HP Virtual Server Environment for HP-UX—Automated, real-
time server resource allocation
Executive summary ........................................................................................................................... 2
HP Virtual Server Environment—Server resource optimization ................................................................ 3
HP Virtual Server Environment in action .............................................................................................. 7
Workload management: HP-UX Workload Manager within a single operating system image ................ 7
HP-UX Workload Manager and psets within a single operating system image ..................................... 9
HP-UX Workload Manager and partitioning: HP Virtual Partitions .................................................... 11
HP-UX Workload Manager and utility pricing solutions: HP Instant Capacity ..................................... 12
HP-UX Workload Manager and high availability: HP Serviceguard .................................................. 15
HP-UX Workload Manager and disaster-tolerant solutions: HP geographically dispersed
cluster solutions .......................................................................................................................... 16
Virtual Server Environment Quick Start Solution.............................................................................. 18
HP Virtual Server Environment—Real value for your business ............................................................... 19
Improved RoIT ............................................................................................................................ 19
Enhanced business agility............................................................................................................ 19
Improved quality of service .......................................................................................................... 19
Reduced risk .............................................................................................................................. 19
Future directions of the HP Virtual Server Environment ........................................................................ 20
HP Global Workload Manager.................................................................................................... 20
Why HP? ...................................................................................................................................... 20
For more information ...................................................................................................................... 21
Businesses are increasingly focused on reducing costs and improving their return on information
technology investment (RoIT). As they strive for agility, the ability for computing resources to be
managed in a way that responds to and reflects rapidly changing business drivers has become a
priority for IT departments. The HP vision for the Adaptive Enterprise is one in which the ability of IT to
supply services is consistently aligned with the demand from the business.
Figure 1. The HP framework for the Adaptive Enterprise
As business cycles vary and demand ebbs and flows, the underlying infrastructure must respond
accordingly. Just as increased exertion results in expansion of the lungs to absorb a greater volume of
air to meet the body’s increased oxygen requirements, the underlying technology infrastructure of a
business must scale to meet the demands made upon it. After the body’s period of exertion is over,
the air intake is reduced and the lungs contract as they continue to serve the needs of the body until
the demand is increased again. In a similar way, the ability to expand and contract to meet demand
without additional resources being required is a fundamental, underlying principle of an Adaptive
To support its Adaptive Enterprise vision, HP is developing an underlying infrastructure that has been
specifically created to reduce cost and complexity while increasing IT responsiveness. A key pillar of
the HP Adaptive Enterprise is virtualization.
Virtualization is an approach to IT that pools and shares resources so utilization is optimized and
supply automatically meets demand.
One key solution for enabling an infrastructure of this nature is the HP Virtual Server Environment
(VSE), which offers an increased RoIT through the optimization of server resource utilization in real
time according to business priorities. In addition to increasing server utilization, the HP VSE also
enables the rapid deployment of computing resources, improves availability, and significantly reduces
This white paper is designed to outline the capabilities and benefits of the HP Virtual Server
Environment and explain how and why it is an important building block for an Adaptive Enterprise. It
is designed to cater to the needs of senior IT managers and data center staff who are in the process
of addressing cost, complexity, resource utilization, and RoIT issues within their enterprises.
HP Virtual Server Environment—Server resource optimization
The HP VSE is composed of seamlessly integrated server virtualization products built around the only
goal-based policy engine available for UNIX®. The VSE assesses whether the performance of a single
server—or cluster of servers—is meeting predefined service-level objectives (SLOs), and it then
dynamically adjusts server resources by managing free pools of resources and/or moving resources
out of underutilized virtual servers. The SLOs assessed by VSE fall into three categories1:
1. Resource utilization: An application or application group can be assigned a minimum and
maximum percentage or share of CPU cycles, and it is also assigned a priority based on business
needs. This percentage or share—also known as CPU entitlement—can be dynamically and
automatically adjusted within the predefined range according to application priority, based on the
utilization of the current entitlement.
2. Performance: A direct measurement of the performance of the workload—for example, a response
time of less than 1 second or a throughput of 5,000 transactions per minute. Degradation in
performance will result in the VSE dynamically allocating additional resources to maintain the SLO.
3. Application load: A measurement of the load placed on an application—that is, the number of
users or processes or the queue length. As load increases, the VSE will automatically reassign
resources to handle the additional workload.
Management of SLOs in the fashion outlined above improves return on IT. The VSE ensures higher
levels of server utilization and increased agility because server resources are automatically adjusted
to meet changing business priorities and application usage levels.
Traditionally, servers have been acquired and deployed on an application basis. For example, an
SAP R/3 project would require dedicated servers for development, test, production, and disaster-
recovery environments. An Oracle® customer relationship management (CRM) project would most
likely require the acquisition and deployment of new, dedicated servers for each of those
environments at additional expense. The planning, budget, and procurement cycles—as well as the
capital expenditure incurred—would have a direct impact on the time to market and return on
investment of the project. Delays in implementation might even erase any competitive advantage that
may have existed when the project began. In addition, research shows that the majority of servers
would be underutilized most of the time.
