Optimistic crash consistency

Chidambaram, Vijay; Pillai, Thanumalayan Sankaranarayana; Arpaci-Dusseau, Andrea C.; Arpaci-Dusseau, Remzi H.

doi:10.1145/2517349.2522726

Cited by 114 publications

(67 citation statements)

References 12 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…TxFlush is used to explicitly write data to persistent memory from the CPU cache. In LOC, durability and atomicity are decoupled, similarly to the approaches of [26,10,34]. TxCommit and TxFlush are combined to provide both atomicity and durability in LOC.…”

Section: A Interfacementioning

confidence: 99%

“…First, a new transaction commit protocol, Eager Commit, enables the commit of a transaction without the use of commit records, traditionally employed for storage systems to record the status of each transaction (which is needed for recovery purposes on system crash) [17,19,20,26]. Doing so removes the need to perform a persistent commit record write at the end of a transaction and eliminates the intra-tx ordering requirement, improving performance.…”

Section: Introduction Emerging Non-volatile Memory (Nvm) Technologmentioning

confidence: 99%

See 1 more Smart Citation

Loose-Ordering Consistency for persistent memory

Shu

Sun

et al. 2014

2014 IEEE 32nd International Conference on Computer Design (ICCD)

124

View full text Add to dashboard Cite

Abstract-Emerging non-volatile memory (NVM) technologies enable data persistence at the main memory level at access speeds close to DRAM. In such persistent memories, memory writes need to be performed in strict order to satisfy storage consistency requirements and enable correct recovery from system crashes. Unfortunately, adhering to a strict order for writes to persistent memory significantly degrades system performance as it requires flushing dirty data blocks from CPU caches and waiting for their completion at the main memory in the order specified by the program. This paper introduces a new mechanism, called Loose-Ordering Consistency (LOC), that satisfies the ordering requirements of persistent memory writes at significantly lower performance degradation than stateof-the-art mechanisms. LOC consists of two key techniques. First, Eager Commit reduces the commit overhead for writes within a transaction by eliminating the need to perform a persistent commit record write at the end of a transaction. We do so by ensuring that we can determine the status of all committed transactions during recovery by storing necessary metadata information statically with blocks of data written to memory. Second, Speculative Persistence relaxes the ordering of writes between transactions by allowing writes to be speculatively written to persistent memory. A speculative write is made visible to software only after its associated transaction commits. To enable this, our mechanism requires the tracking of committed transaction ID and support for multi-versioning in the CPU cache. Our evaluations show that LOC reduces the average performance overhead of strict write ordering from 66.9% to 34.9% on a variety of workloads.

show abstract

Section: A Interfacementioning

confidence: 99%

Section: Introduction Emerging Non-volatile Memory (Nvm) Technologmentioning

confidence: 99%

Loose-Ordering Consistency for persistent memory

Shu

Sun

et al. 2014

2014 IEEE 32nd International Conference on Computer Design (ICCD)

124

View full text Add to dashboard Cite

show abstract

“…Although the fsync latency problem can be alleviated by reducing the delay of the quasi-async requests under the boosting technique, it can manage only the requests in the host queue without controlling the requests on the device queue. OptFS [8] proposed a new system call, called osync, that assures only the ordering of the file operations. It can guarantee the durability of sync requests by receiving notifications from storage.…”

Section: Related Workmentioning

confidence: 99%

Optimizing Fsync Performance with Dynamic Queue Depth Adaptation

Park¹,

Kim²,

Shin

2015

JSTS:Journal of Semiconductor Technology and Science

View full text Add to dashboard Cite

Abstract-Existing flash storage devices such as universal flash storage and solid state disk support command queuing to improve storage I/O bandwidth. Command queuing allows multiple read/write requests to be pending in a device queue. Because multi-channel and multi-way architecture of flash storage devices can handle multiple requests simultaneously, command queuing is an indispensable technique for utilizing parallel architecture. However, command queuing can be harmful to the latency of fsync system call, which is critical to application responsiveness. We propose a dynamic queue depth adaptation technique, which reduces the queue depth if user application is expected to send fsync calls. Experiments show that the proposed technique reduces the fsync latency by 79% on average compared to the original scheme.

show abstract

“…In addition, the lack of well-defined file-system behavior prevents careful applications from optimizing their crash-invariance protocols, thus reducing efficiency. Previous techniques for file-system internal consistency, such as journaling [22,23], soft updates [4], copy-on-write [1,6,13], backpointer-based consistency [3], and optimistic crash consistency [2], have not investigated their (unintended) consequences on application-level consistency.…”

Section: Introductionmentioning

confidence: 99%

Towards efficient, portable application-level consistency

Pillai

Chidambaram

Hwang

et al. 2013

Proceedings of the 9th Workshop on Hot Topics in Dependable Systems

Self Cite

View full text Add to dashboard Cite

Applications employ complex protocols to ensure consistency after system crashes. Such protocols are affected by the exact behavior of file systems. However, modern file systems vary widely in such behavior, reducing the correctness and performance of applications. In this paper, we study application-level crash consistency. Through the detailed study of two popular database libraries (SQLite, LevelDB), we show that application performance and correctness heavily depend on file-system properties previously ignored in research. We define a number of such properties and show that they vary widely among file systems. We conclude with implications for future file-system and dependability research.

show abstract

Optimistic crash consistency

Cited by 114 publications

References 12 publications

Loose-Ordering Consistency for persistent memory

Loose-Ordering Consistency for persistent memory

Optimizing Fsync Performance with Dynamic Queue Depth Adaptation

Towards efficient, portable application-level consistency

Contact Info

Product

Resources

About