A Case for Optimistic Coordination in HPC Storage Systems

Carns, Philip; Harms, Kevin; Kimpe, Dries; Wozniak, Justin M.; Ross, Robert; Ward, Lee; Curry, Matthew; Danielson, Geoff; Karakoyunlu, Cengiz; Chandy, John A.; Settlemyer, Bradley W.; Gropp, William

doi:10.1109/sc.companion.2012.19

Cited by 5 publications

(3 citation statements)

References 16 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Finally, WTF does not provide the atomic operations supported by Týr's design. Such atomic operations have previously been proposed on distributed file systems by [36], but this work unfortunately does not focus on integrating them with a transactional file system, and did not base its evaluation on a real file system.…”

Section: Related Workmentioning

confidence: 99%

Týr: Blob Storage Meets Built-In Transactions

Matri

Costan

Antoniu

et al. 2016

SC16: International Conference for High Performance Computing, Networking, Storage and Analysis

View full text Add to dashboard Cite

Concurrent Big Data applications often require high-performance storage, as well as ACID (Atomicity, Consistency, Isolation, Durability) transaction support. Although blobs (binary large objects) are an increasingly popular model for addressing the storage needs of such applications, state-of-theart blob storage systems typically offer no transaction semantics. This demands users to coordinate access to data carefully in order to avoid race conditions, inconsistent writes, overwrites and other problems that cause erratic behavior. We argue there is a gap between existing storage solutions and application requirements, which limits the design of transaction-oriented applications. We introduce Týr, the first blob storage system to provide builtin, multiblob transactions, while retaining sequential consistency and high throughput under heavy access concurrency. Týr offers fine-grained random write access to data and in-place atomic operations. Large-scale experiments on Microsoft Azure with a production application from CERN LHC show Týr throughput outperforming state-of-the-art solutions by more than 75%.

show abstract

Section: Related Workmentioning

confidence: 99%

Týr: Blob Storage Meets Built-In Transactions

Matri

Costan

Antoniu

et al. 2016

SC16: International Conference for High Performance Computing, Networking, Storage and Analysis

View full text Add to dashboard Cite

show abstract

“…If a conditional write fails, the enclosing transaction is also failed, rolling back any changes previously incurred. More information on uses and performance of conditional updates can be found in previously published work [1] Pessimistic Concurrency Control Sirocco also provides storage semantics that a client library can leverage to implement traditional leased locks. The basic mechanism is the triggered batch, an operation that is similar to a conditional update, but with three important differences.…”

Section: Optimistic Concurrency Controlmentioning

confidence: 99%

Motivation and Design of the Sirocco Storage System Version 1.0.

Curry

Ward

Danielson

2015

Self Cite

View full text Add to dashboard Cite

Sirocco is a massively parallel, high performance storage system for the exascale era. It emphasizes client-to-client coordination, low server-side coupling, and free data movement to improve resilience and performance. Its architecture is inspired by peer-to-peer and victimcache architectures. By leveraging these ideas, Sirocco natively supports several media types, including RAM, flash, disk, and archival storage, with automatic migration between levels.

show abstract

“…Optimistic concurrency control and conditional operations have been used on file data in order to enable richer file system primitives for application coordination [42]- [44]. BatchFS is different from these techniques in that it focuses on metadata concurrency and targets batch applications that do not use the file system as a synchronization mechanism.…”

Section: Related Workmentioning

confidence: 99%

BatchFS: Scaling the File System Control Plane with Client-Funded Metadata Servers

Zheng

Ren

Gibson

2014

2014 9th Parallel Data Storage Workshop

View full text Add to dashboard Cite

Abstract-Parallel file systems are often characterized by a layered architecture that decouples metadata management from I/O operations, allowing file systems to facilitate fast concurrent access to file contents. However, metadata intensive workloads are still likely to bottleneck at the file system control plane due to namespace synchronization, which taxes application performance through lock contention on directories, transaction serialization, and RPC overheads. In this paper, we propose a client-driven file system metadata architecture, BatchFS, that is optimized for noninteractive, or batch, workloads. To avoid metadata bottlenecks, BatchFS features a relaxed consistency model marked by lazy namespace synchronization and optimistic metadata verification. Capable of executing namespace operations on client-provisioned resources without contacting any metadata server, BatchFS clients are able to delay namespace synchronization until synchronization is really needed. Our goal in this vision paper is to handle these delayed operations securely and efficiently with metadata verification and bulk insertion. Preliminary experiments demonstrate that our client-funded metadata architecture outperforms a traditional synchronous file system by orders of magnitude.

show abstract

A Case for Optimistic Coordination in HPC Storage Systems

Cited by 5 publications

References 16 publications

Týr: Blob Storage Meets Built-In Transactions

Týr: Blob Storage Meets Built-In Transactions

Motivation and Design of the Sirocco Storage System Version 1.0.

BatchFS: Scaling the File System Control Plane with Client-Funded Metadata Servers

Contact Info

Product

Resources

About