Many servers today employ virtualization, where multiple operating systems, called VMs, coexist within a single hardware server. This creates the illusion that multiple applications are coexisting peacefully and independently within one piece of hardware.
But at the bottom, all the I/O from all the VMs fight over a single set of disk arms, like the dogs above fighting over a steak. Requests from each VM get shuffled in with requests from other VMs, so that a series of sequential requests from one VM, which would have a low seek overhead because of their proximity to each other, becomes a series of random requests, which has high seek overhead, driving the disks to saturation.
This is a situation where NewServe can really help. By allowing each VM to get its data with less physical I/O, the hardware is able to support more VMs before saturating