Thursday Jul 19, 2007
Thursday Jul 19, 2007
Cluster File Systems' use of ZFS as the disk filesystem inside Lustre is a perfect example of how open source can be used to advance the state of data management. Lustre is a distributed filesystem that can run across thousands of clustered server nodes, all sharing potentially petabytes of data. And like the challenges faced by anyone trying to manage large data sets these days, Lustre needed better scalability, reliability and storage management features than those available in their current internal disk filesystem. So the makers of Lustre had a few choices: start developing enhancements to their existing internal filesystem, build a new one, or pick one up from another open source community. Turns out to be an easy choice - ZFS is a 128 bit filesystem (more scale than any of us can use these days), has built in data integrity through its checksumming algorithms, and handles storage management internally via RAIDZ (no need to define RAID stripes and disk pools separately). Most requirements Lustre had and probably some they haven't even thought of yet are satisfied by ZFS. It's pretty cool that ZFS is solving out-of-this-world problems.
A friend emailed me this picture she took in the Palo Alto Fish Market parking lot of this weird green-glowing thing about to take off...
Click on image for more information.
Seems like ZFS might really be out-of-this-world... [yeah, I know, maybe her camera phone just doesn't take clear shots
]
Here's a picture.
Posted by Kevin on July 19, 2007 at 07:59 PM EDT #