Batch Thread Analytics

The Batch Thread Analytics portal illustrates a set of metrics aligned with execution of batch processes at the thread. This portal is useful for summarization of the state and work performed by the threads for a batch control over a selected date range. This portal is useful for jobs that are run multi-threaded.

Note: To use these analytics, you must populate the batch snapshot must be populated, on a regular basis, using the batch job F1BTANSN - Batch Thread Analytics Snapshot Update..

Note: Not all zones in this portal support broadcast as it is not appropriate in all situations. Unless indicated assume broadcast is not available from the zone.

Note: At the top of each tab in this portal, the selected batch control and date range is shown.

Batch Thread Analytics Search

To use this portal you must first select the batch control and the period you want to analyze using the Batch Thread Analytics search from the search dropdown. To help find the necessary batch control, you can use a variety of filters to refine your search.

Note: Date Range is required At a minimum a date range must be specified. This must be either the predefined period or a custom date range. For custom date ranges, it is recommended to limit the range to analyze less than a thousand (1000) values for performance and readability reasons.

Once the appropriate filters are populated a list of batch controls is returned with the number of executions and the maximum thread limit used in the snapshot for the specified date range and other filters specified. To start the analysis, select the desired batch control from the list.

Batch Thread Analytics - Main

The Main section of the analytics summarizes the state, submission source, performance, and volumes processed by the desired batch control over a date range. This includes analysis in the following areas:

The Batch Thread Status zone summarizes the state of all executions of the desired batch control over the desired date range within the snapshot. In an ideal situation this graph should show a high percentage of complete (successful) executions.
The Batch Thread Reliability Summary zone calculates the total reliability rating for the desired batch control over the desired date range within the snapshot.
The Batch Thread Limit History zone shows the thread limit specified for each batch run for the desired batch control over the desired date range within the snapshot. Ideally the thread limit used should be consistent. Large deviations in values may be due to volume or inefficiencies.

Batch Thread Analytics - Reliability

The Reliability section of the analytics focuses on the reliability of individual job runs for the desired batch control over the desired date range within the snapshot.

The analysis of threads introduces a reliability rating related to the success rate of threads for an execution. The percentage of successful threads over total thread count for each execution is used for the rating using the following criteria:

If all threads (100%) are successful for a batch run then the rating is High.
If the thread failure rate is less or equal to 1% of the total number of threads then the rating is Good.
If the thread failure rate is less or equal to 5% of the total number of threads then the rating is Fair.
If the thread failure rate is above 5% of the total number of threads then the rating is Poor.

The reliability analysis includes analysis in the following areas:

The Batch Thread Reliability History zone shows the reliability rating for each run for the desired batch control over the desired date range within the snapshot. The color and size of the entries in the graph illustrate the reliability and percentage of successful threads.
The Batch Thread Reliability Details zone lists the individual batch runs with their associated reliability rating for the desired batch control over the desired date range within the snapshot. You can navigate to the Batch Run Portal for additional information.

Batch Thread Analytics - Differential

The Differential section of the analytics shows analysis of submission information periods applicable to the desired batch control over the desired date range within the snapshot.

This includes analysis in the following areas:

The Batch Thread Time Differentials zone shows the difference between start and end time for the desired batch control over the desired date range within the snapshot as expressed as a wost case differential. Each differential is shown in seconds. If threads start at the same time and end at the same time then the differential is zero (0). Positive differentials mean differences in start and end times of threads.
The Batch Thread Volume Differentials zone shows the difference between the smallest volume and largest volume for the threads within each run against the average volume. Each differential is shown as number of records that are different compared to the average volume. A reference line is shown to show the relative difference in low and end volumes. In an ideal situation each thread should work on the same amount of records. Variations means variance in processing volume and may result in different thread elapsed times.

Batch Thread Analytics - Retries

Note: If this zone is blank, then there were no automatic retries for the desired batch control over the desired date range within the snapshot.

The Retries section of the analytics lists the total retries per batch run. If the number of retries is above zero, in a consistent way, then there may be some underlying issues that need investigation.

Note: Retries are only performed for specific conditions. Refer to the cloud service documentation for additional information.

Batch Thread Analytics - Zero Processing

Note: If this zone is blank, then there were no zero processing threads found for the desired batch control over the desired date range within the snapshot.

While most threads have records to process, it is possible that a thread is allocated no work to do. This is inefficiency as resources are allocated to start an end the thread at a minimum. This zone highlights the number of threads per batch run with zero processing for the desired batch control over the desired date range within the snapshot. If the number is consistently high then consider altering the thread limits or thread numbers to re-balance the workload.

Batch Thread Analytics - Interquartile Range Analysis

Interquartile analysis is a statistical method used to understand the spread and distribution of a set of data by focusing on the middle 50% of values. By dividing the data into four equal parts, it helps identify the range where most data points fall, making it easier to spot any unusual or extreme values (outliers). This approach is helpful for quickly assessing data consistency and detecting trends, enabling you to make more informed decisions based on typical, rather than exceptional, results.

The Batch Thread Inter Quartile Analysis zone shows the inter-quartile analysis for threads across batch runs for the desired batch control over the desired date range within the snapshot.

To understand the analysis there are different metrics calculated to provide reference:

The median quartile represents the median between the worst and best thread elapsed time. This metric is calculated to provide a reference to assess other metrics against. In an ideal situation all the threads would have this elapsed time.
The upper quartile whisker and lower upper quartile whisker are calculated at the 75% and 25% confidence interval, respectively, from the median. These whiskers represent the acceptable extremes for elapsed times of threads. There proximity to the median illustrates whether the threads are finishing
The high within quartile and low within quartile calculates the thread elapsed time within the whiskers. The proximity of these metrics to the whiskers indicates if threads are ending early and/or late consistently.
The worst thread elapsed time and best thread elapsed time are used as reference for all the above metrics. If these values are above their respective whiskers then they are consider outliers
The difference between the whispers is known as the Inter Quartile Range (IQR). The higher the IQR value the more the variation in the thread elapsed times. If these is little variation, then these values are close to the median.

It is divided into three types of analysis, as different graphs on the zone:

The first analysis is focusing on the upper quartile against the worst performing threads. The relationships between these metrics will illustrate the extent of any variation in the threading and the potential presence of threads that overrun.
The second analysis is focusing on the lower quartile against the best performing threads. The relationships between these metrics will illustrate the extent of any variation in the threading and the potential presence of threads that finish too early.
The last analysis is focusing on the IQR value. If you see a small IQR, it means thread execution times are consistent. A large IQR suggests there is significant variation.