The Aurora Single Level Store Operating System

Tsalapatis, Emil; Hancock, Ryan; Barnes, Tavian; Mashtizadeh, Ali José

doi:10.1145/3477132.3483563

Cited by 9 publications

(3 citation statements)

References 22 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…POS continues the line of research on C/R for different tasks, with a particular focus on OSlevel C/R for GPU applications and on efficient and concurrent C/R execution. Though concurrent OS-level C/R has been extensively explored for processes that only run on the CPU [14,40,24,66,71,10,35,25,77], to the best of our knowledge, existing C/R systems that support GPU all leverage a stop-the-world design [48,47,21,15,67], which we have shown has notable performance degradation or applica-tion downtime in various scenarios. Meanwhile, there are also many task-level C/R designs, e.g., for ML tasks [44,75,16], HPC tasks [57] or leverage NVM for acceleration [56].…”

Section: Related Workmentioning

confidence: 99%

See 1 more Smart Citation

School of Mechanical Engineering, Shanghai Jiao Tong University, 200240, Shanghai, China

Zhuang

Huang

et al. 2019

Mathematical Biosciences and Engineering

View full text Add to dashboard Cite

The crystallization kinetics and melting behavior of nylon 10,10 in neat nylon 10,10 and in nylon 10,10 -montmorillonite (MMT) nanocomposites were systematically investigated by differential scanning calorimetry. The crystallization kinetics results show that the addition of MMT facilitated the crystallization of nylon 10,10 as a heterophase nucleating agent; however, when the content of MMT was high, the physical hindrance of MMT layers to the motion of nylon 10,10 chains retarded the crystallization of nylon 10,10, which was also confirmed by polarized optical microscopy. However, both nylon 10,10 and nylon 10,10 -MMT nanocomposites exhibited multiple melting be-havior under isothermal and nonisothermal crystallization conditions. The temperature of the lower melting peak (peak I) was independent of MMT content and almost remained constant; however, the temperature of the highest melting peak (peak II) decreased with increasing MMT content due to the physical hindrance of MMT layers to the motion of nylon 10,10 chains.

show abstract

Section: Related Workmentioning

confidence: 99%

“…ML tasks, such as training, are vulnerable to GPU failures [20,16,67]. C/R provides fault tolerance by periodically checkpointing and persisting the images [70,77,67]. Upon failure and recovery, the OS simply restores the task using the checkpointed image, creating an illusion that the tasks never failed.…”

Section: Background and Motivationmentioning

confidence: 99%

School of Mechanical Engineering, Shanghai Jiao Tong University, 200240, Shanghai, China

Zhuang

Huang

et al. 2019

Mathematical Biosciences and Engineering

View full text Add to dashboard Cite

show abstract

“…The mechanisms of checkpoint and restore (C/R) has been investigated by OSes for a long time [21,47]. For examples, KeyKOS [30], EROS [59], and Aurora [64] embrace the abstraction of single level store and make the whole-system checkpoint periodically for crash consistency; many researches on Linux [31,41,4,76,66] can dynamically generate applications' checkpoints which can be restored after crashes or on other machines. VAS-CRIU [66] also notices the inefficient of C/R brought by the file abstraction.…”

Section: Related Workmentioning

confidence: 99%

No Provisioned Concurrency: Fast RDMA-codesigned Remote Fork for Serverless Computing

Wei¹,

Lu²,

Wang³

et al. 2022

Preprint

View full text Add to dashboard Cite

We present MITOSIS, an OS primitive to support fast remote fork with minimal resource provisioning for serverless computing. MITOSIS is the first to fork over 10,000 new containers from only one instance across multiple machines within a second. It also allows transparent and efficient intermediate data sharing between producer and consumer serverless functions even across machines. The key enabler for this high speed is an RDMA-OS co-design, which imitates local fork with RDMA's remote memory read capability bypassing extra software overhead for starting and running containers. We also introduce techniques for lightweight connection and memory management based on advanced RDMA transport and fast containerization in a distributed setting. We have implemented MITOSIS on Linux and integrated it to FN, a popular serverless computing framework. Under load spikes in real serverless workloads, MITOSIS reduces the FN tail latency by 95.2% with orders of magnitude smaller memory usage.

show abstract

KalpaVriksh: Efficient and Cost-effective GUI Application Hosting using Singleton Snapshots

Shaikh

Kumar

Mishra

2023

2023 IEEE/ACM 23rd International Symposium on Cluster, Cloud and Internet Computing (CCGrid)

View full text Add to dashboard Cite

The Aurora Single Level Store Operating System

Cited by 9 publications

References 22 publications

School of Mechanical Engineering, Shanghai Jiao Tong University, 200240, Shanghai, China

School of Mechanical Engineering, Shanghai Jiao Tong University, 200240, Shanghai, China

No Provisioned Concurrency: Fast RDMA-codesigned Remote Fork for Serverless Computing

KalpaVriksh: Efficient and Cost-effective GUI Application Hosting using Singleton Snapshots

Contact Info

Product

Resources

About