Michael Schäfer wrote On 03/27/06 20:24,:

>hello,
>
>is there any kind of workaround to get beyond the "1 million files per 
>terabyte"-limitation for volumes larger than 1TB?
>  
>
No.

>I'm running a Helios fileserver for mac-clients which creates a small metadata 
>file for each file. Currently we have 1TB partitions with about 2.5 million 
>files, and we are running out of space. I wouldn't mind long filesystem checks 
>(in fact I would, but I really have no other choice), but I simply need 
>partitions with 2TB each.
>
>Why is it, that I don't have any choice to set the inode density myself? (And 
>see what I get for it ;-)
>  
>
because your usage pattern is not what we had in mind for MTBufs.

As the target for > 1TB UFS was intended to be databases and 
specifically Oracle the nbpi was set at 1MByte (which at the time we 
considered quite low) MTBUFS was never intended to be used by most 
customers, only databases and other users that require a few very large 
files - another reason is that some 3rd party backup tools had a 2TB 
limit for their output files so only applications that had their own 
backup technology could sensibly use MTBUFS and fully protect their data.

The expectation at the time was that MTBUFS would be a short term fix 
to a specific problem (I.e. Oracle) and that ZFS would deal with the 
more general case - fortunately it will allow very large numbers of files and 
not 
require an fsck equivalent so after a system crash it will come back 
immediately.

Remember MTBUFS was *not* intended for the average user, it *was* 
intended for Oracle and a few other customers who wanted a few very 
large files, e.g. Oil industry data and possibly multi media (though 
they often use QFS)---

frankB


_______________________________________________
ufs-discuss mailing list
[email protected]

Reply via email to