> we are generating lots of relatively small (from a data content standpoint)
> NetCDF files (approximately 800 kilobits mostly spread across two variables),
> and sending them over a network. Monitoring this message flow reveals
> substantially higher bandwidth than anticipated.
>
> We are thinking the remaining size contribution is due to the fixed size
> overhead of a NetCDFv4 file.
>
> What is a rough order magnitude size overhead value for a NetCDFv4 file
> containing about 5 dimension declarations, 13 variables, each with about 10
> attributes each, and 15 global attributes. (this may be wy more info than you
> need)
>
So I did the following experiment.
I created the attached size.cdl file.
I then did this command to generate the binary netcdf-4 file size.nc.
ncgen -k 4 -b size.cdl
I then looked at the file size (using "ls -l size.nc").
This gives the basic metadata overhead for the file.
In this case about 46k bytes. Note that because I set
_NoFill for each variable, the file will contain no data.
You can verify this by changing the size of d4 to see if the
total file size changes (it should not).
Just for reference, if I do the same for a completely empty file
netcdf size0 {}
The file size is 360 bytes, which is more or less the minimum
overhead.
=Dennis Heimbigner
Unidata
Ticket Details
===================
Ticket ID: LAJ-287331
Department: Support netCDF Java
Priority: Normal
Status: Open
===================
NOTE: All email exchanges with Unidata User Support are recorded in the Unidata
inquiry tracking system and then made publicly available through the web. If
you do not want to have your interactions made available in this way, you must
let us know in each email you send to us.
Attachment:
size.cdl
Description: Binary data