db-derby-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From <de...@segel.com>
Subject RE: DB Tables on separate hardisks
Date Fri, 27 Jan 2006 18:09:13 GMT
Yeah,
That is exactly what I was thinking. However, sometimes databases get flakey
when you try to use symbolic links. At least this was a dirty trick back in
the Informix Standard Engine Days. However, if Derby is using the standard
Java IO, then you shouldn't have a problem with symbolic links.

But you're going to have a headache every time you want to configure or
modify the database. (adding, dropping and even altering a table can be
problematic.)

One has to ask why they would want to do this? The only thing I could think
of is that you may have some sort of file system constraint and your
database is going beyond it. Even then, Derby wouldn't be the correct
database to use.

Performance? (use different file systems located on different disks?)
That would also be an indication that you're using the wrong product... of
course I'm assuming that you've already gone through a series of code
reviews and tuned the engine to be its most efficient.

While I haven't started hitting the scotch yet, has anyone tried running
Derby from a networked drive? (NFS for example?) What would happen if you
had derby on a public drive that was networked from two different machines
each trying to run Derby in either in embedded or network mode? 

Not that I recommend it, just trying to see where and how it would fail.
(Assuming system lock management would die, or file io ...)

Of course the other idea of running each table in a separate instance may
have merit. You could in theory try to create a Derby MPP database but again
you'd have a lot of work to do.

Oh well, lunch break over. Back to real work.

-G


-----Original Message-----
From: David W. Van Couvering [mailto:David.Vancouvering@Sun.COM] 
Sent: Friday, January 27, 2006 11:26 AM
To: Derby Discussion
Subject: Re: DB Tables on separate hardisks

Well, it's tricky to do symbolic links since Derby creates the tables 
within a directory.  I guess what you could do is create the tables, 
shut down Derby, move the table(s) to a different disk device, create a 
soft link, and start Derby up again.  Hey, that might work! (ever the 
optimist)

David

Michael Segel wrote:
> On Thursday 26 January 2006 3:19 pm, Rajesh Kartha wrote:
> 
>>Only the database logs can be put on a separate device using the
>>'logDevice' attribute
>>
>>http://db.apache.org/derby/docs/10.1/adminguide/cadminlog98.html
>>
>>Please refer the Derby Admin Guide for more information on this.
>>
>>-Rajesh
>>
> 
> Well....
> 
> If you wanted to do something down and dirty, you could always have a
couple 
> of different instances running and then connect to the correct instance to

> write the table out and have each instance point to a different directory.
> (But that's a lot of headache and its going to be a nightmare to
maintain.)
> 
> I guess there may be one other option, that might work.
> 
> Has anyone tried doing either symbolic links or hard links? How will Derby

> handle this? In theory, it should work....
> 
> 
>>On 1/26/06, David W. Van Couvering <David.Vancouvering@sun.com> wrote:
>>
>>>My understanding is this is not possible right now with Derby.
>>>
>>>David
>>>
>>>Melvin Zamora wrote:
>>>
>>>>Hi Derbies,
>>>>
>>>>Would it be possible to have the PK tables on hardisk-A and FK tables
>>>>on hardisk-B using only one database?
>>>>
>>>>to envision:
>>>>
>>>>CUSTOMER_DERBY_DB
>>>>
>>>>CUSTOMER_TBL {PK} [HD-A], CUSTOMER_PURCHASES_TBL {FK} [HD-B]
>>>>
>>>>-Melvin
>>>>
>>>>-----------------------------------------------------------------------
>>>>-
>>>>
>>>>Do you Yahoo!?
>>>>With a free 1 GB, there's more in store with Yahoo! Mail.
>>>><http://us.rd.yahoo.com/mail_us/taglines/mailstorage/*http://mail.yahoo
>>>>.com/
> 
> 



Mime
View raw message