Many companies have heard of the large number of benefits that are afforded by the use of virtual servers and a move away from physical storage space to a disk-to-Web solution. By limiting the amount of time an internal IT team needs to be spending on server maintenance and with the ability to run multiple O/S and applications of one single virtual machine, a company can often find a great deal to like about using VPS technology as their new applications system.
Unfortunately, the nature of the virtual system is that storage becomes a much larger problem than in a physical environment. This is largely due to the random nature of the virtual environment in which data rewrites can happen at any time and without warning across multiple server instances, meaning that I/O usage is inherently random. This unpredictable I/O use, coupled with the often large number of applications running on a virtual server at once means a greater strain on the storage subsystem and lowered performance overall.
Part of the problem lies in the fact that the hypervisor is required to approve many of the storage actions taken on a VM machine, effectively bottlenecking the storage system during random times of high usage. By forcing the storage array to write zeros during formatting or allowing the storage system to remove a VM without other permissions, it is possible to limit the amount of time that a VPS solution spends bogged down due to random server pressure.
Storage solutions are still evolving in order to allow hypervisors to have the most control while still balancing the amount of second-by-second decisions that need to be made, limiting overall slowdown, and viable storage solutions are an essential part of this process.