Diagnosis and Performance Monitoring - Deploying and Managing a Cloud Infrastructure - page 149

Monitoring tools such as CopperEgg provide comprehensive analysis of resources and
applications, including traffic, and CPU and disk performance packaged in a neat and easy-
to-understand UI. Often, the best monitoring software is not really the most detailed but
software that can convey the message of system health easily to anyone reading it.
Since this is cloud computing we are talking about, you have the benefit of scalability.
Your cloud system has the ability to automatically scale with traffic and performance
requirements based on its usage, the extent of which is set by the administrator according
to certain thresholds. For example, the administrator can set the system to create a new
web server whenever the running one is experiencing 80 percent load traffic, giving ample
capacity and buffer time to allow the new server to fully configure. When the new virtual
server is ready to receive traffic, the system can automatically balance the load between all
the running servers so that no one server is being taxed more than the others. All of these
functions are accessible through the management interface being provided by your cloud
service provider.
Hypervisor Configuration Best Practices
The hypervisor is at the heart of every cloud computing platform or system; it sets up the
virtualization layer above it without the need for another OS, which is why it is also called
a bare-metal system. So it follows that for a system to run at peak performance, the hyper-
visor should be one of the healthiest, if not the healthiest, of all of the system components.
There is no single standard for how to configure all hypervisors because each one will have
its own unique features. The provider of the hypervisor will have best practices for setting
up and configuring specific hardware, so look to their documentation for specific needs.
But there are certain concepts that apply to allowing your virtualized system to per-
form better.
Memory Ballooning
Memory ballooning is a special feature implemented between the hypervisor and the
guest OS , and it allows the hypervisor to reclaim memory from the guest OS or virtual
machine. Yet it still allows the virtual machine to make intelligent decisions about which
memory pages should be paged out in case it does not have the required free memory
when the balloon driver is activated by the hypervisor.
In a virtual environment, there are three stacks of memory: the application virtual mem-
ory and guest “physical” memory as well as the actual physical memory of the host machine.
Figure 5.11 illustrates the hierarchy of memory. The hypervisor is between the host and the
guest OS. The guest OS or virtual machine behaves like an actual machine and is not aware
of the hypervisor, sort of like the humans not being aware that they are in a virtual world in
The Matrix . When the guest OS requires memory, the hypervisor provides it by mapping to
the physical host memory.