you.