Why 64-bit Hypervisor Support Is a Big Deal (Hint: More vGPUs)

by Will Wade

Citrix just released XenServer 6.5, their first 64-bit hypervisor – software that lets one machine create and run many “virtual” machines.

So, what makes 64 the magic number?

In the past, we’ve operated on 32-bit systems. But the proliferation of new devices that XenServer supports created the need for 64-bit support.

For XenServer to use a device (say, a GPU), the device needs to be mapped into the input/output (I/O) space of XenServer’s management operating system, also known as “dom0.”

Up until now, XenServer’s dom0 has been a 32-bit Linux operating system. That means all devices need to be mapped into just 4GB of system address space. That’s because 4GB is the most memory a 32-bit operating system can support. So a dom0 can only support a limited number of devices.

64-bit hypervisors
No limits: 64-bit hypervisors unleash greater scalability.

That’s a problem. In today’s servers, there are a lot of devices that need to get loaded into this 4GB address space. Storage controllers, network adaptors, management interfaces … and GPUs.

Most systems today support at least two of our GRID boards. Some support as many as eight. That eats up a lot of space and memory. So with the 32-bit XenServer, eight was the limit for many servers.

With the release of the 64-bit XenServer 6.5, Citrix has removed those memory and software limitations. A 64-bit dom0 OS allows us to load more stuff into a much larger I/O address space.

That lets you scale. And scalability – the ability to pile more users into a system – is one of the biggest questions in virtualization.

And now – thanks to XenServer 6.5 – we can scale to a higher number of users than ever before.


More on NVIDIA GRID and Citrix