Implementing a Virtual Server Environment: Getting Started

The second thing to consider when organizing the partitions is how the resources might be shared.
To do this, you must further analyze each workload to understand its resource usage during peak
processing periods as well as any expected growth. Try to determine when the peak usage periods
occur and how predictable this behavior is. If possible, combine virtual machines or vPars so that
their peak processing periods don’t overlap. (Capacity Advisor is very good at helping with this
task.)
With Integrity VMs, the VM Host allocates the resources based on demand, so you can significantly
increase the utilization of a server when peak processing periods don’t overlap. With vPars, cores
within the same nPar can be dynamically moved from one vPar to another so that resources can be
applied where they are needed most. Memory can also be moved, but you would not want to do this
to address short term peaks. A good example of when moving memory resources might be useful is if
you had an application in one vPar that was busy during the day, and another application in a
separate vPar that ran at night.
If your vPars (or Integrity VMs) are combined in such a way that you have room for extra processor
cores in the nPar, then you can have a pool of iCAP cores with TiCAP usage rights. This is a very
good way to flex the size of these partitions to respond to short-term processing spikes. For vPars,
TiCAP cores can be activated for a specific vPar. For Integrity VMs, TiCAP resources are activated for
the VM Host, and then you can increase the entitlement for a specific VM. If you choose to use
TiCAP, HP recommends that you automate their activation and deactivation with WLM or gWLM.
If you have chosen nPars and need to flex their capacity from time to time, you can move the usage
rights of an iCAP core from one nPar to another by deactivating an iCAP core in one nPar and
activating an iCAP core in another. You can do this manually if the loading is very predictable and
doesn’t change too often; or you can add automation with WLM or gWLM.
Automation
To achieve the most effective benefits of resource sharing or TiCAP resources, you need to use either
WLM or gWLM. If you aren’t currently using WLM, consider gWLM for its ease of use and its ability
to define resource sharing policies for multiple systems. gWLM allows you to combine multiple
workloads with differing demand patterns on a single server to make use of idle capacity, while
guaranteeing that each workload receives its specified entitlement of resources. For multiple
workloads that span partitions or servers, gWLM can move the resources to where they are needed
most as defined by the policies you set up. TiCAP resources may also be enabled when required to
meet the workload’s demand and always disabled when not required which saves money. An
operator or administrator would simply not be able to respond quick enough to shift resources around
to meet short term demand. Additionally, TiCAP resources can be enabled or disabled at the
workload level. The gWLM tool is not as complex as it might sound. For more information about
using gWLM, see the link to the HP Integrity Essentials Global Workload Manager User’s Guide
Version A.03.00.00 at the end of this paper.
High Availability and Disaster Tolerance
It is beyond the scope of this paper to discuss HP Serviceguard high-availability and disaster-tolerant
solutions in detail. However, you might want to consider adding high availability or disaster
tolerance to your virtualized environment as you are planning your virtualized infrastructure. HP
Serviceguard is well integrated with the virtualization products. For example, in a failover situation,
gWLM can have a policy to adjust the workload entitlements on the failover server to provide
appropriate levels of resource usage to the existing workloads, along with the workload from the
failed system. Additionally, iCAP, TiCAP, or GiCAP resources can be activated.
18