Now just imagine an environment where:
• All of those SAP R/3 and CRM server resources could be placed in a dynamic resource pool, with
each request for resources being evaluated and allocated according to business priorities
• Underutilized server resources could be reassigned to alternate application instances without
compromising security or availability
• An unforeseen transaction peak would result in additional resources being added to the pool and
assigned to meet the need as opposed to requests for service being denied
• Disaster recovery no longer requires a set of underutilized, standby systems
• New applications can be deployed with defined SLOs without having to acquire additional
hardware and without initiating a new procurement cycle
For examples, see the “HP Virtual Server Environment in action” section on page 6.
The HP Virtual Server Environment enables all of those benefits through the virtualization—or
abstraction—of server resources so that they are no longer seen as independent, dedicated boxes
with processors and memory, but rather as a pool of dynamic computing resources available to your
entire organization. As business requirements fluctuate, resource requirements are analyzed and
orchestrated across multiple systems using dynamic provisioning, allocation, and management
technologies in accordance with agreed service-level objectives. The VSE automatically grows and
shrinks based on the needs of hosted applications. Resource usage is thus optimized across the
business while IT spending is contained.
Figure 2. The HP Virtual Server Environment
VSE encompasses many fully integrated, complementary components that enhance the functionality
and flexibility of your server environment. These include:
• An intelligent policy engine, HP-UX Workload Manager (HP-UX WLM), which performs real-time
assessment of resource usage and advises and acts based on SLOs and business priorities to adjust
the supply of resources according to demand
• Resource management technologies—HP Process Resource Manager (PRM) and Processor sets
(psets)—that enable granular processor, memory, and I/O allocation to be managed between
multiple applications competing for resources within a single system or virtual partition, thus
eliminating resource contention
• Flexible Virtual Partitions (vPars) that combine software fault isolation and security with the ability to
share resources with other vPars within a system, thus increasing utilization and reducing cost
(Instant ignition [Ignite-UX] of operating systems within new vPars offers quick provisioning of new
environments for development, test, QA, or disaster recovery as required.)
• Utility Pricing Solutions from HP—Instant Capacity , Temporary Instant Capacity, and Pay-per-use
(PPU)—that enable dormant processors and memory to be dynamically activated and paid for as
• Clustering solutions (HP Serviceguard, HP Extended Campus Clusters, HP Metrocluster, and HP
Continentalclusters) that enable the implementation of cost-effective failover and disaster-tolerant
environments without the acquisition of redundant hardware
• HP geographically dispersed cluster solutions such as HP Serviceguard Extension for RAC
(SGeRAC) that extend clustering capability from a single data center solution to a transparent
“virtual application environment” between two data centers up to 100 km apart in an active-active
environment with no redundant hardware
The core functionality of the VSE is delivered by the intelligent policy engine, HP-UX Workload
Manager, which continuously assesses whether a server environment is meeting its SLOs and advises
administrators or adjusts resources automatically when SLOs are not met. This base functionality is
enhanced by the ability of VSE to orchestrate the real-time allocation of all virtual server resources—
virtual partitions, utility pricing solutions, and cluster solutions—assigned to executing applications,
users, and processes.
Seamless integration between these components provides continuous availability and performance of
mission-critical applications such as enterprise resource planning (ERP), supply chain management
(SCM), and CRM.
Furthermore, additional application-transparent toolkits allow for closer integration between HP-UX
WLM and specific applications such as BEA WebLogic and Oracle. These toolkits ensure that
application-specific resource demands, such as BEA queue length and number of free threads or
number of active Oracle users, are assessed by HP-UX WLM; that the system administrator is advised
of the situation and of potential remedies; and that HP-UX WLM in turn acts on the relevant VSE
component so as to meet the SLO.
To get a virtualized BEA WebLogic Server or Oracle database environment deployed quickly, HP has
also developed the Virtual Server Environment Quick Start Solution. It is a pre-tested, qualified, and
supported software package consisting of best-of-breed infrastructure components designed to
substantially reduce the design, testing, installation, and integration processes for these two
This white paper focuses on the key components of the VSE—HP-UX Workload Manager, HP Virtual
Partitions, HP Utility Pricing Solutions, and HP clustering solutions in high-availability and disaster-
tolerant environments—as well as the integration of the components. Take a look in greater detail at
how the key components of VSE work together to provide a unique value proposition for you and
HP Virtual Server Environment in action
The VSE components work together in such a way that the technological strength of each
complements that of the others. Figure 3 illustrates how the modular building blocks of the VSE
expand the scope of dynamic resource optimization, thereby increasing business agility.
