NeophyteThere is consolidation and aggregation, then there is domain aggregation:
Consolidation is where data is combined along the x-axis. For example, if your data is obtained every 1 minute, displaying that graph over 3 months would result in roughly 130,000 points along the time axis (x-axis) per instance. The consolidation function of a graph's datapoint determines how that large number of datapoints is simplified to make the graph display-able and readable. Choosing average as the consolidation means that the graphing engine will combine roughly 5 hours worth of data into a single point and graph that one point, representing the entire 5 hour timeframe.
Aggregation is where multiple instances are combined together vertically according to a function. You would do this if you wanted to know the total bytes of storage free across all instances, for example. You would enable aggregation and choose "sum" as the aggregation method. If you wanted to know the average CPU across all CPUs included in the graph (according to your filters), you would enable aggregation and choose the "average" aggregation method.
Domain aggregation is where you use a statistical formula to describe a single trend line using a single descriptor. You can create a virtual datapoint that uses domain aggregation functions. These functions take into consideration all data currently displayed, so they're calculated at runtime. While the linked page talks about complex datapoints, the percentile functions can only be used in virtual datapoints (not complex datapoints).
