search cancel

SYSVIEW | Comparing Threshold Alerts with RMF Data

book

Article ID: 253679

calendar_today

Updated On:

Products

SYSVIEW Performance Management

Issue/Introduction

Customers may notice that captured RMF data doesn't reflect CPU% threshold alert.

Environment

Release: 17.0; 16.0; 15.0

Cause

The absence of documentation related to an alert event - in this case, a CPU% spike, is related to how RMF and SYSVIEW gather and present captured RMF records.

Resolution

Factors that Lead to Undocumented Events

  • SYSVIEW panels use either "REALTIME" or "INTERVAL" data, with some using both. REALTIME data is just that - data that is gathered and updated when a user refreshes the panel; INTERVAL data is gathered from a saved data collector. SYSVIEW presents collected RMF data based on the RMF data collection intervals.

  • The RMF CPC Capacity Report displays INTERVAL data; that value is displayed using the RANGE parameter. CPU spikes that resolve before the RANGE time and RMF data collection intervals will be witnessed by users but not captured on the report.

  • Finally, we need to consider how SYSVIEW presents RMF data. Depending on the panel, SYSVIEW can present either REALTIME or INTERVAL data. Looking at panels related to RMF, panels RMCAP and CPCGROUP present INTERVAL data in a SUMMARY format - gathered using RMF values with some presenting data from 4-hour averages. These displays, along with RMMSU, allow users to "drill down" to view records in more granular detail. When viewing these, the SrvUTime field identifies the elapsed time between transactions. Looking at a 2-hour window, we see the following in SYSVIEW:

Intv Time       SrvUnits SrvUTime Unused 1hMSU Pct%
  17 11:59:22   12551664 00:05:00    596   151  20%
  29 10:59:21   26480624 00:05:00    429   318  43%
  41 09:59:21   16497472 00:05:00    549   198  27%

In RMF, we do see a discrepancy in values:

Partition  --- MSU ---  Cap     Proc    Logical Util %   - Physical Util % -
             Def   Act  Def      Num    Effect   Total   LPAR  Effect  Total  Time
CA11           0   232  N N N    4.0      46.6    46.8    0.1    31.1   31.2  11.59.00
CA11           0   240  N N N    4.0      48.3    48.4    0.1    32.2   32.3  10.59.00
CA11           0   192  N N N    4.0      38.6    38.7    0.1    25.7   25.8  09.59.00

Are Any Changes Needed to RMF or SYSVIEW?

It's understandable for users to review RMF data when a threshold is triggered; however, the RMF data collection interval might be preventing users from looking at short-term CPU% spikes in the CPC Reports and SYSVIEW displays. This article isn't suggesting altering RMF parameters or not utilizing RMF data but to ensure users are made aware that threshold triggers are a granular action using REALTIME data; something not used in RMF interval-based data collection.

Additional Information