In this work a thorough design space exploration of Solid State Drives (SSD) architectures is presented. Thanks to SSDExplorer, a virtual platform for SSD simulation, designers will get all the elements to understand how to simulate and efficiently design the architecture of an SSD. It will demonstrated that the main figure of merits of these devices, that are the bandwidth, the latency, the quality of service (QoS), the reliability, and the power consumption, are tightly coupled with the features of the underlying storage layer. Moreover, in contrast with nowadays “top-down” design methodologies where the firmware optimization level represents the sole responsible of the SSD behavior, it will be shown that the proposed “bottom-up” design paradigm allows efficiently tuning the memory features with the final user application. As a consequence, the design of extremely high-performing SSD architectures is enabled. Finally, it will be demonstrated that in future SSDs for hyper-scale environments, a “memory-centric” design flow will represent the unique efficient way to build a high-capable and low-power storage device.
Solid State Drives: design challenges for optimum performance-reliability trade-off
2016
Abstract
In this work a thorough design space exploration of Solid State Drives (SSD) architectures is presented. Thanks to SSDExplorer, a virtual platform for SSD simulation, designers will get all the elements to understand how to simulate and efficiently design the architecture of an SSD. It will demonstrated that the main figure of merits of these devices, that are the bandwidth, the latency, the quality of service (QoS), the reliability, and the power consumption, are tightly coupled with the features of the underlying storage layer. Moreover, in contrast with nowadays “top-down” design methodologies where the firmware optimization level represents the sole responsible of the SSD behavior, it will be shown that the proposed “bottom-up” design paradigm allows efficiently tuning the memory features with the final user application. As a consequence, the design of extremely high-performing SSD architectures is enabled. Finally, it will be demonstrated that in future SSDs for hyper-scale environments, a “memory-centric” design flow will represent the unique efficient way to build a high-capable and low-power storage device.I documenti in UNITESI sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/20.500.14242/145678
URN:NBN:IT:UNIFE-145678