Extracting Resource Allocation Data for Chargeback in a HP Virtual Server Environment for HP Integrity Servers

Host2 (VM)
Host3 (VM)
Host4 (VM)
Host1
(non-complex Server)
VSE Workloads (as reported by gwlmreport config)
Complex1 (Complex)
Host6
(VPAR)
Host7
(VPAR)
FSS1
FSS2
Host5
(NPAR)
(NPAR)
Other
HPVM Guest
HPVM Guest
HPVM Guest
Whole OS
Type
No
Yes
Yes
Yes
No
End user
workload?
Physical
Server
Workload
Name
Host1Host1.OTHER
Host1Host4
Host1Host3
Host1Host2
Host1Host1
Whole OS
Whole OS
Other
gWLM-created
gWLM-created
Whole OS
Type
Yes
Yes
No
Yes
Yes
No
End user
workload?
Physical
Server
Name
Complex 1Host7
Complex 1Host6
Complex 1Host5.OTHER
Complex 1FSS 2
Complex 1FSS 1
Complex 1Host5
Figure 9 Example hierarchies and workloads of interest
The noncomplex server is subdivided using Integrity VM with three VM guests. VSE associates four
workloads with this server: one workload for the underlying whole OS workload, one each for the
three VM guests, and one Other workload. Of these four workloads, only the three VM guests
represent user workloads.
The complex server is subdivided into two nPartitions. One of the nPartitions is a whole OS workload
running fss workloads. The other is subdivided into two vPars, each of which is a whole OS
workload. Of all the workloads that VSE associates with the configuration, only the two fss workloads
and the two vPar workloads are of interest from a billing perspective.
Automated control and resource guarantees
A powerful feature of VSE is its automated control. Automated control allows IT to set up resource
pools that can be shared between multiple workloads. The primary automated control engine of VSE
is the gWLM component. This technology can control all workload types.
An additional component of VSE that can result in guarantees is Integrity VM and its entitlement
capability. This technology only applies to VM guest workloads which are not already under gWLM
control. If the VM guest workload is under gWLM control, the the gWLM guarantees take precedence
and those of Integity VM’s entitlement capability are ignored.
Choosing which resource parameter to measure
The resource being shared is the physical server that can have many performance utilization attributes
to use in determining each user’s usage. The existence of multiple parameters does not mean that all
are needed or useful, so you can use discretion in selecting them. Figure 10 uses the analogy of a car
rental, which applies a combination of rental period, mileage, and fuel level and omits parameters
such as tire wear, oil level, and transmission fluid level. In both the car rental and server chargeback
situations, multiple usage paramters are available, but not all are necessarily needed.
11