[Thread Prev][Thread Next][Index]

Re: [ferret_users] THREDDS server, where is the main work done?



Hi Martin,
I asked the question on the netcdfgroup email list, and received these answers:
 On 9/6/2012 6:18 PM, Roy Mendelssohn wrote:
 Hi Ansley:

 I believe the answer is "yes", sort of.  Remember, THREDDS unless you are using the netcdf subset service or another experimental service John C. is working on, does not deliver netcdf files, it delivers data through OPeNDAP, which does not have this compression as part of the description of the protocol.  So the data must be uncompressed and translated to be sent.  however, many OpeNDAP servers, and I believe this is true of TDS, if the client says it can receive zipped files, will zip the OPeNDAP stream and then that will be decompressed when it reaches the client.  One question we (ERD) has been experimenting with is the read penalty with using netcdf4 files in conjunction with OPeNDAP, and what chunking/compression settings have the best tradeoff between performance and space

 I am probably off on some of this, but I believe it is basically correct.

 -Roy
---
On 9/7/2012 10:08 AM, John Caron wrote:
 Correct, thanks Roy.

 Anything thats part of the file format is handled on the server. Standard HTTP deflate (zip) compression is used to send data across the wire, if client allows. the ncstream protocol is experimenting with more complicated schemes, but is not ready for general use.

 Figuring out optimal chunking in netcdf4 is now the right game to play. It depends on the subset request, so you want to optimize your common requests.

 the netcdf group would welcome user experience reports and timings. posting to this list lets us all mull over the issues.

 thanks!

 John

There's some general information that might also be of interest in Unidata's documentation on DAP support:http://www.unidata.ucar.edu/software/netcdf/docs/dap_support.html

We link Ferret with NetCDF libraries that have netcdf-4 enabled.

-Ansley



On 9/6/2012 8:34 AM, Martin Schmidt wrote:
Good morning,

I have a question concerning the data flow when data are stored on a THREDDS server.

After opening an opendap connection to a THREDDS server

use "http://......."

i do some operations and write data to a local disk.

let aaa= ....

repeat/l=1:1000 save/append/file=outputfile.nc aaa

I know, data on the THREDDS server are in netcdf-4 format, using compression. My question is, where
the decompression takes place:

- at the THREDDS server and the uncompressed data are delivered over the web,
- at the local machine, the compressed data are delivered over the web and are uncompressed at the local machine?

Greetings,
Martin Schmidt


[Thread Prev][Thread Next][Index]
Contact Us
Dept of Commerce / NOAA / OAR / PMEL / Ferret

Privacy Policy | Disclaimer | Accessibility Statement