fill out consolidation page

Author: copperlight

docs/concepts/consolidation.md

@@ -1,2 +1,40 @@
+## Overview
 
-TODO
+* Consolidation is automatically applied based upon the size of the query window.
+* There is a hard maximum of 1,440 datapoints per line, which is one day of data with a 1-minute step size.
+* There must be at least 1 pixel per datapoint.
+* The unit is preserved across consolidation.
+* The consolidation function will match the aggregation function used (i.e. avg for avg/sum, max for max).
+
+## Further Explanation
+
+* Step size will change based on the amount of data requested.
+    * For short time windows, with graphs that are the default width, you will get a 1m step size up to ~12 hours.
+    * For longer time windows, such as multiple days, the step size can increase to 2m, 5m, 10m, or more.
+    * The step size can change as the graph width changes - this is most often noticed on very wide monitors.
+* One pixel per datapoint.
+    * In addition to the hard limit on datapoints per line, the pixel width is also considered, in order to be able to
+    render one pixel per data point.
+    * The maximum number of datapoints for a line is the smaller of 1,440 and number of pixels that can be visualized.
+    * This can be demonstrated by viewing a metric for the last day:
+        * If you maximize the graph canvas, the Step size switches to 1m.
+        * If you minimize the graph canvas, the Step size switches to 5m.
+* Unit is preserved.
+    * The y-axis unit is preserved for any step size changes.
+* Consolidation function matches aggregation function.
+    * Let’s say you are looking at a graph of Stream Starts per Second (SPS), which is reported by thousands of instances.
+    * This is typically done with a sum aggregation, because you want to see the total value across the fleet.
+    * If you look at this data across a three hour window with a default canvas size, then it will be a 1-minute step size,
+    and no consolidation is applied.
+    * If you expand the time range to 24 hours, then the amount of data selected is now large enough that it must be
+    combined, in order to graph it.
+    * The step size changes to 5-minute, and a consolidation function is applied to combine multiple minutes of data
+    into a single point on the graph.
+    * Since the aggregation is a sum in this case, the consolidation is an average. This means that each point on the graph
+    is now the average value of the five minutes of data from the previous graph across the same time range.
+    * If you had used a max aggregation, then a max consolidation would be applied. The consolidation function is automatically
+    chosen on your behalf, to provide the most correct value on the graph.
+    * As long as your graph lines are continuous, you should not notice any significant changes due to the application of the
+    consolidation function.
+    * If your graph lines are sparsely populated, with many zero or NaN values, then you will see significant changes due to
+    consolidation, because zero or low values will pull down an average.