What kind of response time are you getting when you retrieve the full list?
The reason I’m asking is that when Asterisk starts, we pre-populate an internal cache (media_index) with all the sounds which, depending on which languages and formats are installed, could easily be 20,000 files. In some instances, users are creating symlinks inside the sounds language directories to other places (like voicemail) which now will also get indexed. In one case, we had a customer with an effective list of over
800,000 sound files. This takes quite a bit of memory. What we’re trying to determine is the value of the cache considering that we’re only caching directory entries (not contents) which the filesystem has already got cached. In tests of the ari/sounds resource, 95% of the response time is constructing the JSON response, even when not using the cache. I.E. The cache made no appreciable difference in the response time.
2 thoughts on - Who Uses The Ari/sounds Resource?
I’m talking about the “sounds” resource itself. https://wiki.asterisk.org/wiki/display/AST/Asterisk+16+Sounds+REST+API
Thanks Sébastien.
What kind of response time are you getting when you retrieve the full list?
The reason I’m asking is that when Asterisk starts, we pre-populate an internal cache (media_index) with all the sounds which, depending on which languages and formats are installed, could easily be 20,000 files. In some instances, users are creating symlinks inside the sounds language directories to other places (like voicemail) which now will also get indexed. In one case, we had a customer with an effective list of over
800,000 sound files. This takes quite a bit of memory. What we’re trying to determine is the value of the cache considering that we’re only caching directory entries (not contents) which the filesystem has already got cached. In tests of the ari/sounds resource, 95% of the response time is constructing the JSON response, even when not using the cache. I.E. The cache made no appreciable difference in the response time.