Durability and crash recovery in distributed in-memory storage systems

Placeholder Show Content

Abstract/Contents

Abstract
This dissertation presents fast crash recovery for the RAMCloud distributed in-memory data center storage system. RAMCloud is designed to operate on thousands or tens-of-thousands of machines, and it stores all data in DRAM. Rather than replicating in DRAM for redundancy, it provides inexpensive durability and availability by recovering quickly after server crashes. Overall, its goal is to reconstitute the entire DRAM contents of a server and to restore full performance to the cluster in 1 to 2 seconds after failures. Consequently, RAMCloud provides continuous availability by recovering from failures so quickly that applications never notice failures. To achieve this, RAMCloud scatters backup data across thousands of disks, and it harnesses hundreds of servers in parallel to reconstruct lost data. The system uses a decentralized log-structured approach for all of its data, in DRAM as well as on disk; this provides high performance both during normal operation and during recovery. RAMCloud employs randomized techniques at several key points to balance load and to manage the system in a scalable and decentralized fashion. In an 80-node cluster, RAMCloud recovers 40 GB of data from a failed server in 1.86 seconds, and the approach scales to recover larger servers in the same period as cluster size scales.

Description

Type of resource text
Form electronic; electronic resource; remote
Extent 1 online resource.
Publication date 2013
Issuance monographic
Language English

Creators/Contributors

Associated with Stutsman, Ryan Scott
Associated with Stanford University, Department of Computer Science.
Primary advisor Ousterhout, John K
Thesis advisor Ousterhout, John K
Thesis advisor Mazières, David (David Folkman), 1972-
Thesis advisor Rosenblum, Mendel
Advisor Mazières, David (David Folkman), 1972-
Advisor Rosenblum, Mendel

Subjects

Genre Theses

Bibliographic information

Statement of responsibility Ryan Scott Stutsman.
Note Submitted to the Department of Computer Science.
Thesis Ph.D. Stanford University 2013
Location electronic resource

Access conditions

Copyright
© 2013 by Ryan Scott Stutsman
License
This work is licensed under a Creative Commons Attribution Non Commercial 3.0 Unported license (CC BY-NC).

Also listed in

Loading usage metrics...