Manage metrics

Reuse existing metrics on startup and reduce the operational memory overhead

Objective

Provide optional ability to reuse existing metrics on startup and reduce the operational memory overhead for stored metrics.

Example & DSL attributes

This example performs two key metric table functions:

1. Truncates the BidMovingAverage table on process startup providing a clean restart.

2. Executes the default compaction query on a 8 hour schedule for the BidMovingAverage metric.

Example

metrics engine:
  ...
  foreach metric compute:
    metrics:
    - name: BidMovingAverage
    ...  
    #
    # Truncate BidMovingAverage on startup
    #
    truncate on start: true
    
    #
    # Manage metric storage
    #
    compaction policy:
      frequency: 8
      time unit: HOURS

Metric Management Attributes

Each metric definition includes management attributes to control how the metric table is managed throughout the Joule process lifetime. Since Joule uses an in-memory database the size of table needs to be managed.

Attribute	Description	Data Type	Required
truncate on start	Truncate metric data on restart. Note if you import metrics using the initialisation DSL element you will need to set this to false	Boolean Default: true
compaction policy	Execute table compaction on a scheduled cycle. See below.	See compaction section

Compaction Policy Attributes

Managing metrics is a key aspect to the performance of the Joule runtime. By applying the compaction policy this will reduce the memory footprint required of the JVM and therefore the number of GC cycles.

Attribute	Description	Data Type	Required
frequency	Frequency metrics are computed. Minimum value cannot be set below 1 for minute or hour time units.	Long Default: 1
compaction query	User defined compaction query. Note: The default query compacts the metric table by deleting all row less than the max createdTimestamp.	String
time unit	Time unit used to set the scheduled processing policy. Supported units: MINUTES, HOURS	TimeUnit Default: HOURS