Re: [agileDatabases] Re: Using FileSystem for transaction? [Sorry if OT.....]
>I coming from the Java background and have read related material before.
> 1. Grab a copy of Stevens' _Advanced Programming in the Unix
> Environment_ and see if that has enough to help you. If you really want
> to understand the issues and know that you're doing the right thing,
> you also need to have a good understanding of how disk drives work,
> at the block read/write level on both sides of the disk's controller,
> how operating systems schedule IO, and how file systems work. Much of
> that you can get from any detailed book on OS design; _The Design and
> Implementation of the 4.4BSD Operating System_ is one example.
However the working hours of current job don't allow me to have enough
personal time to read such a big book , will you have any suggestion of
>Will see if there really chance about this.
> 2. Hire someone who already knows about this stuff to help you fix this
> problem. (Advertisement: I do exactly this sort of consulting.)
>[Non-text portions of this message have been removed]
> Curt Sampson <cjs@...> +81 90 7737 2974
> Mobile sites and software consulting: http://www.starling-software.com
> Yahoo! Groups Links
- On 2008-06-24 00:02 +0800 (Tue), Carfield Yim wrote:
> I coming from the Java background and have read related material before.You need not read the whole book, just the relevant parts. But honestly,
> However the working hours of current job don't allow me to have enough
> personal time to read such a big book , will you have any suggestion of
> shorter tutorial?
to do this properly you need fairly good knowledge of Unix; you'd really
probably be better off just using PostgreSQL or something like that,
where it's guaranteed to work.
Anyway, one form of atomic operations on filesystems can be learned by
carefully examining how systems that use maildir work.
Curt Sampson <cjs@...> +81 90 7737 2974
Mobile sites and software consulting: http://www.starling-software.com
On Mon, Jun 23, 2008 at 11:58:04PM +0800, Carfield Yim wrote:
> Would you share a bit more why it is particular ineffective in solaris? And
> it is ineffective compare to what? Typical Oracle setup? Sorry to ask silly
> question but I really green in this area.
> On Mon, Jun 23, 2008 at 9:16 PM, Cameron Laird <claird@...> wrote:
> > On Mon, Jun 23, 2008 at 12:05:46AM +0800, Carfield Yim wrote:
> > .
> > .
> > .
> > > Thanks a lot for reply, as the size the message per transaction is fixed,
> > > will that possible work if we allocate large enough buffer? BTW we use
> > just
> > > read() and write() at solaris platform, not POSIX function.
> > .
> > .
> > .
> > Bad idea. It's fine to ask the question. Conventional
> > filesystems under Unix--Solaris, in particular--all make
> > this a particularly ineffective approach. There are
> > better ways to spend your time.
The approach isn't particularly ineffective with Solaris; Solaris
is a particular platform in which the approach is ineffective.
In comparison with what? In comparison with *anything* that works.
Conventional filesystems simply do not have the semantics you're
after. They can be implemented in terms of a RDBMS, or filesystem,
or any other persistence store.
It's OK to be green. This isn't an easy area in which to be green,
though; there are a lot of risks. I believe others have recommended
you read Stevens, and you correctly replied that that's a big
investment to make. I don't know any shortcut.