Chapter

Accelerating Virtualized Distributed NVMe Storage in Hardware

April 2021

DOI:10.1002/9781119818410.ch4

In book: Multi‐Processor System‐on‐Chip 2 (pp.69-94)

Authors:

Michail D. Flouris

OnApp Ltd., UK

Modern NVMe Flash devices deliver over 1 million random I/O operations per second (IOPS) at sub‐100 µs latencies, making them orders of magnitude faster than traditional hard drives with rotating disks, especially for random I/O. In this chapter, the authors explain the challenges of high‐performance cloud computing, especially with respect to storage I/O operations, and present their solution to accelerating virtualized distributed NVMe storage with hardware. They briefly discuss the concept of hyperconverged infrastructure (HCI) versus disaggregated cloud infrastructure, providing insights into the advantages of each type of setup. Then, the authors proceed with an overview of the NexVisor platform and explain how it has been designed to minimize I/O overheads and provide I/O performance that is equivalent to servers running baremetal OS.

ResearchGate has not been able to resolve any citations for this publication.

The Datacenter as a Computer, An Introduction to the Design of Warehouse-Scale Machines

Article

Jan 2009

MDev‐NVMe: A NVMe storage virtualization solution with mediated pass‐through, Proceedings of the 2018 USENIX Annual Technical Conference

Jan 2018

B. Peng
H. Zhang
J. Yao
Y. Dong
Y. Xu
H. Guan

The Datacenter as a Computer: An Introduction to the Design of Warehouse‐Scale Machines

Jan 2009

L. Barroso
U. Hülzle

Article

Full-text available

Building a Virtualized Distributed Computing Building a Virtualized Distributed Computing Building a...

January 1970

Abstract Abstract Abstract In this article, we present the realization of a system, termed as InterGrid, for interconnecting distributed computing infrastructures by harnessing virtual machines. The InterGrid aims to provide an execution environment for running applications on top of the interconnected infrastructures. The system uses virtual machines as the building blocks to construct execution ... [Show full abstract] environments that span multiple computing sites. An execution environment is a network of virtual machines created to fulfill the requirements of an application, thus running isolated from other execution environments. These environments can be extended to operate on Cloud infrastructure, such as Amazon EC2. The article provides an abstract view of the proposed architecture and its implementation; experiments show the scalability of an infrastructure managed by InterGrid and how the system can benefit from using Cloud infrastructure.

Article

Building a Virtualized Distributed Computing Building a Virtualized Distributed Computing Building a...

January 2009 · IEEE Internet Computing

In this article, we present the realization of a system, termed as InterGrid, for interconnecting distributed computing infrastructures by harnessing virtual machines. The InterGrid aims to provide an execution environment for running applications on top of the interconnected infrastructures. The system uses virtual machines as the building blocks to construct execution environments that span ... [Show full abstract] multiple computing sites. An execution environment is a network of virtual machines created to fulfill the requirements of an application, thus running isolated from other execution environments. These environments can be extended to operate on Cloud infrastructure, such as Amazon EC2. The article provides an abstract view of the proposed architecture and its implementation; experiments show the scalability of an infrastructure managed by InterGrid and how the system can benefit from using Cloud infrastructure.

Article

What Modern NVMe Storage Can Do, and How to Exploit it: High-Performance I/O for High-Performance St...

July 2023 · Proceedings of the VLDB Endowment

NVMe SSDs based on flash are cheap and offer high throughput. Combining several of these devices into a single server enables 10 million I/O operations per second or more. Our experiments show that existing out-of-memory database systems and storage engines achieve only a fraction of this performance. In this work, we demonstrate that it is possible to close the performance gap between hardware ... [Show full abstract] and software through an I/O optimized storage engine design. In a heavy out-of-memory setting, where the dataset is 10 times larger than main memory, our system can achieve more than 1 million TPC-C transactions per second.

Conference Paper

Environment Virtualized Distributed Storage System Deployment and Effectiveness Analysis

July 2015

In recent years, with the booming server virtualization, many applications extend topics, such as: PaaS [1], IaaS [2], SaaS [3] and other cloud services concept, but whether it is the kind of virtualization technology, after end storage I / O performance and high availability is a key factor in quality of service. This paper will explore vSAN distributed storage system, the differences in ... [Show full abstract] different principles, and understand the differences in the effectiveness of a number of different data node configuration, for the production of environmental planning reference.