Deterministic service guarantees for nand flash using partial block cleaning

Choudhuri, Siddharth; Givargis, Tony

doi:10.1145/1450135.1450141

Cited by 58 publications

(42 citation statements)

References 8 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…These studies propose either partial cleaning policies with the help of additional memory [18,19], file system support [20], or pre-emptible garbage collection [21], but they do not consider the context of RAID. To guarantee average I/O latency while migrating data, [22] presents a control-theoretic approach which dynamically adjusts the speed of data migration by periodically measuring I/O performance of the magnetic storage devices.…”

Section: Motivation and Related Workmentioning

confidence: 99%

Reliability and performance enhancements for SSD RAID

McEwan

Komsul

2017

Microprocessors and Microsystems

View full text Add to dashboard Cite

NAND based solid state storage devices are almost ubiquitously used in safety-critical embedded devices, and recent advances have demonstrated RAID architectures specific to solid state storage devices resulting in increased data reliability, with architectural enhancements to solve the age convergence problem. However, these techniques require devices to be taken off-line while components are replaced-consequently these devices are of limited use in hard real time systems. There are further real time issues in that the conventional architectures ignore other characteristics of solid state devices such as garbage collection and meta data management. In this paper we investigate techniques that support the replacement of aged devices in the array in such a way that we provide continuous system reliability. We also improve the performance overhead of the reconstruction process using a novel data migration policy. The techniques are implemented and tested in a trace-driven simulator, and results demonstrate that average I/O response time is improved by up to 39% with improvement by up to 45% in its standard deviation, overheads in terms of device replacement time are negligible, and read performance is improved by an average of 8%.

show abstract

Section: Motivation and Related Workmentioning

confidence: 99%

Reliability and performance enhancements for SSD RAID

McEwan

Komsul

2017

Microprocessors and Microsystems

View full text Add to dashboard Cite

show abstract

“…To shorten the latency introduced by garbage collection, the partial garbage collection mechanism is proposed in GFTL [2]. With this mechanism, for the same garbage collection operation in Figure 1(a), the operation is divided into several small steps, and each step interleaves with the service of a read or write request.…”

Section: B Garbage Collectionmentioning

confidence: 99%

“…In [2], partial garbage collection is proposed to divide a garbage collection operation into several small steps and the time of each step is not longer than that of one erase operation. By interleaving each partial garbage collection step with the service of a read/write request, the worst system latency is deterministic, as the longest operation time in NAND flash is to erase a block in which it cannot respond to I/O requests.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Optimizing deterministic garbage collection in NAND flash storage systems

Zhang

Wang

et al. 2015

21st IEEE Real-Time and Embedded Technology and Applications Symposium

View full text Add to dashboard Cite

NAND flash has been widely adopted as storage devices in real-time embedded systems. However, garbage collection is needed to reclaim space and introduces a lot of time overhead. As the worst system latency is determined by the worst-case execution time of garbage collection in NAND flash, it is important to optimize garbage collection so as to give a deterministic worst system latency. On the other hand, since the garbage collection does not happen very often, optimizing garbage collection should not bring too much overhead to the average system latency.This paper presents for the first time a worst-case and average-case joint optimization scheme for garbage collection in NAND flash. With our scheme, garbage collection can be postponed to the latest stage so improves the average system latency. By combining partial garbage collection and over-provisioning, our scheme can guarantee that one free block is enough to hold all pages from both write requests and valid-page copies. The experiments have been conducted on a real embedded platform and the results show that our technique can improve both worstcase and average-case system latency compared with the previous works.

show abstract

“…The conclusions in [13], supports our motivation for the lack of real-time, deterministic guarantees for flash. The results on wear level and details on benchmark performance is in [14].…”

Section: Related Workmentioning

confidence: 99%

Real-Time Access Guarantees for NAND Flash Using Partial Block Cleaning

Choudhuri

Givargis

2008

Software Technologies for Embedded and Ubiquitous Systems

Self Cite

View full text Add to dashboard Cite

Abstract. Increasing use of NAND flash in newer application domains has been possible due to lowering cost per GB, consumer demands for storage and advantages of NAND flash over traditional disks. However, NAND flash has its idiosyncrasies resulting in asymmetric read/write times due to garbage collection and wear leveling requirements. Such asymmetric (non-deterministic) read/write times poses a challenge for the adoption of NAND flash in real-time systems.We present the implementation details of a flash translation layer called GFTL that guarantees strict upper bounds on read/write times that are comparable to a theoretical ideal case. Such guarantees are made possible by dividing the source of non-determinism into deterministic intervals using our proposed approach called partial block cleaning. Using partial block cleaning, the process of garbage collection is divided into several smaller, deterministic steps. Partial block cleaning comes with an overhead of additional space requirements. We provide a proof on the limit of the additional space requirements.

show abstract

Deterministic service guarantees for nand flash using partial block cleaning

Cited by 58 publications

References 8 publications

Reliability and performance enhancements for SSD RAID

Reliability and performance enhancements for SSD RAID

Optimizing deterministic garbage collection in NAND flash storage systems

Real-Time Access Guarantees for NAND Flash Using Partial Block Cleaning

Contact Info

Product

Resources

About