Figure 3. Building blocks of the Virtual Server Environment
As the base capability, workload management enables application stacking within an operating
system image with or without utilizing processor allocation (psets). Partitioning enhances this
capability by enabling application stacking across operating system images within the same hard
partition or server. Utility pricing solutions provide flexible capacity management and pricing schemes
that offer you the ability to instantly ignite additional CPU capacity to meet SLOs or to pay HP on the
basis of only the actual resources used. Integration with high-availability clusters and multiple data
center disaster-tolerant solutions maximizes resource optimization and reduces risk.
The following section looks first at the intelligence behind the HP-UX WLM component of VSE and then
explains how each building block within the VSE is orchestrated to provide the dynamic virtual server
environment that your organization needs to improve operational efficiencies and increase return on
Workload management: HP-UX Workload Manager within a single operating system image
At the heart of the VSE is an enhanced version of HP-UX WLM, the only goal-based, intelligent policy
engine for UNIX available today. It provides real-time monitoring of application SLOs to ensure that
they are maintained through the dynamic re-allocation of processor resources as required. This core
functionality enables otherwise underutilized resources to be re-allocated from idle applications to
those that require them to maintain performance and response times.
Figure 4. Workload management in a VSE: HP-UX WLM
To begin with, the system administrator must decide on the SLO for each application. Within the
context of HP-UX WLM, SLOs can be entitlement-based or goal-based. With an entitlement-based
SLO, either HP PRM—an HP-UX resource management component of VSE—or HP-UX WLM tries to
grant a specified percentage of available computing resources to the associated application
workload, such as 40 to 80% of processor capacity for PRM Group A, as depicted in Figure 4.
With goal-based SLOs, HP-UX WLM actively changes the associated workload’s PRM resource
allocation to best meet the SLO. For example, if the SLO of a priority application cannot be
maintained using 40% of processor resources, HP-UX WLM will increase the processor allocation until
such time as the SLO is honored. These SLOs are based on one of two goal types:
1. Metric goals: Goals based on a metric, such as performing n transactions per minute, or a
response time of less than x seconds, as pictured in Figure 4.
2. Usage goals: Goals based on how efficiently workloads use their CPU allocations—if a workload
is not using its full allocation, its allocation is decreased; if a workload is using a high percentage
of its allocation, the allocation is increased.
With HP-UX WLM, the system administrator creates one or more SLOs for defined workloads
composed of executing applications. Each SLO is assigned a priority in addition to metric or resource
usage goals. As the applications execute, HP-UX WLM compares the performance metrics or usage
against the defined goals and automatically adjusts the CPU entitlements (the amounts of CPU that are
available to the workloads) to achieve each goal, maintaining the relative priorities at all times.
HP-UX WLM enables you to run the system at close to 100% utilization and still count on the
performance of your mission-critical applications. This is accomplished by placing one or more critical
applications that have performance requirements on a system along with multiple lower-priority
workloads that have little or no performance requirements. HP-UX WLM will allocate resources to
ensure that the critical applications get the resources they need to meet their performance
requirements, and then it will allocate spare CPU cycles to the lower-priority workloads. HP-UX WLM
maximizes the use of CPU resources while ensuring that the most critical applications perform
according to the defined SLOs. The investment in historically single-application, low-utilization systems
can be recouped by utilizing those systems to run multiple workloads, all managed by HP-UX WLM.
Resources required by mission-critical applications that peak at different times can be shared instead
The implementation of HP-UX WLM enables the following:
• Multiple applications can be consolidated within a single operating system image to utilize excess
capacity while ensuring that the highest priority applications still have access to the resources they
need during peak times. This reduces the need for a high server asset count and eases the
management burden associated with large server farms.
• System resources can be dynamically re-allocated in response to changing priorities, conditions that
change over time, resource demand, and application performance.
HP-UX Workload Manager and psets within a single operating system image
One example of HP-UX WLM in a VSE is its success in managing resource utilization in BEA
WebLogic environments executing on large HP servers. Many data centers want to simplify
deployment, minimize management overhead, and reduce licensing and support costs by running
multiple BEA WebLogic instances on larger systems that are better able to handle fluctuating
Because of the Java™ Virtual Machine, BEA WebLogic’s architecture scales best in environments of up
to four processors. The benefit of allocating BEA WebLogic instances to systems that scale beyond
four processors—such as HP Superdome—is limited because of the sharing of CPU cycles and the
underutilization of available resources.
BEA WebLogic requires dedicated CPUs, but installation across multiple smaller systems increases
deployment and management complexity, negating the benefits available through consolidation on
larger, scalable systems. Moreover, dedicated CPUs without the flexibility of re-allocation according
to need causes underutilization and increased cost.
Psets are an HP-UX feature that enables the customer to group processors together within a single
operating system instance. Representing a group of dedicated processors within a system, psets
provide a mechanism for dynamic CPU resource management. They enable users to partition large
systems into more than one virtual machine in terms of the processor resources only. A BEA WebLogic
instance assigned to a pset will only utilize processors within that assigned pset, ensuring CPU
resource isolation for applications and users. Therefore, multiple BEA WebLogic instances can run in
a single operating system on a single HP Superdome, each with a dedicated set of CPUs. This can,
however, create underutilization of the system based on fluctuating demand.
