Information Technology Reference
In-Depth Information
in line with the goals of cloud computing and would especially help scientific research that
needs supercomputing levels of power.
But current implementations of cloud shared memory allows for only a type of shared
cache that exists between virtual machines.
Virtual CPU
The virtual CPU (vCPU) is the logical counterpart of the physical CPU and is meant to
work as such. Each virtual machine is allocated one vCPU by default but may be assigned
more CPUs or cores depending on the guest operating system and the number of CPUs the
host contains. The virtual CPU does not directly correspond to a physical core or CPU but
is represented as time shares on the physical CPU's resource stack; it is not a one-to-one
assignment.
In a Type 2 hypervisor, this allows for more virtual machines to run on a single host
without affecting the host's performance adversely up to a saturation point where there are
too many timeshares that need to be accommodated, leading to various performance hits
for all VMs as well as the host. But in a Type 1 hypervisor, the bare metal one that takes
the place of an OS that runs the host machine, all of the CPU's processing power can be
allocated to virtual machine vCPUs because the host does not actually require CPU time on
its own because it is not running any sort of OS or software except the hypervisor itself.
As mentioned, the number of vCPUs that can be assigned to a virtual machine depends
on the type of guest OS that is being run and the underlying hypervisor. Different hypervi-
sors from different providers also have different limitations. This can also differ according
to the subscription level that you are using. Higher subscription levels usually allow for
more powerful virtual machines with more processors or computing power, more RAM,
and more storage, which of course comes at increased costs. This is quite expected of the
grid computing model that cloud computing uses; you pay for what you use, so if you use
more, you have to pay more.
Storage Virtualization
We have discussed in the previous sections how virtual disks work and some details regard-
ing virtualization. To recap, storage virtualization is creating a logical equivalent of storage
devices within a virtualized environment. So we are effectively creating a simulation of sepa-
rate storage devices, which in application are just single file formats located on the physical
storage medium of the host. But for the virtual machine using it and to the user, it all appears
in the same way that a physical storage medium would in a typical operating system.
Shared Storage
Cloud computing is a multitenant solution and is all about pooling resources and smart
reallocation to those that need it. CPU time, main memory (RAM), network bandwidth,
and of course storage are all gathered into a large share pool and allocated and reallocated,
processes we refer to as provisioning and deprovisioning. So in essence, all tenants or users
Search WWH ::
Custom Search