This is a silly press release, by a silly marketing department with no grasp on reality. Must be something they are smoking.
In reality, customers install no more than 10 nodes. The performance of the cluster falls off a cliff as you add more nodes than that. That negates anything else that they've published, all of which assumes linear scalability in every measure.
In reality, you can't get 12000 disk IOPS out of a 12 drive SAS or 36 drive SATA system as they are claiming. Max possible disk IOPS for those drives types would be 4800 (12 drive SAS) or 7200 (36 drive SATA), and only under ideal full sequential IOPS testing.
In reality, they aren't delivering 5.2 PB of USABLE capacity, even if we are generous and concede that they could deliver a 144 node cluster. At best, they might achieve over 4.5PB of usable capacity (if we're generous about overhead), but ONLY if they could actually manage parity across 144 nodes.