Castle: Re-inventing storage for Big Data

In this talk, Tom Wilkie looks at Castle, an open-source project, is a ground-up overhauling of RAID, file systems, and the POSIX interface

The standard Linux storage stack wasn’t designed for write-heavy big data workloads, nor is it well-suited to modern hardware: large, slow SATA disks, SSDs or many cores. Castle, an open-source project, is a ground-up overhauling of RAID, file systems, and the POSIX interface. It is released under the GPL and runs as part of the Linux kernel. Our target is 1 million random inserts per second to disk on a $1,000 commodity box, and we’re nearly there. Castle is also the core of the Acunu Data Platform, which delivers up to 100x higher performance for applications written for Cassandra and other tools.

YOU MAY ALSO LIKE:

• Microservices from the very start with Russel Winder! (in London on 26th June 2017)
• Infiniteconf 2017 - the conference on Big Data and Fast Data (in London on 6th - 7th July 2017)
• Russ Miles' Microservices Workshop (in London on 4th - 6th September 2017)
• Uncle Bob's Advanced TDD (in London on 30th - 31st October 2017)