Domain Intelligence Offered by SAS IT Resource Management

SAS IT Resource Management delivers sets of supplied reports that communicate trends, forecasts, predictions, and analytical results. It also delivers SAS analytical calculation and illustration capabilities as well as specific implementations of these capabilities for IT performance management and capacity planning. One such example of these special implementations is the exhaust forecast chart. This type of chart calculates and graphs the amount of time that is predicted for a particular aspect of an IT resource to reach a threshold.

As illustrated in the preceding exhaust forecast chart, the actual measurements (shown as yellow points) are used to create a forecast for the CPU Busy Percent with 2 Standard Deviations (shown in the purple line). This purple line indicates the variations over the reported time period (in this case, a span of four months) and the forecast (for the next two months). Next, this chart shows the trend line (the green smooth line on the chart). Finally, the upper and lower confidence boundaries are presented as an indication of the accuracy of the predicted and forecasted values presented on the chart.

A primary job function of the capacity manager is to appropriately size the IT enterprise to accommodate the needs of the business that it services in terms of the availability and performance of IT resources. Successfully performing this task requires the capacity planner to:

Exhaust reports are an excellent means to accomplish this task. Exhaust reports are possible through the delivery of SAS statistics and advanced analytic capabilities with SAS IT Resource Management. For any IT element for which a fixed capacity exists (such as CPU, disk, memory, bandwidth, and so on), exhaust reports are essential. These reports can predict, with adequate lead time to respond, the exhaustion of those IT resources. Exhaust reports can also be used to anticipate the obsolescence of resources from the IT enterprise. SAS analytics can forecast and predict any aspect of the IT data available from the SAS IT Resource Management solution.

The forecast chart is configured to accommodate the amount of time that will be required to procure, install, and configure additional components to meet the demand for this server to ensure that the IT enterprise sustains the business that it supports.

The preceding example used data from a server to illustrate exhaust reporting. IT performance data from Windows and UNIX systems is derived from many sources. SAS IT Resource Management supports adapters for many of them, including those from HP, BMC, Microsoft, Demand Technology, and many UNIX operating system providers. Therefore, a consolidated IT performance data warehouse, such as the one delivered by SAS IT Resource Management, enables this data to be analyzed in a coordinated and consolidated fashion. SAS IT Resource Management also delivers special key metric summarizations of this data that present the most common measurements for these systems across the multiple adapters that deliver them. This facilitates long-term trending and analysis for and across all systems that are represented in the IT data warehouse.

For more information about key metric summarizations, see “What Are Key Metrics?” in the “Data Model Appendix” in SAS IT Resource Management: Administator's Guide.

Tile charts are rectangular maps. (Tile charts are sometimes referred to as treemaps.) They enable you to view a large quantity of hierarchical data in a limited space. Each unique category combination is represented by a rectangular tile whose size and color are determined by response variables. These tiles are placed in a hierarchical arrangement. You can drill down into a tile to view it in more detail. You can also link to reports that are associated with tiles.

To maximize the performance and capacity of any IT environment, you can create a consolidated IT performance data warehouse and provide the tools to analyze and statistically enrich the data within it. Given this premise, tile charts are ideal for use by performance managers and capacity planners to provide both a first glance and a detailed view that enables the evaluation of large volumes of data. The scenarios that illustrate the value of this report technique include the evaluation of large volumes of data for outlying measurements that are likely indicators of issues that need to be addressed in the IT infrastructure. Tile charts have been used by SAS IT Resource Management customers for consolidating and initializing servers, balancing workloads, creating high availability resource groups in support of critical business processing, orchestrating and defining the cloud, network tuning, and so on.

Tile charts are used by performance managers and capacity planners to evaluate data from a large number of resources that would otherwise be extremely difficult to work with. In the preceding display, the size of each tile shows the CPU count and the color indicates the CPU Busy Percent weighted mean. This provides an overall status of the VMware environment from a cluster perspective for a period of time. You can drill down to view the guests that are allocated to them.

