Sripathi Raj wrote:
This is on a NFS on Windows XP Xeon - 2.8 Ghz, 1 Gig RAM and the database
size is 395 MB. I'm connecting to the database from Perl. There is no
discernible difference b/w Perl and sqlite shell.
NFS? Hmm... You have been following the locking threads, haven't you? ;)
How long does it take to read the entire file? Don't forget that network
bandwidth is << than disk bandwidth. The disks in my Samba server may do
>60MB/sec but my 100Mb network restricts this to about 8.5MB/sec.
That's around 50 seconds for a straight read of your 400MB file before
you even do anything with it. If you have to search indexes, go back to
read data etc etc then it's bound to be slow. If your network is at all
loaded the performance will be even worse.
It takes about 5 seconds to copy my 43MB test database to a Samba server
(1GHz Athlon, new SATA2 disks) over 100Mb CAT5. The SQLite shell takes a
good 9 seconds for your count(gid) query. It's 8.5 seconds from Python.
If I repeat the test the time goes down to about 0.4 seconds - XP cached
the data.
If I (roughly) simulate lots of network traffic by listing all the files
on the server the test takes about 28 seconds.
If I "saturate" the network by copying a ~4GB file from the server to my
PC the test takes about 43 seconds.
That's a >100x slowdown without even trying.
Martin
-----------------------------------------------------------------------------
To unsubscribe, send email to [EMAIL PROTECTED]
-----------------------------------------------------------------------------
