 |
|
|
|
Q: What is HP-UX WLM?A: HP-UX Workload Manager is a software product that automatically allocates resources and manages application performance through the use of prioritized service-level objectives (SLOs). It allows you to automatically change CPU allocations for your workloads in response to a variety of events or metrics, including CPU utilization, date/time, and application performance. WLM is most effective managing applications that are CPU-intensive. It adjusts the CPU allocation (cores) of a workload based on the current needs and performance of the applications in that workload. You can set up workloads based on: - nPartitions that use Instant Capacity
- HP-UX virtual partitions
- HP Integrity Virtual Machines (Integrity VM) hosts
- Resource partitions, which can be:
- Whole-core: HP-UX processor sets (pSets)
- Sub-core: Fair Share Scheduler (FSS) groups (WLM creates FSS groups using HP Process Resource Manager, also known as PRM)
WLM manages CPU resources within a single HP-UX instance as well as within and across hard partitions and virtual partitions. WLM is optimized for moving cores among hosts such as virtual partitions and nPartitions. Using hosts as workloads, WLM manages workload allocations while maintaining the isolation of their HP-UX instances. The tools WLM uses to manage CPU resources depend on the software enabled on the complex—such as Instant Capacity, Pay per use, and virtual partitions (vPars).
WLM can manage nested workloads, with workloads based on FSS groups and pSets inside virtual partitions inside nPartitions.
|
| Q: What system resources does HP-UX WLM control?A: HP-UX WLM adjusts CPU resources to satisfy your specified SLOs, across partitions or within a single HP-UX instance. Within a single HP-UX instance, WLM can also control real memory and disk bandwidth, but not based on SLO definitions. |
Q: When should I use HP-UX WLM?A: HP-UX WLM is useful if you want to: - Run multiple workloads on a single system and maintain performance of each workload
- Prioritize workloads on a single system, adjusting the CPU allocations based on the workloads' goals
- Ensure that critical workloads have sufficient resources to perform at desired levels
- Manage by SLOs within and across virtual partitions or nPartitions
- Adjust resource allocations by automatically enabling or disabling SLOs based on time of day, system events, or application metrics
- Enable SLOs associated with a Serviceguard package failover
- Adjust the number of cores in a partition or pSet to meet SLOs
- Grant a workload dedicated CPU and memory resources in the form of a pSet
- Create Secure Resource Partitions (in conjunction with HP-UX 11i v2 Security Containment feature) based on pSets or FSS groups, providing file and resource isolation and automatic resource allocation
- Grant a workload CPU resources in direct proportion to a metric, such as number of processes in the workload
- Set and manage user expectations for performance
- Set minimum and maximum amounts of CPU and memory resources available to a workload
- Monitor resource consumption by applications or users through HP Glanceplus, WLM tools, or PRM tools
|
Q: What is the ideal environment for HP-UX WLM?A: You will benefit most from HP-UX WLM if your environment meets one or more of the following conditions: - You run more than one workload concurrently on a server. The workloads can all run under one instance of HP-UX or in separate partitions, each with its own instance of HP-UX. These workloads can be multiple database servers, a database server and an applications server, or any other combination of workloads, provided that they are on HP 9000 servers running HP-UX 11i v1 (B.11.11) or later, or on HP 9000 or HP Integrity servers running HP-UX 11i v2 (B.11.23) or later.
- You have CPU-intensive workloads that can be prioritized.
- You have an important workload with end-user performance requirements.
- You want consistent performance from applications under varying application and system loads.
- You run Serviceguard and need to ensure proper prioritization of workloads after a failover.
