Using SAP with HP Virtualization and Partitioning

Computer Resources: Dedicated vs. Shared
When considering a given partitioning or virtualization technology for any software deployment, the
primary difference is whether the technology shares resources or has dedicated access to them.
When an application shares resources then, by its very definition, these resources are divided and
distributed to other applications for their use. The result is that the application performance is not
likely to execute as fast as it might if the resources are dedicated to that application.
By “dedicated resources” we don’t mean statically, dedicated hardware here. It rather stands for
providing the application with a certain guaranteed amount of resources at any time. That is, the
access to resources does not need to be constant, it can change over time.
By “shared resources” we mean having a pool of resources from which every application takes as
much as it wants, thus not guaranteeing exclusive resource access to any application.
Some heuristics for deciding whether a workload requires dedicated resources or is a good fit for
resource sharing are:
• Application testing, especially with respect to correctness, rarely requires high processing power
and is very well suited to resource sharing. The periodic nature and high levels of automation
present in testing also make it a good candidate for resource sharing.
• More formal quality assurance (QA) testing may include response time feasibility or be constructed
so that successful execution requires some level of sustained processing. In those scenarios, QA
processing is probably better suited to dedicated resources. Otherwise, QA processing may also
have short-term, periodic processing requirements that may lend it to efficiently sharing resources.
• Application development can benefit from the flexibility and rapid deployment inherent in resource
sharing technologies such as virtual machines. The inherent ability in virtual machines to quickly
deploy a system and provide OS configuration and patching capability to a development team
makes it an excellent choice.
• Low utilization production workloads are among the best candidates for hardware consolidation,
especially if these workloads do not have response-time requirements. Be aware that utilization is
often provided as an average and there may be periodic (e.g., end of week or end of quarter)
requirements for dedicated resources.
• High utilization workloads are usually better deployed on systems that can provide dedicated
access to computer resources. Rarely are they appropriate for virtualized processing solutions.
• Workloads that are performance sensitive, having stringent Service Level Agreements (SLAs) with
regard to completion time or response time are best deployed on platforms that provide dedicated
processing power.
• Computer capacity is often a driver for deciding on whether to use shared hardware or dedicated
hardware solutions. Workloads that require scores of mass storage units, network interfaces, or
processing cores are almost always better suited for dedicated hardware solutions.