Hardware Management - Deploying and Managing a Cloud Infrastructure - page 233

meant to serve applications through a network. The latter requires more bandwidth than
compute or memory resources. This has to be considered when selecting the service package
to use. You also need to consider whether the service provider can actually provide enough
network resources for high-networking-demand applications.
As for storage, we have what we call storage tiering, which we discuss later in this chapter.
Tiering is mostly related to backups and allows you to select the best storage resource for your
application in terms of performance, availability, and cost. But if you are just in need of stor-
age capacity, the continuing advancement in disk technology has ensured that this resource is
abundant and cheap. Even in the cheaper packages, storage capacity is ample. Besides, if you
need more than what is provided in your service package, you could always use more; you just
have to pay extra.
Entitlement/Quotas (Shares)
Entitlement and quotas come predefined in the service package. This is standard in order to
prevent overutilization from a single source, even if that source can afford it.
These limits are set as hard or soft limits. For example, if you allocated a static hard
drive of 100 GB for your VM instance, then you are stuck with that amount for that
specific drive. There is no way to increase unless you have a spare virtual drive. That is a
hard limit.
A soft limit, on the other hand, is one that is more flexible, though it is still considered
a limit. A good example is your bandwidth quota. Your network connection will not be cut
off if you reach it, but there are always consequences if you go beyond the soft limit. Usually
you would incur additional costs, which may be higher per gigabyte than the standard rate of
your package, or your bandwidth and network performance might be intentionally reduced.
The hard and soft limits of the service package should be highlighted so that the customer is
able to incorporate them into their planning, and automatic notifications should be set up so
that there will be a notification whenever you are nearing these limits.
Dynamic Resource Allocation
Dynamic resource allocation is there simply to make our lives easier. Instead of the admin-
istrator manually evaluating the resource utilization of virtual machines and then assign-
ing more resources where it's deemed necessary, specialized monitoring algorithms in
the hypervisor react to different situations and allocate or remove resources dynamically
based on the business logic or settings that are present for each scenario. These settings
could include the amount of traffic, process utilization, and even cost-to-performance
ratios. With dynamic allocation, precious time is saved by the administrator, not to men-
tion that a human operator may not be fast enough to manually cope with dynamic
workloads.
Resource Pooling
In a tenant's view, resource pools are hierarchical abstractions of virtual resources that can
be arranged according to importance or weight. Pools are hierarchical, so there can be parent
and child pools. The parent represents a higher order of virtual machines that have priority