WLM solves this problem by being able to allocate and migrate CPUs between BEA WebLogic
instances based on priority. Figure 5 depicts an HP Superdome partition with eight processors
running two BEA WebLogic instances as well as other workloads. Each BEA WebLogic instance has
been assigned to a pset that includes a minimum of one dedicated CPU (CPU 6 and CPU 7). Both
pset 1 and pset 2 have been assigned an additional CPU (CPU 5 and CPU 4, respectively) from the
default pset pool, which holds all CPUs not allocated to specific psets.
Figure 5. HP-UX WLM managing multiple BEA WebLogic workloads
Assessing the information obtained using the WLM BEA WebLogic Server Toolkit, HP-UX WLM will
note that WebLogic 1 has a queue containing 37 outstanding requests. To handle this workload
based on its assigned priorities, HP-UX WLM will migrate CPU 3 from the default pset to pset 1 to
manage the queued requests, as illustrated in Figure 6. Immediately, the number of outstanding
requests in the queue is reduced (from 37 to two) and free threads are made available (from zero to
HP-UX WLM will also be informed that WebLogic 2 no longer requires CPU 4 because it has no
queued requests and the number of free threads available has increased (from four to 11). CPU 4
will, as a result, be migrated back to the default pset pool for use by other workloads.
Figure 6. HP-UX WLM de-allocates processors when not required
HP-UX Workload Manager and partitioning: HP Virtual Partitions
This orchestration of resources by HP-UX Workload Manager is significantly enhanced when
implemented in an environment incorporating HP Virtual Partitions (vPars) in a VSE.
HP vPars are a unique feature that enables multiple operating system instances to execute within a
single system or hard partition (an HP nPartition incorporating electrical isolation). Each vPar can own
a specified amount of memory, one or more ranges of physical memory, a specified pool of CPUs, or
a set of one or more I/O cards within the server. HP vPars can be dynamically created and modified
using software commands. Each application is then able to run in an environment where performance
is maximized because its own unique operating system configuration requirements are met.
Security is maintained because each operating system instance is isolated from all others. In the event
of an operating system panic within one vPar, applications executing in alternate vPars are not
affected. Moreover, greater flexibility is achieved because each server is able to run multiple versions
and patch levels of the operating system. This means that applications can exploit new operating
system features without having to wait for all applications on the system to be ported to the same
operating system version. In addition, vPars can be created on the fly for testing new versions of the
operating system or application without having to replicate the deployment environment.
HP-UX WLM can resize virtual partitions by dynamically re-allocating CPUs from one virtual partition
to another. The policy-based rules of HP-UX WLM and HP PRM determine the prioritization of each
application within the partition, enabling the allocation of CPU, memory, I/O, and capacity
according to specific processing requirements and assigned priority. In the event of an application
within a clustered partition failing over, resources can automatically be re-allocated, minimizing the
user impact resulting from resource contention.
CPUs that are not allocated to any specific vPar within the system are known as “floating” CPUs and
are assigned by HP-UX WLM according to need. Figure 7 represents a single system with six
processors and four vPars, each possibly executing a different patch level of HP-UX 11i along with
different kernel tuning parameters, according to the application requirements. Each vPar is assigned a
dedicated CPU, while vPars A and B each have one additional “floating” CPU currently assigned to
them—CPU 4 and CPU 5, respectively. Specific SLOs have been set for applications A and B,
residing in vPars A and B, respectively. Application A has an SLO that states its online response time
must be less than one second. Application B’s SLO is not as stringent, stating that response times of
less than five seconds are acceptable.
Figure 7. Partitioning in VSE: HP-UX WLM managing HP vPar resources
If the response time of Application A exceeds one second, HP-UX WLM will assess the situation.
Because Application A has a higher priority than Application B, HP-UX WLM will dynamically de-
allocate “floating” CPU 5 from vPar B and re-allocate it to vPar A so the SLO will be met. In doing so,
HP-UX WLM has the ability to flex, or adjust, the resource assignment based on business priorities.
This integration between HP-UX WLM and vPars in a VSE enables development, test, and QA or pre-
staging environments to be deployed on the same systems as production workloads. Should the
production workload exceed its configured resource threshold, HP-UX WLM will automatically
re-allocate resources assigned to the lower priority, non-production partitions, thus leveraging the full
value of the investment.
HP-UX Workload Manager and utility pricing solutions: HP Instant Capacity
Flexible capacity and cost management is provided by HP Utility Pricing Solutions to minimize risk
while enhancing agility for your business. These solutions are designed to enable you to cater to
demand fluctuations without having to make a significant up-front investment in processing resources
that may only be utilized at a later date. In a VSE, when a requirement is recognized by HP-UX WLM,
dormant resources can be activated and utilized to ensure that business processes are managed and
maintained according to the agreed-upon service-level agreements (SLAs) from which the SLOs are
HP Utility Pricing Solutions fall into two categories: metered capacity, such as PPU, or Instant
Capacity. In the case of metered capacity, all of the CPUs in a system are activated and monitored on
a daily basis. A monthly average is derived from the daily CPU utilization metrics, and that average
serves as the basis for your monthly payment. Integration with HP-UX WLM in a PPU environment
ensures the optimization of the resource utilization to the minimum required to meet your SLOs,
optimizing your monthly costs.