- You want more control over resource allocation than PRM provides.
|
| Q: When should I use WLM vs. gWLM?A: HP-UX WLM is optimized to manage multiple workloads on a single system. It offers more in-depth management capabilities for customers who want to use customized policies in their environment. HP-UX WLM also does not require a central management server. HP Integrity Essentials Global Workload Manager (gWLM) was designed to manage multiple workloads on multiple systems. It requires a central management server, but can easily deploy straight-forward policies to multiple workloads providing simpler management. Also, the central management server provides centralized reporting. HP gWLM is optimized for a centralized IT department managing many servers. |
| Q: What kind of applications can HP-UX WLM manage?A: HP-UX WLM adjusts CPU allocations and is best-suited for managing applications that are CPU-intensive. The quickest way to manage an application is by CPU usage: you can define a CPU usage goal as part of a WLM SLO. WLM also provides the flexibility to base CPU allocations on any metric you provide. You can define goals based on this metric. Metric goals are based on performance data and require understanding of that data. HP recommends using usage goals instead—usage goals can be implemented immediately without prior knowledge of workload performance. HP-UX WLM allows you to automate changing resource allocations based on something that can be measured on the system. If for any reason you want to change allocations on the fly, and if the changes are in response to a measurable system event, the time, or the date, then HP-UX WLM can do that. |
| Q: How does HP-UX WLM guarantee response-time?A: It doesn't. HP-UX WLM doesn't remove the need for capacity planning or application and system maintenance. HP-UX WLM is used to allocate the existing system CPU resources in response to application performance or other system metrics or events. If giving an underperforming application access to additional CPU resources doesn't help, an alarm (EMS event) can be generated to notify the administrator that the SLO is not being met. |
| Q: What happens if the application needs more than its maximum CPU allocation to meet its SLO?A: An SLO violation occurs when the performance of a workload does not meet the workload's SLO. When such a violation happens, HP-UX WLM triggers an EMS event. You can create EMS notifications based on such events to notify a system administrator. The administrator can then determine whether the SLO should be relaxed so that it can be met with the current CPU allocation or if the workload should be granted more CPU resources to meet the current SLO. |
| Q: Can HP-UX WLM control resources on other systems over a network?A: HP-UX WLM controls only the CPU resources on the systems where it is installed. It cannot control multiple systems across a network unless it is installed and configured on every system. You can, however, run WLM in each virtual partition on a system, moving cores among virtual partitions as needed. Also, on a system divided into nPartitions with Instant Capacity software (formerly known as iCOD software) installed, WLM can simulate core migration through core deactivation and activation. |
| Q: How can HP-UX WLM control a metric in a three-tier client-server architecture?A: HP-UX WLM can only control CPU resources on a single system. It cannot prioritize network traffic or client-server applications. |
| Q: How secure is HP-UX WLM?A: Only individuals with root access can activate a new configuration. Any user can run a syntax check of a configuration file. In addition, you can configure WLM to secure its communications using SSL. Also, the HP-UX 11i v2 (or later) Security Containment feature provides file and processes isolation. WLM's integration with this feature allows you to create Secure Resource Partitions, which combine the isolation of Security Containment and the resource allocation controls of WLM. WLM's network interfaces are designed to operate correctly to defend against attacks in a moderate to high threat environment, such as a demilitarized zone (DMZ). (A DMZ is a computer or small subnetwork located between a trusted internal network, such as a private corporate LAN, and an untrusted external network, such as the public internet.) You can use network protections such as firewalls to provide an additional level of defense. |
| Q: What is the overhead of HP-UX WLM?A: The WLM daemon itself consumes very little of system resources— typically no more than 3% on a single-CPU system. Its CPU utilization—as a percentage of the total CPU available—is less on multiCPU systems. In addition, WLM uses the Fair Share Scheduler (FSS). Under this scheduler, the performance characteristics of some applications can be different from what is seen under the default HP-UX timeshare scheduler. The amount and nature of the differences depend upon the application. For most applications, both schedulers produce comparable performance results. If you are currently running PRM and have CPU capping enabled, then the FSS is already enabled much like WLM will use it. Consequently, you would not encounter scheduler-induced differences. Lastly, improperly tuned WLM configurations can also degrade application performance—giving the false impression of a large overhead for WLM. |
|
|
|
|
|
|
Q: How difficult is HP-UX WLM to set up? Do I get a template file or examples to modify?A: You configure WLM by creating a text file. You can create one manually or use the WLM configuration wizard or the WLM GUI. This text file follows a syntax that is simple and easy to use. There is no template file. However, there are numerous example configuration files that illustrate how to use the various features of HP-UX WLM. For configurations with usage-based goals or with goals based on a fixed number of CPU shares or a shares-per-metric allocation, creating the configuration is typically an easy process. WLM tracks CPU utilization for you. With performance metric goals, most of the effort is in getting metrics for your applications.