From this single image, the performance manager can identify the utilization issues that need attention. These issues are easily identifiable because they are represented on the tile chart by boxes with the most intense colors (oranges and reds). Starting with this one quick view, the performance manager can prioritize issues.

Achieving benefits from virtualization technologies lies within the implementation, measurement and management of the clusters, hosts and guests that are established in that environment. By effectively managing the virtual infrastructure, the perception that virtual machines are established without a corresponding cost can be dispelled. By measuring the utilization, availability and performance of this environment, the anticipated value of virtualization can be proven and opportunities to further increase that value can be identified.

In the preceding chart, each box represents a VMware Cluster and the VMware guests that are allocated to the Cluster on a particular day, and for each hour of that day.

Three-dimensional vertical bar charts show the relationships between three variables or statistics.

Presenting volumes of data from a large quantity of resources with vertical block charts is most useful when the relative magnitude of the bars is more significant than the exact magnitude of any particular bar.

In the preceding display, the vertical bar chart shows the CPU Ready Percent for all guests that are allocated to a VMware cluster. In VMware environments, closely monitoring the CPU Ready Percent measurement is critical. This measurement is the percentage of time for which a virtual machine has work that is ready to be performed but for which there is no physical machine ready to process that work. In well-tuned environments, this measurement should never exceed 10%. It should typically be a value of less than 5% in order to ensure that the needs of each of the consumers of virtual resources are satisfied. Any value in excess of this recommendation is cause for additional IT performance evaluation.

For an alternative view of this data, right-click anywhere on the chart to display pie charts for each guest in the cluster. Pie charts create simple, group, or stacked charts that represent the relative contribution of the parts to the whole by displaying data as wedge-shaped "slices" of a circle. Each slice represents a category of data. The size of a slice of the pie represents the contribution of the data to the total chart statistic.

Right-click a pie chart to display the graph toolbar.

As shown in the preceding display, the graph toolbar provides a button () that enables you to rotate the graph. You can also zoom into the graph by clicking .

The pie charts enable performance and capacity analysts to gain a detailed perspective for a particular measurement that is associated with a designated resource. In the preceding example, if a particular cluster had a guest with a disproportional value for CPU Ready Percent, that value would be easily identifiable in a pie chart view.

Tabular reports enable you to generate cross-tabulation tables, also known as contingency tables, from your data. Tabular reports are used extensively for communicating performance and capacity information. They typically present measurements in ranked orders to focus the report consumer's attention most appropriately.

For the purposes of IT performance management and capacity planning, tabular reports are good for displaying measurements to view and understand the relationship (cause and effect or correlation) between them. Examples include the analysis of directly related measurements like CPU and Run Queue or Memory and Paging. Tabular reports are also useful for readily identifying suspect performance values that warrant further investigation.

The preceding tabular report shows the available CPU Busy, hardware millions of instructions per second (MIPS), hardware millions of service units per second (MSUs), and CPU Active Time for all engine types that are available on a particular mainframe system for each hour of the day. The report includes standard, System z Integrated Information Processor (ZIIP), and System z Application Assist Processor (ZAAP) engine types. The report is sorted by All Engines CPU Hardware Mips Total. It presents the consolidation of key metrics data that is fundamental to managing the CPU performance and capacity of an IBM mainframe system.

The System CPU Busy Percent measurements that were used to create this report came from RMF data. SAS IT Resource Management obtains RMF data through MXG. For other sources of data that are made available to SAS IT Resource Management through MXG, see Supported Adapters in SAS IT Resource Management: Administator's Guide.

SAS IT Resource Management accommodates the analysis of mainframe and special-purpose processors as well as the unique applications that operate on these systems. SAS IT Resource Management can provide an analysis of the eligible Java work (that is, work that can be performed on a lower cost special-purpose engine) so that the cost advantages of implementing special- purpose engines can be explored. In the preceding report, the MSU values are calculated using appropriate CPU active times (based on the type of processor) with a hardware service unit factor that corresponds to the CPU model on which the resource is consumed. The MIPS values in the report are calculated from the resulting MSU values by multiplying with an MSU-to-MIPS conversion factor that also corresponds to the CPU model.

