DELL Open Manage Enterprise Power Manager User Manual

DELL Open Manage Enterprise Power Manager

USER MANUAL

Best Practices to Analyze Multivariate Metrics in OpenManage Enterprise Power Manager using Parallel Coordinates

Abstract

This technical white paper provides information about the best practices for using the Analyze Multivariate Metrics feature—using parallel coordinates—in OpenManage Enterprise Power Manager plugin.

December 2023

Revisions

December 2023: Initial release

Acknowledgements

Author:
Maria Dhinesh Singarajan, Software Senior Engineer, OpenManage Enterprise Power Manager

Support:
Soumyadeep Basu, Technical Content Developer, CE&T

The information in this publication is provided “as is.” Dell Inc. makes no representations or warranties of any kind with respect to the information in this publication, and specifically disclaims implied warranties of merchantability or fitness for a particular purpose.

Use, copying, and distribution of any software described in this publication requires an applicable software license.
This document may contain certain words that are not consistent with Dell’s current language guidelines. Dell plans to update the document over subsequent future releases to revise these words accordingly.
This document may contain language from third party content that is not under Dell’s control and is not consistent with Dell’s current guidelines for Dell’s own content. When such third-party content is updated by the relevant third parties, this document will be revised accordingly.

Copyright © 2023 Dell Inc. or its subsidiaries. All Rights Reserved. Dell Technologies, Dell, and other trademarks are trademarks of Dell Inc. or its subsidiaries. Other trademarks may be trademarks of their respective owners. [12/8/2023] [Technical White Paper] [598]

Executive summary

This technical white paper provides recommendations on using the Dell OpenManage Enterprise Power Manager plugin’s Analyze Multivariate Metrics feature—using parallel coordinates—to effectively analyze the metrics in a data center. The white paper provides insights into the usage of various features of parallel coordinates to analyze multivariate metrics and manage the data center in a more efficient way. A parallel coordinate graph allows data center analysts to visualize, analyze, and evaluate complex multivariate device metrics in Power Manager. Analysts can also use this graph to correlate between various multidimensional time-series metrics.

Audience

This technical white paper is intended for data center managers and server administrators who have prior knowledge about using Dell OpenManage Enterprise Power Manager.

Introduction

1. What is Parallel Coordinates plot?

The Parallel Coordinates visualization technique is a way of visualizing high- dimensional multivariate data. It consists of parallel lines placed at equal distances, with each line representing a different numeric axis for different metrics. It enables data analysts to compare many quantitative variables together looking for patterns and relationships between them. They are appropriate for comparing multiple numerical variables simultaneously when those variables have different magnitudes (different scales) and different units of measurement. The idea is to find patterns, similarities, clusters, positive, negative, or no particular relationships in multidimensional datasets.

1.1 Parallel Coordinates in OpenManage Enterprise Power Manager
In OpenManage Enterprise Power Manager, you can view and analyze metrics collected from devices belonging to different physical rack server groups, simultaneously. This can be used by data center analysts and data engineers to observe patterns and draw conclusions. Each metric corresponds to a vertical axis and each data element is displayed as a series of connected points along the axes.

2. Analyze Multivariate Metrics

To analyze and identify the power anomalies, add the devices and groups to Power Manager. For more information about adding the devices and groups to Power Manager, see the Power Manager User’s Guide at OpenManage Enterprise Power Manager Documentation. OpenManage Enterprise Power Manager provides a graphical view using which metrics collected from different devices can be analyzed together.

Note—The devices that are added to Power Manager have a default monitoring duration set to seven days. You can change the duration from the Duration drop-down menu.

2.1 Visual Controls
This section describes how to use Visual Controls section within the Analyze Multivariate Metrics feature.

• Expand the Visual Controls accordion in the left pane.
• Summary Statistics section contains Minimum (Min), Average (Avg), Maximum (Max), and Instant.
• Metrics section displays all the supported graph metrics.
• You can select any combination of metrics to analyze the data.
• When you select or clear any check box under Metrics, the graph reloads to synchronize the latest changes.
• You can reorder the axis horizontally for better comparison. Click, hold, and drag any metric along the
• Y- axis horizontally to move it. Drag the metric to the required position and release the mouse button to place the metric in its new position.

Filter and Search

This section describes the Filter and Search feature in Analyze Multivariate Metrics. To use the Filter feature, do the following:

1. Click, hold, and drag to select any region of any metrics on the Y-axis.

2. To clear the filter, click outside the selected region of the metric’s axis.

Figure 2 Select a Filter metrics on the Y-axis

Search box in the Metric sources section is used to search the data by Rack or device names.

