[ambiate]

RAID over IP via broadcast? From a security standpoint, it would seem there are many hurdles to jump. Not too long ago, I talked to Dickinson about what you guys do. I never considered the amount of harddrives necessary to facilitate Swiftstack's goals. Of course the software challenges are interesting, but the amount of hardware and maintaining risk/DR/costs/etc. Oh man. Hats off to you guys.

[notmyname]

It's not RAID over IP. Each drive itself is a network endpoint.

On the practical side, this means that the drive is "connected" to one or more servers (and can be re-homed on the fly). The storage system itself is responsible for coordinating the communication across the cluster of Kinetic drives. In our case, the coordinating storage system is OpenStack Swift.

On the humorous side, this means that technobabble like "Can you ping the boot record?" or "What does a traceroute to the directory show?" actually sorta make sense now.

[ambiate]

RAID over IP was a joke. I remember reading this story about someone flooding routers to turn them into a broadcasts to view the flow of packets over the network. I imagined the same thing, then said, hey with HDDs and broadcast, you could simulate RAID. I made a long stretch on that one.

[mbreese]

That's probably not too far off. If you wanted to do mirror a write, you could imagine broadcasting that same packet to a set of devices. The devices would have to figure out whether or not they needed to store the packet. Otherwise, you'd be sending N packets to the same switch for N-way mirroring. In a typical hardware RAID, the sector write is sent to the RAID controller once.

It's probably not too long of a stretch.

[toomuchtodo]

Writing with multicast and using tcp only for the control plane? Interesting concept.

[wires]

lol @ technobabble