| United States-English |
|
|
|
![]() |
HP Capacity Advisor Version 4.1 User's Guide > Chapter 2 FeaturesData Collection |
|
Data collection is at the heart of Capacity Advisor and a task is created during configuration to automatically collect utilization data from all licensed VSE resources added to HP SIM. (See “First Data Collection (or the Automated Nightly Collection)”.) HP recommends that once you have become familiar with using Capacity Advisor, you should review this scheduled task to understand what it is doing and make sure that it fits your needs. This automated data collection task gathers historical data from all your systems so that the data is available when you need it for analysis. In the data collection infrastructure, each component plays a role, as shown in Figure 2-1. It all begins with data collection. Collecting Data for a Workload with a HP Serviceguard PackageStarting with VSE Management Software version 3.0, you can associate a VSE-monitored workload with a Serviceguard package. This association will allow VSE Virtualization Manager to show workloads moving from one system to another within minutes of the Serviceguard migration. It will also allow Capacity Advisor to show an uninterrupted history of utilization data for a workload across these migrations. Associating VSE monitored workloads and Serviceguard packages requires that Virtual Vault has been installed on the Central Management Server (CMS) and that the corresponding version of the Utilization Provider has been installed on all member systems of the Serviceguard cluster. All nodes of the cluster must be running the appropriate version of the Utilization Provider. For specific version information, see the HP Systems Insight Manager screen: Help About HP Systems Insight Manager. This help topic compares the potential benefits and possible drawbacks that accompany the use of agentless data collection versus using the Utilization Provider (UP) to collect data from systems licensed for Capacity Advisor. Table 2-1 A Comparison of Agentless Versus Utilization Provider Data Collection
Differences in data imported from HP Performance Management Pack (HP PMP) affects how you should interpret resulting analysis and scenarios as compared to data collected from the Utilization Provider. Capacity Advisor normally measures memory collected from the Utilization Provider, which returns the sum of memory allocated to the processes on the system. HP PMP collection provides the total memory less the zeroed, free, and standby memory. This means that HP PMP counts memory used by the operating system, whereas the Utilization Provider does not. The memory use reported by HP PMP is about 20% higher than memory use reported by the Utilization Provider. Because Capacity Advisor assumes a different memory amount than HP PMP provides, Capacity Advisor may overestimate the memory needed when using HP PMP data to analyze stacking several application instances in a system running a single instance of an operating system. While either the Utilization Provider data or the HP PMP data can be used to plan memory size in a new configuration, you should not use a mix of HP PMP and Utilization Provider data when looking for a trend in memory utilization. HP PMP does not report the correct network IO when network interface cards (NICs) are being teamed for network fault tolerance (for example, by using HP Network Config utility). That is, for a system with 2 NICs teamed as one single logical NIC, Utilization Provider reports the addition of TX and RX Mbps for the teamed NIC. HP PMP adds the metrics from the 2 NICS, ignoring that they are working as a teamed NIC. For a teamed configuration, the HP PMP network utilization curve will be always twice that of the Utilization Provider data. If you know you have teamed NICs, or if you suspect your imported network data is twice as high as it should be, you will want to instruct Capacity Advisor to ignore the imported data. When logged in as a user authorized to use the Capacity Advisor toolbox on the CMS, you will be able to mark data as invalid. You can do this from the Profile Viewer, or by using the capprofile(1M) command. |
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
|||||||||||||||