KVM superior than ESX and Xen a place: not only was a very good performance, it can help solve the issue of maintenance of device drivers. There is some truth.
I / O performance is critical to a hypervisor. At the same time, I / O is also a big burden to maintain, because there needs to be supported by a large number of hardware devices, a large number of I / O protocol, high availability, as well as the management of these devices.
VMware choice performance, but the I / O protocol stack on which the hypervisor. Unfortunately, VMware kernel is proprietary, it means that VMware had to the development and maintenance of the entire protocol stack. That would mean that the development would slow down the speed of your hardware may have to wait some time before we can get the support of VMware.
Xen can be chosen to maintain this path, it will all of the I / O operation on the Linux guest inside, that is, the so-called domain-0 inside. Reuse Linux do I / O, Xen's defenders do not have to rewrite the entire I / O protocol stack has been. Unfortunately, however, so the expense of performance: a disruption of Xen have been necessary for the activation in order to switch to domain 0, and everything had been an additional layer of mapping.
Is not to say that Xen has completely solved the problem maintainability: Xen domain 0 kernel is still the oldest Linux 2.6.18 (even though 2.6.25 has also been available. Now Xen has been adopted in the domain 0 pv_ops to solve this problem)
KVM is how to deal with it then?
Like VMware, like, I / O is to be put to the implementation of the context of hypervisor, so performance will not have any damage. Like Xen, like, KVM reuse the entire Linux I / O protocol stack, so users KVM naturally get the latest drivers and I / O protocol stack improvements.
Subscribe to:
Post Comments (Atom)
No comments:
Post a Comment