Module 5
Resource
Management
Resource Management
In cloud computing, resource management refers to the process of
managing the usage of resources in a cloud environment. This
includes compute resources (like CPU and memory), storage
resources (like disk space), and network resources (like bandwidth).
Resource management is crucial for optimizing the performance and
cost of a cloud environment.
It involves monitoring resource usage, allocating resources, and
managing resource capacity.
Resource Management
Resource management is a core function of any man-made system, it
affects the three basic criteria for the evaluation of a system: performance,
functionality, and cost
An efficient resource management has a direct effect on performance and
cost and an indirect effect on the functionality of the system
Cloud resource management requires complex policies and decisions for
multi-objective optimization.
Effective resource management is extremely challenging due to the scale of
the cloud infrastructure and to the unpredictable interactions of the system
with a large population of users.
Resource Management
Resource management becomes even more complex when
resources are oversubscribed and users are uncooperative.
In addition to external factors, resource management is affected
by internal factors, such as heterogeneity of hardware and
software systems, the scale of the system, the failure rates of
different components, and other factors.
How resources are managed in cloud?
The strategies for resource management associated with the basic
cloud delivery models, IaaS, PaaS, SaaS, and DBaasS are different
In all cases, the cloud service providers are faced with large
fluctuating loads which challenge the claim of cloud elasticity.
In some cases, when a spike can be predicted, the resources can
be provisioned in advance, e.g., for web services subject to
seasonal spikes. For an unplanned spike the situation is slightly
more complicated.
.
How resources are managed in cloud?
Auto-scaling can be used for unplanned spikes of the workload
provided that:
(a) there is a pool of resources that can be released or allocated on
demand; and
(b) there is a monitoring system enabling the resource management
system to reallocate resources in real time.
Auto-scaling is supported by PaaS services, such as
Google AppEngine.
Resource Management
Centralized control cannot provide adequate solutions for management
policies when changes in the environment are frequent and unpredictable.
Distributed control poses its own challenges since it requires some form of
coordination between the entities in control.
Autonomic policies are of great interest due to the scale of the system and
the unpredictability of the load when the ration of peak to mean resource
demands can be very large.
POLICIES AND MECHANISMS FOR RESOURCE MANAGEMENT
A policy refers to the principles guiding decisions, while mechanisms represent
the means to implement policies.
Cloud resource management policies can be loosely grouped into five classes:
1. admission control,
2. capacity allocation,
3. load balancing,
4. energy optimization, and
5. QoS guarantees
POLICIES AND MECHANISMS FOR RESOURCE
MANAGEMENT
The explicit goal of an admission control policy is to prevent the system from accepting
workload in violation of high-level system policies. For example, a system may not accept
additional workload which would prevent it from completing work already in progress or
contracted.
Capacity allocation means to allocate resources for individual instances; an instance is an
activation of a service
Load balancing and energy optimization can be done locally – they both are correlated and
affect the cost for providing services.
Load balancing - evenly distribute the load among the set of servers.
- An example
POLICIES AND MECHANISMS FOR RESOURCE
MANAGEMENT
An important goal of cloud resource management is minimization of
the cost for providing cloud service and, in particular, minimization of
cloud energy consumption.
QoS is the aspect of resource management probably the most difficult
to address and, at the same time, possibly the most critical for the
future of cloud computing.
In common, the resource management strategies jointly
target performance and power consumption.
POLICIES AND MECHANISMS FOR RESOURCE
MANAGEMENT
The Dynamic Voltage and Frequency Scaling (DVFS) techniques such as Intel’s
SpeedStep and AMD’s PowerNow lower the voltage and the frequency to
decrease the power consumption
Intel’s SpeedStep will slow down the CPU to save electricity and also reduces
heat and noise of the PC. These techniques have migrated virtually to all
processors including the ones used for high performance servers
Processor performance decreases, but at a substantially lower rate than the
energy consumption, as a result of lower voltages and clock frequencies
- CPU with 1.8 GHz saves 18% of the energy required for maximum
performance
POLICIES AND MECHANISMS FOR RESOURCE
MANAGEMENT
Cloud resource allocation techniques must be based on
a systematic approach, rather than on ad hoc methods.
The four basic mechanisms for the implementation of resource
management policies are:
• Control theory
• Machine Learning
• Utility-based approaches
• Market-oriented mechanisms
POLICIES AND MECHANISMS FOR RESOURCE
MANAGEMENT
Control theory uses feedback mechanisms to guarantee system stability and
to predict transient behavior. Feedback can only be used to predict local,
rather than global behavior. Kalman filters have been used for unrealistically
simplified models.
Machine learning techniques do not need a performance model of the
system, a major advantage.
Utility-based approaches require a performance model and a mechanism to
correlate user-level performance with cost.
Market-oriented mechanisms do not require a model of the system, e.g.,
combinatorial auctions for bundles of resources
Referen
ces
Marinescu, Dan C. Cloud Computing: Theory and Practice.
Morgan Kaufmann, 2017.