Hi

I read that BDB supports databases up to 256 TB in size.

I have a customer requirement to store about 10 TB of data.
Considering using BDB. Not very knowledgable on mass storage options.
I would really like a simple Linux solution.

Thinking of spreading this across multiple (up to 20) IDE drives on
Linux.
I can manage creating the multiple small databases myself. Any
opinions ?

But my real question is : Somebody has told me that Storage Array
Networks are the way to go.Will BDB work on this ?? I read in the docs
that BDB needs to memory map the file and I would imagine this couldnt
work with SANs.

If anybody knows this or can suggest an alternative, that would be
great.

Thanks
Chad

Chad,

BerkeleyDB only needs to memory map what we call the environment region files.
These hold data which is shared amoung multiple processes accessing
the same databases. Please see:
http://www.sleepycat.com/docs/ref/env/region.html
Two things to note, first these files need not be in the same file
system as actual data, see:
http://www.sleepycat.com/docs/api_cx..._data_dir.html
Second, system memory can be used via shmget(2) instead of mmap if this is
not available at all:
http://www.sleepycat.com/docs/api_cx...#DB_SYSTEM_MEM

BerkeleyDB data storage should work on a SAN so long as it conforms
to the POSIX file system implimentation. You should be careful of
such a device that caches disk writes and does not write them through
to stable storage when the operating system is told to fsync(2) the file
as this can lead to data corruption after a system failure.

Michael Ubell
Sleepycat Software.

Ps. region.html contains a bug. Bullet 3. should say:
"If mmap is not available, the UNIX shmget interfaces MAY be used, ...."