Cloud service providers are deploying large numbers of industry standard, X86-based, systems. For the most part, however, they're focused on purchasing the lowest cost systems rather than purchasing more expensive systems that have special features that increase levels of performance, scalability, or reliability. This also means that they seldom purchase fault tolerant, everything duplicated, systems that avoid disasters.
If asked, the decision-makers at these firms would be very likely to point out that their profitability is tied very closely to reducing the cost of hardware acquisition and maintenance. Even a small increase in the cost of systems would simply overwhelm their margins. So, they focus on using various forms of virtualization software to get to the same goal of avoiding disasters.
Cloud service providers appear to be relying on the tools offered by suppliers of processing virtualization, application virtualization and management software designed for virtual environments as a way to address the need for never-fail systems even though these tools will not offer the near-instantaneous response offered by special purpose hardware.
I predict that we will see a new class of processing virtualization technology, software that lives just below the hypervisor, to be announced in the coming years. This software will duplicate the thought behind today's fault tolerant systems and respond to hardware outages much more rapidly and produce much higher levels of availability.
Suppliers, such as Stratus and Marathon Technologies (recently acquired by Stratus), are likely to lead this charge. It is also possible that VMware, Citrix and Microsoft may produce technology that addresses the same requirement.