HP Instant Capacity solutions provide additional flexibility based on the increase of performance
requirements over time, whereas transaction volatility is catered to by metered capacity. HP Instant
Capacity comes in three variations:
• Standard Instant Capacity: With the standard Instant Capacity program, you can deploy a server
fully loaded with CPUs but pay only for the processors you plan to use on day one. When needed,
the dormant processors can be activated instantly with a simple UNIX command. This is a real
benefit in the case of new projects where systems are sized based on a future, anticipated
transaction volume. In addition, if a CPU failure is detected, high availability is provided by the
automatic activation of an Instant Capacity processor and the shutdown of the defective processor.
Because the total number of active processors does not change in this case, there is no cost
associated with this transaction.
• Temporary Instant Capacity: In the case of an environment where extra processing power is only
required on rare occasions (for example, peak holiday periods or year-end processing), Temporary
Instant Capacity allows processors to be activated and then deactivated. One Temporary Instant
Capacity 30-day license is applied to each system so that any number of Instant Capacity CPUs
installed on the system can be activated and deactivated for an accumulated time of 30 days within
a three-year period. Should additional capacity over and above the 30 days be required, more
Temporary Instant Capacity licenses can be purchased.
• Cell board Instant Capacity: Cell board Instant Capacity for the HP rp7410 and rp8400 servers
extends the value of Instant Capacity to include complete cell boards loaded with CPUs and
memory. Designed for systems with hard partitions, cell board Instant Capacity can be transferred
between partitions based on demand and then activated following a reboot. Each partition requires
at least one active cell board, with the cell board, one processor, and one memory module
activated. Additional processors and memory modules are activated by the administrator as
The availability of HP Instant Capacity solutions within the VSE provides two key benefits:
• Available resources are optimized so activation of additional Instant Capacity resources is delayed
as long as possible, saving you money.
• If Instant Capacity resources are required to meet SLOs, activation is almost instantaneous, so
procurement cycles and downtime for upgrades are eliminated and no revenue-generating
opportunities are lost.
As with the other key components of the VSE, Instant Capacity is seamlessly integrated to provide you
with the efficiency and return on investment that you have come to expect from HP.
As discussed previously, HP-UX WLM monitors the SLOs of the individual workloads to determine how
to allocate resources in order to best optimize the entire system resource pool based on business
priorities. In cases where Instant Capacity resources are available but inactive, HP-UX WLM monitors
these resources and can either activate them dynamically to meet its SLOs or simply inform a system
administrator that Instant Capacity resources could be activated. HP has found that many customers
prefer to have their system administrator analyze the situation and choose whether to activate the
resource and incur the cost rather than receive an unexpected invoice. Because the time required for
the administrator to make an informed decision based on the cost of the resource compared to the
impact of the SLO violation is often minimal, this process has also been implemented in HP-UX WLM2.
When an SLO is missed on a system where all resources are fully utilized and Instant Capacity
resources are available, the system administrator is notified both of the SLO violation and of the
availability of the Instant Capacity resource that can be activated to resolve the situation. If the Instant
Capacity resource is activated, it will become available to HP-UX WLM and will be utilized to meet
the SLO. In environments where virtual partitions are installed, administrator-activated Instant Capacity
CPUs become “floating” CPUs, which can be allocated by HP-UX WLM to vPars according to demand
Figure 8. HP-UX WLM orchestrating Instant Capacity CPUs between nPars
A new release of HP-UX—scheduled for late 2003—will allow integration between HP-UX WLM,
nPars, and Instant Capacity resources. This integration allows the activation and deactivation of
Instant Capacity CPUs to be synchronized across nPars so no additional Instant Capacity or software
licensing costs are incurred.
Figure 8 depicts the process that HP-UX WLM follows when attempting to allocate available Instant
Capacity resources in order to meet SLOs. Each nPar includes two active, dedicated CPUs and two
Instant Capacity CPUs. Both Instant Capacity CPUs in nPar A are active, while both Instant Capacity
CPUs in nPar B are inactive. If the workload executing in nPar B violates its SLO, HP-UX WLM will
assess the possibility of transferring the Instant Capacity license from one active Instant Capacity CPU
in nPar A to an inactive Instant Capacity CPU in nPar B. If there are unused resources in nPar A or if
nPar B has a higher priority—as is the case in this example—HP-UX WLM will dynamically deactivate
an Instant Capacity CPU in nPar A and then activate one of the inactive Instant Capacity CPUs in
Integration of HP-UX WLM with HP Instant Capacity 6 requires that the administrator be notified. The automatic
activation of Instant Capacity CPUs is not available in this configuration.
nPar B. This process ensures that the total number of active Instant Capacity CPUs within the system
remains constant. This means that the customer is not charged for the activated Instant Capacity CPU
or for additional per-processor software licenses in situations where both nPars are executing the
same database or application. If the SLOs for both nPar A and nPar B cannot be met by the existing
active CPUs, HP-UX WLM will notify the administrator, who can choose to activate one or more
additional Instant Capacity CPUs in nPar B.
