Abstract
3D NAND Flash is the preferred storage medium for dense mass storage applications, including Solid State Drives and multimedia cards. Improving the latency of these systems is a mandatory task to narrow the gap between computing elements, such as CPUs and GPUs, and the storage environment. To this extent, relatively time-consuming operations in the storage media, such as data programming and data erasing, need to be prioritized and be potentially suspendable by shorter operations, like data reading, in order to improve the overall system quality of service. However, such benefits are strongly dependent on the storage characteristics and on the timing of the single operations. In this work, we investigate, through an extensive characterization, the impacts of suspending the data programming operation in a 3D NAND Flash device. System-level simulations proved that such operations must be carefully characterized before exercising them on Solid State Drives to eventually understand the performance benefits introduced and to disclose all the potential shortcomings.
Highlights
The data storage in enterprise scenarios, including High Performance Computing (HPC) and cloud-based services, requires the low latency and the high throughput that only Solid State Drives (SSD) architectures can deliver [1]. 3D NAND Flash-based SSDs are the preferred solution due to the offered large storage density, the lower total cost of ownership (TCO), and the inherent higher reliability with respect to Hard Disk Drives (HDDs) [2].Among the many parameters of the drive that exercise the reliability/performance trade-off [3], one of the key aspects that is perceived by the users of the SSD platforms concerns their Quality of Service (QoS) [4,5,6]
Through a co-simulation framework that accounts for the measured 3D NAND Flash program suspend characteristics, the program suspend’s impact on the figures of merit of an SSD, including the throughput, latency, QoS, and power consumption
The program operation was implemented in 3D NAND Flash memories by following two consecutive steps: first, the data to be programmed in a specific memory location are transferred from the host and loaded in an on-memory structure called page-buffer [27]; second, an iterative algorithm based on the Incremental Step Pulse Program (ISPP) and verify concept [28] is applied on the target Flash page
Summary
The data storage in enterprise scenarios, including High Performance Computing (HPC) and cloud-based services, requires the low latency and the high throughput that only Solid State Drives (SSD) architectures can deliver [1]. 3D NAND Flash-based SSDs are the preferred solution due to the offered large storage density, the lower total cost of ownership (TCO), and the inherent higher reliability with respect to Hard Disk Drives (HDDs) [2]. To further complicate the picture, we must remember the 3D NAND Flash characteristics: read and write operation granularity is at the page level (from 4 kB to 16 kB in state-of-the-art devices [11]); since the in-place data update operation is prohibited, there is the need of a block erase operation (sized several pages and lasting from 10 ms to 15 ms [9]) that deletes all the invalid data to make room for the updated ones This free space-reclaim operation, known as garbage collection (GC) [12,13], can be very long if it encompasses many blocks (several hundreds of milliseconds) and is managed by the SSD controller during the idle times. Simulations are performed with synthetic benchmarks using different read/write ratios and different micro-architectural drive parameters for design space exploration
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