| Q: How do I decide what are reasonable goals when I am setting up performance goals?A: Instead of performance goals, HP suggests using CPU utilization goals, which you can implement immediately without knowledge of workload performance. Setting up performance goals depends on your application and the data collector being used. Some testing may be required to determine reasonable goals for your situation. |
| Q: Does HP-UX WLM have a GUI for editing the configuration file?A: HP-UX WLM does have a GUI (see the wlmgui(1M) manpage). It also has a configuration wizard (see the wlmcw(1M) manpage). WLM manpages are available at Information Library. |
Q: Where do I get the metrics for my goal-based SLOs?A: Metrics are available from a number of sources: - CPU utilization metrics, which HP suggests using as WLM can track these metrics for you
- GlancePlus metrics, which you can easily use by specifying one of the GlancePlus data collectors in your configuration file (these data collectors are included with HP-UX WLM)
- Oracle database metrics, which can be retrieved using the WLM Oracle data collector
- Apache server metrics, which can be retrieved using the WLM Apache data collector
- BEA WebLogic Server metrics, which can be retrieved using the WLM BEA WebLogic Server data collector
- SNMP metrics, which can be retrieved using the SNMP data collector
- Existing metrics (if you already maintain application-specific or system-specific metrics, you can use that data).
- Dummy transactions, which can be created to give a rough feel for the application performance (these transactions require the same types of processing as the real transactions of interest; the process generating the dummy transactions would then report the performance to HP-UX WLM)
- ARM-instrumented applications (ARM is the Application Response Measurement standard for application instrumentation)
|
| Q: What if I don't want to define an application as an HP-UX WLM workload?A: You do not have to explicitly define all applications as workloads. When managing workloads within a single HP-UX image, any applications you do not place in workload groups are placed in the default workload group OTHERS. These applications then share the resources granted to the OTHERS group. Consequently, you should be sure to grant OTHERS enough resources for all its applications. When managing virtual partitions or nPartitions, you can treat an entire partition as a workload, where any processes running in the partition form a workload. |
| Q: Can I define SLOs for a specific set of users or UNIX groups?A: Yes. Simply assign the users or UNIX groups to a workload group, then define an SLO for that workload group. In more specific terms, it depends on the application. Whatever you define as a workload group can have an SLO. For applications where different processes run under different HP-UX user IDs or UNIX group IDs, and those processes can be put in different workload groups, you can have different SLOs for each of those workload groups. Some applications have their own notion of users, independent of HP-UX user IDs. If those applications support putting their processes in different workload groups, different SLOs can be applied to those groups. Oracle, for example, does not support putting the different Oracle processes from the same Oracle instance into different workload groups. (Oracle does, however, support partitioning one Oracle instance from another by putting each in its own workload group.) For managing the resource allocations among Oracle users, use the Oracle Database Resource Manager. |
| Q: Can I define SLOs for a specific set of processes based on my own criteria?A: Yes, beginning with WLM A.03.01, you can define process maps. Process maps associate a specific workload group with a script or command and its arguments that gather and output process IDs to be placed in that group. WLM spawns the command or script at 30-second intervals. At each interval, WLM places the identified processes in the specified group. The WLM SAP Toolkit, in conjunction with the HP Serviceguard Extension for SAP (SGeSAP) product, takes advantage of process maps, providing a script that enables you to place specified SAP processes in specific workload groups managed by WLM. |
| Q: Can I implement HP-UX WLM when I do a server consolidation?A: Yes, HP-UX WLM is very useful when consolidating servers and can be implemented during the application consolidation phase. |
|
|
|
|
|
|
Q: What is a data collector?A: A data collector is a script or program that provides data to HP-UX WLM, typically for the purposes of gauging performance against an SLO.
