Seagate connects Hadoop and Lustre in an open sourcery ceremony

Lustre? meh

Never much liked lustre, maybe it doesn't suit my workloads, or my hardware, or wasn't installed properly. The bumf says that it stripes writes (thence reads) over many disks thus eliminating the bottleneck from only reading X bits per second from a single disk.

Doesn't fix latency limitations of the drives, and of course adds comms overhead to spread the info around. In theory this should still be good for megamega big files but in practice I and the other admins and developers on our system never got much joy from it.