Asked this question via Google but didn't realize it wasn't
making it to the mailing list (I hadn't subscribed via the
postgres website).
=====
Hello all,
Do the Postgres large object functions support non-blocking operation?
I'm working on a single-threaded, event-driven server that needs to
store incoming speech from a large number of clients and async LO
operations would come in really handy (at least this seems to me like
a good approach). Unfortunately the docs only explicitly mention support
for PQputline, PQputnbytes, PQsendQuery and PQendcopy. I'm researching
a move to Postgres and would like to know it it can handle this or if I should
look at different techniques(*).
Regards,
Iker
(*) I could also buffer all the speech for each client and save it into a
single bytea column (asynchronously) but then I have to take the memory
hit and the CPU hit for escaping the rather large strings (most speech
segments will be around 30K but some could go as high as 200K - for
several hundred concurrent clients at peak this could be an issue). Alternatively,
I could use multiple bytea rows and do my own chunking but this still incurs
the escaping hit and it seems like a lot more work. Are there any other
techniques I can consider?