HP-UX WLM works not only across the virtual partitions within an HP 9000 server, but also across
hard partitions on both HP 9000 and HP Integrity servers. While CPU resources cannot be moved
between hard partitions at this time, HP-UX WLM does integrate with nPars and Instant Capacity to
offer some dynamic capabilities. If a system has inactive Instant Capacity CPUs, WLM can turn off a
CPU in one hard partition and then turn on an available CPU in another hard partition. Because the
total number of active CPUs in the system does not change, customers are not charged for activating
the Instant Capacity CPUs. HP-UX WLM can trigger license transfer through either set policies (such as
time of day) or application SLOs.
HP-UX WLM is thus able to orchestrate workloads executing on a single system or in a hard or virtual
partition, workloads in environments where Instant Capacity resources are available, and, in the
future, workloads in multiple hard partitions with available inactive Instant Capacity CPUs. This
ensures that SLOs are met in the event of resource contention. While resolving resource conflicts
according to business priorities is the main focus of HP-UX WLM, its strength lies in its ability to
accomplish this even in times of resource outage.
HP-UX Workload Manager and high availability: HP Serviceguard
HP Serviceguard is the foundation software designed to protect mission-critical applications on HP-UX
and Linux operating systems from a wide variety of hardware and software failures. Integration of
Serviceguard within the VSE ensures application availability while maintaining SLOs through dynamic
resource re-allocation—even between multiple, distributed data centers.
With Serviceguard, up to 16 servers and server partitions (hard or virtual) are organized in an
enterprise cluster that delivers highly available application services to LAN-attached clients.
HP Serviceguard monitors the health of each server and partition and rapidly responds to failures in a
way that minimizes or eliminates application downtime. Within Serviceguard, applications are
“packaged” and then transferred to alternate servers or partitions within the cluster in the event of an
operating system or system failure. In addition, “packages” can be rolled over to alternate servers or
partitions to maintain services while operating system or hardware upgrades take place on the
original system. In each situation, integration with HP-UX WLM ensures that the application priorities
and corresponding performance SLOs are honored.
In the event of either an unplanned failover or a planned rollover for system maintenance,
HP Serviceguard will fail over the affected packages to an alternate system within the cluster. Figure 9
depicts two systems within a data center cluster environment. The two systems could be either two
independent servers or two partitions (hard or virtual) on the same server.
In the event that Server 1 becomes unavailable for service because of an interruption (a system error,
a major unplanned event, or an operator intervention, such as a rolling upgrade), Package A will be
restarted on Server 2 as defined in the HP Serviceguard configuration file. HP-UX WLM will
automatically optimize the system resources by re-allocating and balancing them according to
business priorities. Because of its higher priority as defined within the HP-UX WLM configuration file
that is shared across the cluster, Package A will be allocated additional resources at the expense of
Packages B and C to minimize the impact on its SLO. Because HP-UX WLM supports inactive SLOs,
packages are always guaranteed resources on the receiving node. At the same time, because all
inactive SLOs receive a combined maximum allocation of 1% CPU and 1% memory, available
resources can be fully utilized at times when application packages have not been transferred.
Figure 9. High availability: HP-UX WLM integration with HP Serviceguard
Furthermore, if Instant Capacity resources are available on the receiving system (Server 2), HP-UX
WLM will notify the system administrator that these can be activated to meet the SLOs of all
packages. When activated, HP-UX WLM will use the Instant Capacity resources to restore the
performance of the affected applications. This capability provides a compelling business case
because the RoIT of a high-availability solution that includes VSE is significantly better than a
traditional high-availability environment involving oversized servers that cater to additional workloads
and are often underutilized.
HP-UX Workload Manager and disaster-tolerant solutions: HP geographically dispersed cluster
Properly preparing for disasters can be an expensive undertaking because multiple data centers are
involved in creating a disaster-tolerant environment. Subscription to a third-party disaster recovery
(DR) site raises concerns about security, time, and resource sharing, as well as the very real threat of
not having priority if multiple companies require use of the facility in the event of a national disaster.
For companies with multiple data centers and stringent time and data point recovery requirements,
catering for disasters in-house is an alternative solution—but it can also be an expensive undertaking.
