A user has a large 3D variable, about 17.5 GB, stored with time varying most slowly:

netcdf cvx3 {
dimensions:
	latitude = 1617 ;
	longitude = 1596 ;
	time = 1698 ;
variables:
 ...
	float var(time, latitude, longitude) ;
 ...
}

• Accessing all data for a specific time (10.3 MB) is fast (0.25 sec)
• Accessing all data for a specific location (6792 bytes) is very slow: (10.3 sec)
• Users want to access data both ways
• Data currently stored as netCDF classic file contiguously.
• Possible solutions:
  • Rewrite data with time dimensions varying fastest:

    	float var(latitude, longitude, time) ;
        

    But then accessing data at specific time is very slow.
• Keep multiple copies of data for different forms of access. Takes too much space, can lead to update anomalies.

2011 Unidata NetCDF Workshop > Example of the Importance of the Chunk Cache