SAS IT Resource Management adapters enhance the value of IT mainframe performance measurements by providing mainframe processor performance data in the following formats:

IBM Large System Performance Reference (LSPR) tables are owned by and available from IBM. They are used as reference information to enable the calculation of MIPS and MSUs from the expended CPU time. Hardware MSUs are used in the physical engine type calculations for systems and logical partitions (LPARs). Software MSUs are used for application measurements.

For more information, see “How MSU and MIPS Columns Are Included in Staged Tables for the SMF Adapter” in the “Working with MXG Code to Stage Data” section of Appendix 2, “Data Sources Supported by SAS IT Resource Management Adapters” in the SAS IT Resource Management: Administrator’s Guide.

Stacked bar reports show the relative magnitude of data by displaying bars of varying height. Each set of color-coded segments in a bar represents a category of data. Stacked bar charts are most useful when the relative magnitude of the segments in the bars is as significant or more significant than the exact magnitude of any particular bar. Stacked bar charts display the color-coded segments for each category of data, one on top of the other.

Reports of this type are essential to understanding the allocation of specific resources to workloads for a host system, network component, or virtual system environment. These reports also help create load-balancing strategies and enable users to adjust their workloads to minimize the use of resources.

Evaluating workload measurements such as transaction counts, I/O and exception rates, response times, and CPU, disk, and memory performance attributes is crucial for ensuring service quality and assessing charges that should be allocated to business units that share computer resources. Stacked charts are useful for understanding the parts of a whole entity, whether that entity is a resource, a segment of time, or any other quantifiable object. Facts that are communicated by stacked charts, along with other data, are effective in determining the proper allocation of resources to the business tasks that must be accommodated. For example, certain business processing must take place in a given time period. Therefore, the resource to accommodate this processing must be available.

The presentation of any data in a stacked bar chart enables you to make quick comparative observations. Viewing the resource utilization and workload components of an IT resource over the hours of a day enables the IT performance manager to understand how that resource can best be allocated and used.

In the preceding report, the CPU resource (specified in software MIPS) is segmented by the applicable active workloads defined to the system (typically through the RMF Workload Manager). Workload usage of the resource can be reviewed and analyzed. This analysis enables capacity planners to understand exactly how much of the resource is being used by workloads and when those resources are being used.

Line plots connect data points with straight lines. Points are connected in the order in which they occur in the input data set. From an IT perspective, line plots are most often used to show a particular aspect of IT performance over a designated period of time.

Line plots show the mathematical relationships between multiple numeric variables by revealing trends or patterns of data points.

Line plots are used extensively for the purposes of performance management and capacity planning. They are a quick and easy way to understand variations on a single metric or comparative cause and effect relationships between multiple variables. For example, the CPU utilization of a resource over the course of an hour, day, week, or month can yield valuable information about the business cycles that are being accommodated by that resource. These facts are essential to planning for the capacity of that device. Similarly, evaluating the relationship between CPU utilization and system run queues, as well as memory utilization and paging rates, yields facts that can help capacity planners configure resources that can best accommodate assigned workloads.

Establishing and understanding the utilization and availability profiles of resources and sets of resources is useful for managing the performance and capacity of those machines. This type of time-based analysis enables you to readily identify exceptions to anticipated performance activities.

For example, at 9:00 a.m. resource utilization increases to accommodate the beginning of the work day, and you can observe user login activity. You can observe a decrease in utilization around noon when employees are at lunch. This observation is followed by another peak that ends around 5:00 p.m. when the work day ends. Overnight, peaks and valleys in utilization illustrate batch processing activities, routine system maintenance work such as backups, and lulls that are associated with business inactivity. Viewing this behavior over long periods of time and over various periods of time (days, weeks, quarters, or months) enables IT profiles to be characterized. These profiles provide information that is useful in the following ways:

Business cycles and the performance and capacity of IT systems all change over time. The good news for IT performance managers is that these changes can be analyzed and reported on over different time periods. This information enables IT performance managers to make the correct and best IT business decisions. For example, performance measurements of a particular resource can be analyzed over time to assess its performance, utilization, and availability. From this assessment, the IT response to business events, economic drivers, or natural weather occurrences can be analyzed. If necessary, adjustments to the configuration of IT enterprise can be identified to accommodate these occurrences in the future.

