PAGE 28 Workload Resource Guarantees with gWLM VSE allows IT to control the level of resources given to workloads using gWLM and Integrity VM. One aspect of this control is the ability to set guarantees. In many cases, servers are sized – and therefore the cost to IT is based – on these guaranteed levels, making this an attractive parameter to use in chargeback. In the next example, the policy has an Owned Size as well as a Min Size and Max Size: In such cases, the Owned Size value is the guaranteed amount of resources for workloads with this policy. CLI: gwlmreport config command Workloads that are being actively managed by gWLM are part of gWLM Shared Resource Domains (SRDs). The following example is a section of the gwlmreport config command output. The example includes a list of managed SRDs and the workloads that are in them. ######### #### SRDs ######### … SRD - va01sv.srd Mode : Managed Status : Deployed Interval: 15 seconds Workload - va01sv.zko.hp.com.OTHER Policy : Owns_2-Max_4 Compartment Type: hpvm Workload - vmwin1 Policy : owns.5-max-remaining Compartment Type: hpvm Workload - vmhpux2 Policy : Owns_1-Max_Remaining Compartment Type: hpvm Workload - vmhpux1 Policy : Owns_1-Max_Remaining Compartment Type: hpvm Workload - va00vm00.zko.hp.com Policy : owns. In this output, note that the Owned value for the policy can be determined. In cases where the GUI reports only a minimum and maximum value, the gwlmreport output includes an Owned value that is equal to the minimum value. Note also that many more policies might be defined and listed than those shown to be actually in use. Workload Resource Guarantees with Integrity VM For Integrity VM guests that are not being managed by gWLM, the amount of guaranteed CPU is known as an entitlement. The net amount of guaranteed physical core processing is the entitlement percentage multiplied by the number of virtual CPUs (vCPUs) of the virtual machine. GUI: IntegrityVM Manager You can use the Integrity VM Manager tool to obtain information about virtual machines. The hpvmstatus command runs on a specific Integrity VM Host server and shows the summary properties of the virtual machines on that server. In the following example, the –r flag specifically shows the vCPUs and entitlement values: HPVM HOST # hpvmstatus -r [Virtual Machine Resource Entitlement] [Virtual CPU entitlement] Percent Cumulative Virtual Machine Name VM # #VCPUs Entitlement Maximum Usage Usage ==================== ===== ====== =========== ======= ======= ================ vmlin1 1 1 80.0% 100.0% 0. Listing Physical Servers This section describes how to identify the physical servers in the VSE environment. A feature of HP SIM is its ability to organize and present a vast array of different systems for management and monitoring. It also includes optional features to allow the user to add information for asset tracking. It provides both GUI and CLI interfaces. va00-cplx_USE4452HL6 is a complex server divided into two nPar workloads. va01sv is a noncomplex server with 4 VM guest workloads running on it. It also includes the VSE-created Other workload. va02cm is a noncomplex server with two user-defined fss workloads and the additional VSEcreated Other fss workload. legacy01, legacy02, and legacy03 are noncomplex servers with whole OS workloads. CLI: mxnode and mxquery commands HP SIM keeps many parameters about systems it has detected in its database. For a complex (cell-based) physical server o Device type: complex For a all other (noncomplex/non-cell-based) physical servers o Device type: Server o Device subtype: HP Integrity or Device subtype: HP9000 o None of the following entries Device subtype: nPar Device subtype: vPar Device subtype: HPVMGuest The following is a simple shell loop that extracts all node records from the mxnode output and looks for physical servers: CMS # cat ./listphys. All c-Class Racks All HP Integrity VM Guests All nPartitions All Virtual Partitions Events older than 90 days … CMS # From a HP SIM perspective, the All Systems output can encompass many different types of devices that HP SIM can reach on the network. Therefore, HP SIM might return many entries that do not relate to VSE