In some instances, redundant hardware is installed at enormous cost in a remote data center, and
data is mirrored in case of a disaster. In other cases, each data center is populated with both
production and non-production systems (for example, development, QA, or tests). In the event of a
disaster at one site, all non-production systems at the alternate sites are reconfigured to run the
These traditional disaster-tolerant or business continuity models have been replaced with solutions that
adapt to meet the changing needs of both the business and the computing environment. The HP
virtualization of servers across different data centers means that all available resources are fully
utilized, both in an everyday environment and in the event of a disaster. There is no longer a need for
idle equipment to stand by in the event of a disaster.
HP geographically dispersed cluster solutions (GDCS) are built on an HP Serviceguard foundation
and deliver the same capabilities as Serviceguard, only over a far greater distance. The HP GDCS
suite consists of Extended Campus Cluster, Metrocluster, and Continentalclusters. HP Extended
Campus Cluster enables automatic and bidirectional failover of mirrored mission-critical data and
applications between data centers located up to 100 km apart. HP Metrocluster enhances those
capabilities by adding asynchronous and synchronous data replication between disk arrays over the
same distance. HP Continentalclusters is the most flexible of the GDC solutions, supporting up to 32
nodes over an unlimited distance and allowing any type of data replication. In all cases, resource
utilization and performance are optimized with the addition of a VSE.
Figure 10. Disaster-tolerant solutions: HP-UX WLM reconfigures the landscape following a disaster
Figure 10 illustrates the integration of HP-UX WLM and HP Extended Campus Cluster, Metrocluster,
and Continentalclusters solutions. Regardless of the distance between data centers, in the event of a
disaster, “packages” can be rolled over to servers in alternate data centers and workloads prioritized
to honor SLOs.
If your environment requires disaster-tolerant protection for Oracle9i RAC while still needing the
adaptive infrastructure to maintain SLOs in the event of a failure, Serviceguard Extension for RAC
(SGeRAC) is also part of the Serviceguard portfolio of high-availability products for HP-UX.
Enhancements to this product extend its capability from a single data center solution to providing a
transparent “virtual application environment” between two data centers up to 100 km apart. In all
cases, resource utilization and performance are optimized with the addition of a VSE.
HP disaster-tolerant solutions are recognized as being among the best in the industry. Implementing
HP VSE significantly reduces the high costs associated with traditional disaster-tolerant or business
continuity environments requiring hot standby resources. The integration of HP GDCS, HP-UX WLM,
and Instant Capacity allows all available servers to be fully utilized at all times. In the event of a
disaster, mission-critical applications will be re-allocated resources from lower-priority workloads and
additional capacity requirements will be fulfilled by HP Utility Pricing Solutions, allowing the system to
flex based on demand.
Virtual Server Environment Quick Start Solution
To get a virtualized BEA WebLogic Server or Oracle database environment deployed quickly, HP has
developed the Virtual Server Environment Quick Start Solution. It is a pre-tested, qualified, and
supported software package consisting of best-of-breed infrastructure components. The Virtual Server
Environment Quick Start Solution is designed to substantially reduce the design, testing, installation,
and integration processes for these applications. The solution contains all of the required components
and patches. The VSE Quick Start Solution includes components and tools for:
• Virtual Server Environment techniques (as described previously)
• System management
• Web serving
• Cluster management
• Directory serving
• HP-UX operating environment management
In addition to BEA WebLogic Server and Oracle9i Database, the Virtual Server Environment Quick
Start Solution includes best-of-breed software to take advantage of HP best practices in the areas of
consolidation (application stacking), management, and virtualization.
Processor set tools contained in the VSE Quick Start Solution allow Java Virtual Machines (JVMs) to
function optimally on SMP machines by restricting each JVM to a subset of four or fewer processors
where they perform best. Additionally, the same technique can be used to stack Oracle9i Database.
The result is that you can consolidate (“stack”) multiple WebLogic Servers or Oracle instances onto
one operating system image without compromising performance.
The VSE Quick Start Solution also includes powerful tools from HP OpenView to enable your IT staff to
better manage its IT infrastructure and BEA WebLogic and Oracle installations. HP OpenView tools
enable you to monitor, control, and report on the operation of your infrastructure, including your
network, systems, storage, databases, applications, services, and the Internet.
HP-UX Workload Manager can assign processors to the tasks that require them, when they require
them, synchronizing your IT supply and demand. This means that the system can adapt to bottlenecks
by shifting resources on the fly, away from, for example, a BEA WebLogic installation whose load
requires only one processor for its current task.
The Virtual Server Environment Quick Start Solution is flexible, can be customized through HP
Services, and is supported by HP Customer Support. This solution enables you to demand more from
your BEA WebLogic Server or Oracle9i Database installations—more utilization, more agility, and a
better return on IT.
HP Virtual Server Environment—Real value for your business
The previous explanation of what the HP VSE is and how it works should give you an idea of the
benefits that your organization could gain from its use. While the technological benefits are many, the
key business values derived from implementing HP VSE within the server environment can be
summarized in the following sections.
The industry plays up the reduced TCO and increased RoIT that go hand-in-hand with consolidation.
True, servers and mass storage devices have increased in scalability. But new servers are often
oversized and redundant and only increase the management complexity of the data center.
