In article <8sklos$rb4$1 at snic.vub.ac.be>,
Guy Bottu <gbottu at bigben.vub.ac.be> wrote:
> Dear colleages,
>>We are busy installing an SRS 6 server on a PC with Linux. We hit on a
>problem : with the last release of EMBL the size of some indexes grew
>beyond the limit of 2 Gbyte (not supported by a conventional file system
>on 32 bit machine). More in particular, it are the FeatureDescription
>and the Authors indexes who became too big, although I suspect that
>Authors is abnormally large because of some bug.
>>Therefore this question : Has anyone already succeeded in making SRS 6
>work on a PC under LINUX with index files bigger than the limit ?
Well, you could split your data into smaller sub-databases, and then
modify your queries to search them together. That would work right now.
Work is in progress to support larger file sizes on Linux, but I think
most of it needs to wait for the 2.4.0 kernel, which includes a number
of new filesystems, such as ext3. I don't know whether ext3 is 64-bit,
but I have a feeling it might be.
You might also consider looking into SGI's xfs filesystem, which they
have donated to the Open Source community. This may well support 64-bit
files (after all, it does on IRIX, but then the MIPS processors on IRIX
boxes are 64-bit). I believe there are patches available to add the xfs
filesystem to the Linux 2.2.x kernel series.
You need to be confident about patching and compiling development
kernels to try this route though...