Resource Manager is a critical component for managing a scalable
data center service that uses Microsoft® Provisioning
Framework (MPF). Without effective management of resources in
the data center, system resources are likely to be over- or
underutilized. Overutilization can lower quality of service and
violate a subscriber's service level agreement. Conversely,
underutilization can inflate operating costs.
Resource Manager establishes an optimal utilization model that
generates a notification whenever resources are over- or
under-allocated to resource consumers. You can use Resource Manager
Allocate consumers a percentage of total capacity available on
resources such as servers.
Allocate based on either virtual used or actual used capacity.
Virtual used capacity is the maximum amount that a consumer is
entitled to have, such as an amount that would be guaranteed in a
service-level agreement. Actual used capacity is a forecast of
average actual consumption, perhaps derived from consumption
statistics collected over time. For example, in a Microsoft®
Exchange installation, each consumer might be entitled to 30 MB of
disk space but in actuality, the average user might consume only 15
Allocate based on the rank of a resource. For example, suppose
you have a group of servers available for particular consumers. You
could rank the individual servers and allocate consumers to the
highest-ranked servers first.
Manage consumers and resources as groups (for operations), as
types (to simplify data entry and maintenance), and as instances
(when dealing with individuals).
Establish mappings between consumers and resources to simplify
selection of groups and instances for allocations, queries, and
other operations. Consumers and resources eligible for
consideration are selected for candidate sets.
Update allocations and shift consumers between resources as
demands and equipment availability change over time.
Resource Manager supports one allocation model, the block model,
in which each resource instance has a maximum capacity, and each
consumer instance is allocated a given percentage of that capacity.
During allocation, the block-model algorithm evaluates each
consumer and attempts to allocate it to the highest-ranked resource
instance. If that resource is already fully allocated, the
allocation function tries the next highest-ranked resource
instance, and so on, for all resource instances and all consumers.
Whenever multiple resources share the same rank, consumers are
allocated to the first resource with sufficient capacity.