Moreover, the things that really count—security, availability, performance, and user assurance—have
not arrived either at the same pace or at an affordable price. HP VSE provides a far greater RoIT than
many companies expect because it provides an immediate return through radically improved
utilization of your current infrastructure.
VSE’s integrated, consolidated architecture reduces the amount of hardware required to run current
and future workloads; allows production and non-production application instances to execute on a
single system; automates IT tasks; provides tangible savings on floor space, power, and software
licensing fees; improves asset utilization; reduces the need for additional staffing through
simplification of the environment; makes the IT budget go much further on revenue-generating
opportunities; and aligns IT cost with business revenue.
Enhanced business agility
In today’s world, the old adage “time is money” certainly holds true. Globalization has compressed
the world into a tight, highly competitive environment where a moment’s hesitation can result in
millions of dollars in lost revenues. As companies evolve from decades of monolithic structure building
to highly agile, on-the-fly, opportunistic, accountable business units, their IT infrastructures must be
aligned to support this new business model.
VSE supports this new model. With VSE it is easier, faster, and less costly to turn around a business
request for more or less capacity; for a flexible, adaptive server architecture; or for usage-based
billing. With VSE, companies can be confident that no matter where opportunities might arise in our
unpredictable world—no matter how the business ebbs and flows—their server infrastructure will be
there to support their endeavors.
Improved quality of service
The HP VSE provides real-time provisioning of business services. Companies can capitalize on new
opportunities almost immediately because they no longer have to budget for new hardware and then
wait for delivery. Off-the-shelf or in-house applications can be installed on existing systems
characterized by either low utilization or intermittent demand. IT departments can be more confident
of meeting agreed-upon SLOs.
Most companies purchase hardware during the development phase of a project with a three-year
system life span in mind. This means that the systems are usually sized to meet the performance
requirements anticipated at the end of the third year. However, research shows that approximately
35% of projects never require more capacity than that utilized at the end of the first year. This means
that a lot of unused, redundant hardware is lying around. In addition, 30% of all projects are stopped
within the first year and 50% fail. Because of accounting procedures, the hardware purchased is
rarely transferred to other projects and is generally written off. Other projects that might produce a far
greater return for the company never get off the ground because of either lack of funds or
management skepticism as to the ability of the projects to succeed.
The HP VSE helps to alleviate these concerns and save your budget. The implementation of either
vPars or HP-UX WLM—or both—reduces the over-provisioning associated with new projects. By
developing new applications in partitions on current servers, additional hardware costs can be
eliminated for projects with an uncertain future. Even for projects that do succeed and do require new
hardware, Instant Capacity or PPU solutions eliminate the temptation to oversize, thus saving
considerable expense. User satisfaction is ensured through the use of HP-UX WLM and the re-
allocation of resources to meet business demand. The enormous cost associated with managing the
risk of disasters is significantly reduced by the integration of high-availability systems through
clustering, optimized workloads, and HP on-demand solutions.
Future directions of the HP Virtual Server Environment
As the individual capabilities of HP VSE evolve—workload management, partitioning, utility pricing,
and clustering—its level of integration will be significantly enhanced. Integration between HP-UX
WLM, nPars, vPars, and Instant Capacity resources will increase its flexibility and dynamic attributes.
HP Global Workload Manager
HP Global Workload Manager (gWLM) is an enhanced release of HP-UX WLM. The integration of
gWLM with the VSE will optimize resources in a multi-server, heterogeneous environment composed
of HP-UX, Linux, and Microsoft® Windows® platforms. The complexity and management challenges
associated with multi-tier applications and Web services will be reduced by VSE’s ability to anticipate
and adjust the virtual resource requirements across the entire service based on the demand generated
at the front end.
HP gWLM will enable VSE to provide dynamic resource allocation for applications and business
services within vertical and horizontal scaled environments, such as clustered environments. VSE will
ensure that each service can access the resources it requires to meet its SLOs. This will be done by
adjusting the resource allocations between resource management groups and virtual partitions and, in
a clustered environment, moving applications from one server or hard partition to another based on
resource demand and priority.
In the executive summary at the beginning of this white paper, the HP Adaptive Enterprise vision and
its meaning to your organization was explained. The HP Virtual Server Environment, built around the
only UNIX goal-based policy engine—HP-UX Workload Manager—is a significant step toward the
realization of the HP vision of the Adaptive Enterprise.
John Madden, a senior analyst with Summit Strategies, stated the following with regard to VSE:
“Smart enterprises want to prioritize and automatically adjust infrastructure resource workloads based
on business objectives. This helps set user expectations for performance and clearly communicates the
business value IT brings to the enterprise. HP delivers these capabilities today and, in fact, has a
strong lead in offering customers the most comprehensive set of technologies relative to hard
partitioning, virtual partitioning, and goal-based workload management.”
The orchestration and real-time allocation of server resources according to business priorities that is
enabled by the HP VSE reflects the real world and helps ensure your company will have an agile IT