On 14. Jan 2007, at 10:58 , karl wettin wrote:
In the original post you mention 2-10 million documents. How much
is that is bytes?
On my development machine I have 1.5 million documents and those are
weighing in at
~950MB. I suspect that for production we will add more fields, so it
would be bit bigger
than that. To be able to scale for some time, I think 30-40 GB index
size is the upper limit.
If you can fit it in RAM, I think that http://issues.apache.org/
jira/browse/LUCENE-550
could be your friend. It keeps each document as an instance in
memory. Extend the
InstantiatedIndexDocument class with an association to the user
owning the document.
I wrote the index for this particular case where I due to very
heavy load needed to
retreive some 100,000 aggregated class instances per second, based
on a field value.
Yeah, I have seen that, and it looks awesome! Unfortunately, I think
I cannot use it here,
though it might come in handy for some other problem.
As I understand it, I must create it from a RAMDirectory/FSDirectory,
right?
Thanks,
-k
--
Kay Röpke
http://classdump.org/
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