| Q: Why do I need to collect data?A: HP-UX WLM requires data about the workloads it is managing so that it can determine whether changes in CPU allocation are needed to meet defined SLOs. With SLOs based on CPU utilization goals, WLM collects data for you. With metric goals, you need to provide data to WLM. |
Q: How can I collect data? Does WLM provide any standard collectors? Do I need to write my own program to do this?A: There are several methods for collecting data: - GlancePlus data collectors (included with HP-UX WLM)
- WLM Oracle data collector
- SNMP data collector
- Scripted data collectors invoking the wlmsend command
- Scripted or binary data collectors providing data on stdout
- Binary data collectors providing data through the native HP-UX WLM data collection API
For information on using collectors included with HP-UX WLM, see the wlmrcvdc(1M) and wlmsend(1M) manpages. Also see these manpages for information on creating scripted data collectors. For information on using the HP-UX WLM API, see libwlm(3). |
|
|
|
|
|
|
Q: Are there any tools for viewing the state of the HP-UX WLM world? For example:
- What applications are currently running under HP-UX WLM?
- What resources have been allocated to these applications?
- What level of resource consumption have they reached?
|
A: WLM provides a number of ways for monitoring WLM operations: - Data based on SLOs, metrics, or workload groups is available through a command-line interface (the wlminfo utility) and through a graphical user interface (the wlmgui utility).
- WLM reports its configuration via the EMS interface, which is available through most of the monitoring tools.
- WLM also generates a statistics log file, /var/opt/wlm/wlmdstats, when wlmd is invoked with the -l option.
- Additionally, GlancePlus and PRM (prmlist, prmmonitor) can be used to track the actual allocations and usage, in real time.
- HP-UX accounting records can be used to track historical usage.
Also, refer to Information Library for Using HP PerfView to view HP-UX Workload Manager statistics. |
| Q: Is it possible to monitor resource consumption by applications or users through GlancePlus or PRM tools?A: Yes, in addition to WLM's own wlminfo utility, you can monitor consumption by PRM groups within GlancePlus. |
| Q: How are the EMS components configured? Is this done transparently by installing the product?A: EMS notifications are not automatically configured. You must use SAM or SMH (the enhanced version of SAM) to configure EMS components for notification. |
| Q: Can I send status information to monitoring tools other than EMS?A: Using SAM, you can specify that EMS notifications be sent to any EMS client, such as VP/O. EMS can send notifications in the following formats: SNMP, TCP, UDP, syslog, email, console, or textlog. Most of the popular monitoring tools (VP/O, Unicenter, Patrol, Tivoli) can receive and process SNMP traps. |
| Q: Are there alarms or notifications when applications reach limits?A: WLM generates EMS events. You can then create EMS notifications to pass to VP/O to alert someone or some process. |
| Q: Can HP-UX WLM be integrated in an Openview framework? That is, can I have several servers with HP-UX WLM and visualize it from a VP/O (IT/O) window?A: Yes. There are commands to monitor the running of an HP-UX WLM configuration. Using SAM or SMH, you can create an EMS monitoring request to monitor the status of HP-UX WLM and its SLOs. You can specify that notification be sent via SNMP, which can be monitored using VP/O. |
|
|
|
|
|
|
Q: Can I still use my PRM configurations after I install HP-UX WLM?A: HP-UX WLM will only take control once it is activated. You can have HP-UX WLM installed and still use PRM in its normal operation. Once HP-UX WLM is started, the PRM configuration it creates replaces the current PRM configuration—if any. As of WLM A.03.01, PRM is no longer included with the WLM bundle. If PRM C.03.00 or later is already on the machine on which you must install or upgrade WLM, you can continue to manage FSS and pSet-based workload groups (just as if PRM had been installed with WLM—when you purchase WLM, you receive a PRM license that enables you to continue to use PRM). If you are installing WLM for the first time on a machine, you can use a strictly host-based configuration (no FSS or pSet workload groups). However, to manage FSS and pSet-based workload groups, you must install PRM (C.03.00 or later) separately. Additionally, HP-UX WLM provides a script that converts your PRM configuration file into a template for an HP-UX WLM configuration file.