3. Example Insights

Some examples of observed patterns in a sample testbed are provided in the following screen shots. It uses last 24 hours dataset of racks containing five servers.

1. Server with average power consumption of 550 Watt

• Time window 9 P.M. to 8 A.M. of continuous peak workload. Not loaded as much at other times of the day.
• CPU usage is the major contributor of the power draw. So, it is a CPU intensive workload.
• CPU utilization is within 40% of total capacity, and system usage is 35%. This signifies that this server can take up to 150% more workload for full utilization.
• Inlet temperature averages at 29 degrees Celsius, which might not be ideal for this server.

2. Server with average CPU utilization of 90%

• Time window 6 P.M. to 8 A.M. of continuous workload.
• System usage is in the vicinity of 80%. This server may not be able to service more workload of CPU intensive tasks.
• Average power consumption is within 300 Watt. This server might not consume more than 350 Watt at maximum load.

3. Server with average power consumption of 230 Watt

• Constant power consumption throughout the day.
• CPU and memory bandwidth utilization is 0%.
• This could be a candidate for idle server that can either be powered off to save power, or repurposed.

4. Server with average storage subsystem power of 25 Watt

• Constant power consumption throughout the day.
• Adequate cooling at 20 degrees Celsius throughout the day. All other servers are cooled at above 28 degrees Celsius. Therefore, the server may be part of a different rack server group.
• Primarily a storage server having more than one enclosure, controller and array disk, drawing the power.
• Not ideal for CPU or memory intensive workloads. However, it can take I/O intensive workload for full system utilization.

5. Server with average system usage more than 40%

• Only server in the group where system usage varies between 40% and 80%.
• Average system usage of every other server is always less than 40%.
• This group of servers is under-utilized as a whole.

6. Server with high power usage on weekends

• Figure 5 shows power usage between 400 to 500 Watt on weekends – Saturday and Sunday.
• Figure 6 shows power usage between 250 to 300 Watt on Weekdays – Monday to Friday.

4. Recommendations

Unselect the metrics having low variance in measurements. Constant and categorical data are usually not suited for parallel coordinates analysis. In the following example, I/O Utilization Avg and Total FPGA Power Avg can be removed because values hardly vary.

Avoid over-plotting the view by selecting too many metric types and measurements. It might make it impossible to observe anything. Keep the number of dimensions on the screen to less than 12 at a time to keep it readable and observable. Here is an example scenario that you must avoid.

• If you have fully loaded 42U racks, then do not select more than one rack at a time. The visualization works best when the measurements are a few thousands and not more. Try to keep your observable set under comfortable limits.
• Unselect the metrics which are not supported by the servers in the rack. If the metrics are not supported, the graph lines will reach the bottom “Unsupported” line. You can skip the unsupported metrics for better visibility.

To reorder the metrics, click, hold, and drag the metrics horizontally along the Y-axis. You can use this to compare metrics by placing them adjacently.

5. REST API details

URI: https:///api/PowerService/TimeSeriesMetrics
This URI returns metrics retrieved from all the monitored devices contained in the specified group.
HTTP success response code 201

6. Conclusion

The monitoring and management of data center servers are critical tasks and Power Manager simplifies these tasks for you. Power Manager provides a great insight into the data center, allowing the administrators to take necessary actions, helping you to maximize energy efficiency. This helps in making the data center more efficient in terms of energy and cost, thus helping move towards a more greener data center.

7. Technical support and resources

• Knowledge Base Article about Dell OpenManage Enterprise HTML
• Dell OpenManage Enterprise Power Manager Version 3.2 User’s Guide PDF HTML
• Dell OpenManage Enterprise Power Manager RESTful API Guide version 3.2 HTML
• Dell OpenManage Enterprise Power Manager 3.2 Release Notes PDF
• Dell Technical Support http://www.dell.com/support

Specifications:

• Product Name: OpenManage Enterprise Power Manager

• Version: December 2023

• Author: Maria Dhinesh Singarajan

• Support: Soumyadeep Basu

FAQ

Q: Who is the intended audience for this technical white paper?

A: The intended audience for this technical white paper is data center managers and server administrators with prior knowledge about using Dell OpenManage Enterprise Power Manager.

Read User Manual Online (PDF format)

Read User Manual Online (PDF format)  >>

Download This Manual (PDF format)

Download this manual  >>