K-T
> I think the average number of files contributing to a subchunk
> might be 1, or at most a few,
indeed :)
> unless the subchunks are a lot bigger than I thought.
We expect few hundred subchunks per chunk.
And we said ~1 sq deg per chunk
> And are all the files open at all times? Or only when a table
> is accessed?
Only when table is accessed. But if we start dealing with a very
large number of files, the sheer number of files in the file system
will start to become scary (the few files-per-subchunk will help though)
> Is it ridiculous in terms of overhead or scheduling or something
> else to have one MySQL server (and hence process) per chunk?
There is a fixed overhead for things like connection, checking access
control, opening schema information, etc. I am not sure if we want
to pay that overhead. Imagine "select count(*) from Object" - we'd
be making 20K connections, checking authorization 20K times, and
opening schema 20K times to do that one query.
Jacek
########################################################################
Use REPLY-ALL to reply to list
To unsubscribe from the QSERV-L list, click the following link:
https://listserv.slac.stanford.edu/cgi-bin/wa?SUBED1=QSERV-L&A=1
|