servers. Similarly, HP SIM’s multiplatform support means that All Servers might return more than just the VSE managed physical servers that are of interest. HPVMHost vParMonitor nPar vPar HPVMGuest CMS # The highlighted “INCLUDE” and “EXCLUDE” flags indicate which kinds of servers match this query. Notice that this query lists all noncomplex physical servers except those that are also VM Host servers. A custom query that does not exclude these servers is easy to create. grep "^DeviceName" | awk '{printf "%s Non-Complex\n",$2}' CMS # This procedure produces the following output: CMS # ./listnoncmplx. Workload and Physical Server Associations Earlier sections of this paper demonstrate how to list workloads and physical servers in the environment. This section describes how to combine the information to determine the physical server of a workload. This is straight forward with the GUI tools but requires some filters to put into a form that is suitable in a larger script. GUI: VSE Manager All VSE Resources page As described earlier, the VSE Manager All VSE Resources page shows the complete hierarchy. CLI: gwlmreport config tool As described earlier, the output of the gwlmreport config tool can be used to list all workloads. A filter can be used to remove all but the end “leaf” workloads of the hierarchy that are associated with end-user processing. This output contained a line for each workload with the workload’s name, type, and host: CMS # ./listwklds2.sh Generating historical config report. Please be patient. This may take several minutes. vmhpux1 VM va01sv.zko.hp.com vmlin1 VM va01sv.zko.hp. gwlmreport config | awk -v RS="" '{printf "%s %s %s %s\n",$1,$3,$6,$9}' | grep " ^Workload " | awk '{printf "%s %s %s\n", $2, $3, $4}' | grep -v "Other" > $WKLDS FILE # filter workloads to leafs for w in `cat $WKLDSFILE | grep Whole | awk '{print $1}'`; do cat $WKLDSFILE | grep -v Whole | \ awk ' BEGIN {WM=0} \ {if (match($2,"'$w'")) {WM=1; printf "%s %s %s\n",$1,$3,"'$w'"}} \ END {if (WM==0) {printf "%s Whole %s\n","'$w'","'$w'"}}' done > $WKLDSFLTRD # for each workload, identify host type using mxnode for Physical Server Serial Numbers and Asset Tags If an organization is interested in the relationship of physical servers to cost, it can be useful to provide more than just a server name in resource allocation data. HP SIM discovers and maintains server serial numbers and provides a field for customers to use for an optional asset tag. CLI: mxnode and mxreport commands For complex physical servers, the output from the mxnode command contains both the asset tag and serial number. In the following example, the –ld flag is used with the mxnode command to obtain this information for a node that is known to be a complex: CMS # cat ./cmplxsn.sh mxnode -ld $1 | \ grep -E "CreatorSerialNumber|^Asset" CMS # This command produces the following output: CMS # ./cmplxsn. "va01sv","Server","ia64 hp server rx4640","USE4452N2H","0001","33541676","HP-UX","HP","","" "va02cm","Server","ia64 hp server rx2600","US40576064","0002","8377064","HP-UX","HP","","" 45 Physical Server CPU Cores and Memory As described earlier, IT might want to allocate the costs of physical servers across its workloads proportionally. If this is the case, then you must first determine the CPU core and memory configuration of the physical server. GUI: HP SIM System page The HP SIM System page for a physical server can be used to obtain component information for servers. CLI: mxquery and mxreport commands CPU Cores For noncomplex physical servers, the mxreport command produces a report that shows the number of CPU cores. The “CPU – Servers” query lists all CPU cores for all noncomplex servers with one line of output per CPU core. hp.sim.aggr.alerts: 16 DeviceName: va00-cplx_USE4452HL6 DeviceType: 19 DeviceAddress: ProductName: OSName: DeviceAddress: DeviceKey: 10 FullDNSName: AssociationType: AssociatedDeviceType: 0 AssociatedDeviceName: AssociatedDeviceKey: 0 TimeStamp: 1191273186704 … CMS # mxquery –e “All nParitions”: Lists all partitions of all complexes. DeviceName is the name of the partition, and AssocatedDeviceName is the name of the complex to which it