| Q: How do PRM, GlancePlus, and HP-UX WLM work together?A: HP-UX WLM defines workloads within a single HP-UX image as workload groups, which are based on PRM groups. GlancePlus can generate PRM Group reports. WLM allows monitoring through the EMS monitor. You can use a variety of tools (any that can receive and process SNMP data) to monitor the status of HP-UX WLM and the defined SLOs. PRM tools (prmlist and prmmonitor) enable you to monitor the current PRM configuration and resource usage by PRM group. |
Q: Can HP-UX WLM, gWLM, and PRM be used to manage the same system at the same time?
A: With one exception, the HP products HP-UX WLM, gWLM, and PRM should not be used to manage the same system at the same time. Using these products to control resources at the same time can cause inconsistent behavior and undesirable performance. However, you can use WLM and PRM at the same time if the PRM configuration uses FSS groups only (no pSet-based groups) and the WLM configuration is strictly host-based. (A strictly host-based configuration does not include a prm structure; it is designed exclusively for moving cores across partitions or for activating TiCAP or PPU cores.) You might want to use both products on the same system to take advantage of certain features of PRM that are not included with the latest release of WLM, such as PRM’s CPUCAPOFF mode, enabled with the prmconfig –M CPUCAPOFF command. (In this mode, a PRM group’s upper bound for CPU resource consumption is determined by the CAP value, available on HP-UX 11i v3 (B.11.31) or later. For more information, see the HP Process Resource Manager User’s Guide or prmconfig(1M).)
|
| Q: How does HP-UX WLM integrate with Serviceguard?A: You should install WLM on each node in the Serviceguard cluster. Also, you should create a single WLM configuration file to handle all the Serviceguard packages in the cluster. Include the WLM command sg_pkg_active in the configuration to enable SLOs for a package when the package becomes active on a given node. Distribute this file to all the nodes in the cluster. Activate HP-UX WLM on each node. For detailed step-by-step information on configuring WLM with Serviceguard, see the white paper "More efficient high availability and resource utilization through manageability". |
Q: How does HP-UX WLM integrate with SAP?A: WLM and its SAP Toolkit (SAPTK), in conjunction with the SGeSAP product, take advantage of the WLM process map feature to enable you to identify different SAP processes and place them into separate workloads. The SAPTK process ID identifier, called wlmsapmap, enables you to identify and isolate entire SAP instances (or just subsets of an instance's processes) as a separate workload. For example, you can use wlmsapmap to collect SAP batch and dialog (interactive) processes and place them in separate workload groups. WLM can prioritize and assign specific SAP processes to workload groups. You can: - Isolate an entire SAP system as a workload
- Isolate an SAP instance as a workload
- Separate SAP processes from an instance and place them in a workload
For more information, see the white paper, "Using HP-UX Workload Manager with SAP", also available at /opt/wlm/toolkits/sap/doc/sap_wlm_howto.html .
|
| Q: Is HP working with ISVs to "ARM" their products?A: Because ARM is an industry standard, there is continued effort to have ISVs ARM their applications. With HP-UX WLM, HP hopes that this effort will be accelerated. The HP ARM implementation is available through GlancePlus. Install GlancePlus on each system where you want to track ARM data. |
|
|
|
Printable version
