Tools and Utilities - Professional NoSQL - page 313

DS is a keyword that stands for data source . DS is essentially what I have been calling metric so far.
variable_name identifi es the data source. In the example, cpu is a variable name for holding the
CPU utilization value. data_source_type defi nes the type of value stored for a data source. The
value in this example is GAUGE . The possible values for a data_source_type are as follows:

COUNTER — Records the rate of change over a period. The values in this case are always
increasing.

➤

DERIVE — Similar to a COUNTER but can accept negative values.
ABSOLUTE — Also records rate of change but the current value stored is always in relation to
the last value. The current value is different from the last value. In math terms, it's always
the “delta.”

➤

➤

GAUGE — Records actual value and not rate of change.
RRDTool records values at a defi ned interval. In the example, myrrddb.rrd would expect a CPU
utilization value to be available every 60 seconds. The RRDTool database, unlike an RDBMS, expects
a value to be made available at a predefi ned interval. This means that if it doesn't get a value, it records
it as UNDEFINED . The heartbeat value, which in the example is 120 seconds, is when the database thinks
the value is not present and then records it as UNDEFINED . If values don't come exactly as defi ned,
RRDTool has the capability to interpolate values if the record still arrives within the heartbeat interval.
The last two values, min and max, are boundary conditions for values. Data source values outside
these values are recorded as UNDEFINED . In the example, I assume the CPU utilization is a percentage
utilization value and therefore 0 and 100 mark the boundary conditions for such a measure.

➤

The last two lines depict the aggregation functions on the time series data. In the database create
statement in the example, the last two lines are as follows:

RRA:AVERAGE:0.5:60:24 \
RRA:AVERAGE:0.5:1440:31

RRA , like DS , is another keyword. RRA stands for Round Robin Archive. The RRA defi nitions follow
this format:

RRA:consolidation_function:xff:step:rows

consolidation_function is an aggregation function. AVERAGE , MINIMUM , MAXIMUM, and LAST could
be possible consolidation_function values. In the example, two RRA defi nitions are included.
Both average data points. Consolidation functions operate on the values captured from the data
source. Therefore, in the example of CPU utilization, RRA values will be aggregates of the per-minute
CPU utilization recordings. step defi nes the aggregation bundle and rows specifi es the number of
aggregated records to be saved. In the example, a value of 60 for steps implies that the average
is calculated on the basis of 60 data points of the data source recordings. The recordings are every
minute so this means the averages are for every hour, because an hour has 60 minutes. The number of
rows to be archived is 24. Therefore, the fi rst RRA records average CPU utilization on a per-hour basis
and keeps records for a day.

The second RRA defi nition is an average CPU utilization for a day, and 31 days (or a month's worth)
of data is stored for this consolidation function.