In the following example, the CPU utilization of a particular UNIX machine is being evaluated. SAS IT Resource Management can obtain UNIX data through the following adapters:

This chart shows the CPU Busy activity of a machine for the hours of the day over the course of a week. Analysis of this type facilitates load balancing and redistribution of workload activities as well as the overall capacity planning function.

An IT performance data warehouse benefits UNIX environments because it is the single best method to accommodate the many resource, enterprise systems management, and freeware monitoring tools that are used to measure the performance and manage the availability of these systems.

Comparing the three lines on the following graph provides a perspective that shows how seemingly small variations in the maximum and mean memory measurements for a particular UNIX device relate to the paging rates that are associated with that device. This long-term (monthly) analysis can prompt more detailed memory analysis activity to ensure that the system is configured with an adequate quantity of memory for the processing allocated to this device.

Horizontal bar charts show the relative magnitude of data by displaying horizontal bars of varying length.

Horizontal bar charts illustrate the comparative volume of a measurement, typically over a time period. From a performance management and capacity planning perspective, bar charts are used for comparative analysis and to get a good perspective on the measurements being analyzed. You might consult these charts to prioritize issues or to quantify their severity.

The one element of the IT infrastructure that is common across large portions of the enterprise is the network. The best source of information that can be used to measure the performance and to assess the health of that network is a repository of Simple Network Management Protocol (SNMP) data.

The preceding chart shows the network interfaces that have the highest incoming and outgoing bit rates over the course of a day. Charts of this type establish an understanding of what is normal and expected activity with respect to network traffic. Deviations to measurements of this type and the trends associated with them should be more thoroughly analyzed.

SAS IT Resource Management provides a reference table that identifies the type of devices that are charted. (For example, in the preceding chart, the reference table was used to determine that the device measurements pertained to the traffic on a Type1 device.)

For information about this lookup table, see “Working with the ITMS_SNMP_DeviceType Lookup Table” topic in the “Accessing Raw Data for SNMP” section of Appendix 2, “Data Sources Supported by SAS IT Resource Management Adapters” in the SAS IT Resource Management: Administator's Guide.

SNMP measurements are essential for understanding the status and components of a network and the devices that are serviced by the network. These measurements can show the status of the network's routers, hubs, and switches as well as the status of the network's links and interfaces. Network administrators and performance analysts must be able to understand the behavior and error rates that are associated with network devices over time. In that way, they can identify and replace faulty hardware components that contribute to the overall degradation of the network. SNMP also communicates the utilization status and highlights instances where thresholds that were established for that utilization have been exceeded. All of this information, and its analysis, is used to best allocate traffic to the network and to tune its flow across individual network components.

SAS IT Resource Management enables you to import MIB definitions that can be used to create the metadata for SNMP staged tables.

Prior releases of SAS IT Resource Management used the repository of SNMP data that was created by the HP Network Node Manager and the Tivoli NetView snmpColDump utility. Newer releases of the HP and Tivoli network management system products do not support the snmpColDump utility. Accordingly, SAS IT Resource Management now supports data collection of SNMP data from a round-robin database to which the data is written. As a convenience for our customers, SAS IT Resource Management continues to support the collection of SNMP data from the repository created by the HP and Tivoli snmpColDump utility.

For information about the support that is provided for these two data sources, see the “Accessing Raw Data for SNMP” section of Appendix 2, “Data Sources Supported by SAS IT Resource Management Adapters” in the SAS IT Resource Management: Administator's Guide.

Bubble plots display information from several perspectives to illustrate the relationships of those perspectives.

In the preceding display, Run Queue Length measurements from Microsoft System Center Operations Manager for a particular machine is evaluated for each day of the month and each hour of the day. This analysis enables performance managers to understand when Run Queue Lengths are at their peak so that systems can be sized or work can be prioritized and allocated or reallocated to alternate devices.