Towards building a high-performance, scale-in key-value storage system

Kang, Yangwook; Pitchumani, Rekha; Mishra, Pratik; Kee, Yang-Suk; Londoño, Federico Uribe; Oh, Sangyoon; Lee, Jongyeol; Lee, Daniel D. G.

doi:10.1145/3319647.3325831

Cited by 27 publications

(18 citation statements)

References 8 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…This has recently enabled porting RocksDB and BlueFS ( § 4.1) to run on host-managed SMR drives, and an effort is underway to store object data on such drives next [3]. In addition, the Ceph community is exploring a new backend that targets a combination of persistent memory and emerging NVMe devices with novel interfaces, such as ZNS SSDs [9,26] and KV SSDs [51].…”

Section: Exploring New Interfacesmentioning

confidence: 99%

File systems unfit as distributed storage backends

Aghayev

Weil

Kuchnik

et al. 2019

Proceedings of the 27th ACM Symposium on Operating Systems Principles

View full text Add to dashboard Cite

For a decade, the Ceph distributed file system followed the conventional wisdom of building its storage backend on top of local file systems. This is a preferred choice for most distributed file systems today because it allows them to benefit from the convenience and maturity of battle-tested code. Ceph's experience, however, shows that this comes at a high price. First, developing a zero-overhead transaction mechanism is challenging. Second, metadata performance at the local level can significantly affect performance at the distributed level. Third, supporting emerging storage hardware is painstakingly slow. Ceph addressed these issues with BlueStore, a new backend designed to run directly on raw storage devices. In only two years since its inception, BlueStore outperformed previous established backends and is adopted by 70% of users in production. By running in user space and fully controlling the I/O stack, it has enabled space-efficient metadata and data checksums, fast overwrites of erasure-coded data, inline compression, decreased performance variability, and avoided a series of performance pitfalls of local file systems. Finally, it makes the adoption of backwards-incompatible storage hardware possible, an important trait in a changing storage landscape that is learning to embrace hardware diversity. CCS Concepts • Information systems → Distributed storage; • Software and its engineering → File systems management; Software performance.

show abstract

Section: Exploring New Interfacesmentioning

confidence: 99%

File systems unfit as distributed storage backends

Aghayev

Weil

Kuchnik

et al. 2019

Proceedings of the 27th ACM Symposium on Operating Systems Principles

View full text Add to dashboard Cite

show abstract

“…This high CPU utilization comes from multiple software stack layers required to translate from key space to device block space including key-value data management, file system, block I/O layer, device driver, etc. [16]. Another problem of modern key-value stores are high I/O write amplification [20,30].…”

Section: Introductionmentioning

confidence: 99%

“…Although Log Structure Merge Trees significantly reduce the WAF compared to B-Trees [25], they still suffer from relatively high WAF (∼10-40x) due to background compaction [20,23,30]. Recently, both academia [11,14,31] and industry [16] have proposed key-value interfaced devices to replace the conventional software-based key-value engines built on block interface devices. Key-value SSDs simplify the software stack for key-value store applications, reducing their overall CPU/memory usage [16].…”

Section: Introductionmentioning

confidence: 99%

Kvraid

Qin¹,

Reddy²,

Gratz³

et al. 2021

Proceedings of the 14th ACM International Conference on Systems and Storage

Self Cite

View full text Add to dashboard Cite

Key-value (KV) stores have been widely deployed in a variety of scale-out enterprise applications such as online retail, big data analytics, social networks, etc. Key-Value SSDs (KVSSDs) provide a key-value interface directly from the device aiming at lowering software overhead and reducing I/O amplification for such applications.In this paper, we present KVRAID, a high performance, write efficient erasure coding management scheme on emerging key-value SSDs. The core innovation of KVRAID is to use logical to physical key conversion to efficiently pack similar size KV objects and dynamically manage the membership of erasure coding groups. Such design enables packing multiple user objects to a single physical object to reduce the object amplification compared to prior works. By applying out-ofplace update technique, KVRAID can significantly reduce the I/O amplification compared to the state-of-art designs. Our experiments show that KVRAID outperforms state-of-art software KV-store with block RAID by 28x in terms of insert throughput and reduces CPU utilization, tail latency and write amplification significantly. Compared to state-of-art KV devices erasure coding management, KVRAID reduces object amplification by ∼ 2.6 compared to StripeFinder and reduces I/O amplification by ∼ 9.6 when compared to KVMD and StripeFinder for update intensive workloads. CCS CONCEPTS• Hardware → Emerging interfaces; Emerging architectures; • Computer systems organization → Redundancy; Reliability.

show abstract

“…Starting from Active Disks [11,14,17,18,22], moving high-selectivity data access functions to storage devices gains increasing research interest mainly because of the conceivable benefits [10] such as reducing the size of data transmission between hosts and storage devices, reducing total power consumption, increasing overall resource utilization, and simplifying the application design by leveraging high-level storage semantics. For example, key-value smart storage devices can substitute the translation layers from key-value down to physical block address, which includes a key-value to file translation in the front, a file to logical block address translation in the middle, and a logical block address to physical block address translation at the bottom.…”

Section: Introductionmentioning

confidence: 99%

MBWU: Benefit Quantification for Data Access Function Offloading

Jian-shen

Kufeldt

Maltzahn

2019

Lecture Notes in Computer Science

View full text Add to dashboard Cite

The storage industry is considering new kinds of storage devices that support data access function offloading, i.e. the ability to perform data access functions on the storage device itself as opposed to performing it on a separate compute system to which the storage device is connected. But what is the benefit of offloading to a storage device that is controlled by an embedded platform, very different from a host platform? To quantify the benefit, we need a measurement methodology that enables apple-to-apple comparisons between different platforms. We propose a Media-based Work Unit (MBWU, pronounced "MibeeWu"), and an MBWU-based measurement methodology to standardize the platform efficiency evaluation so as to quantify the benefit of offloading. To demonstrate the merit of this methodology, we implemented a prototype to automate quantifying the benefit of offloading the key-value data access function.

show abstract

Towards building a high-performance, scale-in key-value storage system

Cited by 27 publications

References 8 publications

File systems unfit as distributed storage backends

File systems unfit as distributed storage backends

Kvraid

MBWU: Benefit Quantification for Data Access Function Offloading

Contact Info

Product

Resources

About