belongs. "va00np00","Intel(R) Itanium(R) 2 Processor","1500","1" "va00np00","Intel(R) Itanium(R) 2 Processor","1500","2" "va00np00","Intel(R) Itanium(R) 2 Processor","1500","0" ... CMS # The following is an example of a simple script that issues the three commands and processes the output to count CPU cores in a complex: CMS # cat cmplxcores. The mxreport “Inventory - Servers” command also provides the same information about partitions of a complex physical server. To find the total memory for a complex, the amount in each partition must be added together. The following example script uses the mxquery “All nPartition Servers”, mxquery “All nPartitions” and mxreport “Inventory – Servers” commands to report the amount of memory for complexes in the environment. For example: CMS # cat ./cmplxmem. Physical Server iCAP The iCAP and TiCAP of HP’s Utility Pricing Solutions for HP Integrity and HP 9000 servers allow customers flexibility in how they pay for and share computing resources. When iCAP and TiCAP are used, you must gather additional information from VSE for chargeback. In iCAP configurations, the number of physical CPU cores and the amount of physical memory in a server might exceed the number and amount that the customer is entitled to use through purchased usage rights. Step 1: Verify Target Systems o Select the systems you selected in the previous screen. Step 2: Task Confirmation o Click Run Now at the lower right. Task Results o If the tool definition and system selection are correct, the tool runs properly and the “Status:” field indicates “Complete”. o The Stdout tab under Target Details displays the output of the remotely run icapstatus command, as shown in the following example screen: The next section describes the fields in the preceding output. --------------------------------Number of cells without usage rights: 0 Number of inactive cells: 0 Amount of memory without usage rights: 0.0 GB Amount of inactive memory: 0. Elapsed Time :1 second 288 milliseconds Node :va00np00.zko.hp.com Status :Complete Exit Code :0 STDOUT : Software version: B.08. for i in `mxquery -e "All nPartition Servers" | grep "^DeviceName" | awk '{print $2}'`; do printf "%s " $i let CMEM=0 for n in `mxquery -e "All nPartitions" | awk -v RS="" '{ printf "%s %s\n",$34,$12}' | grep $i | awk '{print $2}'`; do let NMEM=`mxreport -e -n -S $n "Inventory Servers" -x CSV | sed 's/\"//g' | grep "^$n" | awk -v FS="," '{print $6}'` let CMEM="$CMEM + $NMEM" done let ULMEM=`mxexec -t "remote icapstatus" -n $n | grep "Amount of memory without usage rights" | awk '{print $7}'` let CMEM="$CMEM Physical Server TiCAP If a complex server configuration uses TiCAP as well as iCAP, additional cores without usage rights in a server can be activated and deactivated using separately purchased temporary capacity. TiCAP activations of cores represent a change in the pool of resources a workload is drawing on and, therefore, might be important to computing the relative utilization of a workload. This script produces the following output: COMPLEX PARTITION 09/10/07 19:30:04 09/10/07 20:00:04 09/10/07 20:30:04 09/10/07 21:00:04 09/10/07 21:30:05 09/10/07 22:00:04 09/10/07 22:30:04 09/10/07 23:00:04 09/10/07 23:30:04 09/11/07 00:00:04 … 10/02/07 00:00:04 10/02/07 00:30:05 10/02/07 01:00:04 10/02/07 01:30:04 10/02/07 02:00:05 10/02/07 02:30:04 COMPLEX PARTITION # ./icapdebits. Summary IT organizations that require resource allocation data for chargeback can extract that information from the HP VSE environment using the many GUI and CLI tools provided. This paper describes the parameters that might be of interest and provides examples of how to use specific tools to determine their usefulness for a chargeback reporting period. For more information For more information about HP VSE virtualization technologies, visit the VSE website at www.hp.com/go/vse. © 2007 Hewlett-Packard Development Company, L.P. The information contained herein is subject to change without notice. The only warranties for HP products and services are set forth in the express warranty statements accompanying such products and services. Nothing herein should be construed as constituting an additional warranty. |