On 2016-06-21 22:29, Alexander Harm wrote:
My advice: DON’T.
I agree with this. We tried using CouchDB for image storage because the
images logically "belonged" to some data entities that were in Couch.
The benefit was that if you deleted an entity, the images would be
deleted too, but the drawbacks were far greater than the benefits. The
most important one was that the database grew so large that syncing it
took so long that it wasn't practical any more. The CouchDB sync
protocol just isn't geared towards moving a very large number of
binaries quickly.
To illustrate how unwieldy it can become, imagine replacing the CouchDB
component with a tar.gz-server that can read and write to a single
tar.gz-archive. You really don't gain anything apart from having
bottlenecks in odd places.
Regards,
Michael.
--
Michael Zedeler
70 25 19 99
[email protected]
dk.linkedin.com/in/mzedeler | twitter.com/mzedeler | github.com/mzedeler
