Re: LRU latches

"Brian Peasland" <peasland_at_usgs.gov> wrote in message news:3A9A758B.94A69F5D_at_usgs.gov...
> Comments inline:
>
> > of the server or "shadow" processes). Latches control both the LRU list
 for
> > the buffer cache and the redo log buffer.
>
> I didn't think that the redo log buffer worked on a Least Recently Used
> algorithm?!?

It doesn't. The original statement is (to use a Sybrand term, "Hogwash"!)

>The log buffer doesn't work on the same principles as the
> buffer cache so it doesn't need an LRU (or MRU) algorithm does it? While
> there is a latch for the log buffer, there is not a LRU latch to my
> knowledge. Can anyone else clarify this? Or is it just me?

There are latches. The redo allocation latch and the redo log copy latch. Both pretty untunable in 8i.

>
> > Therefore, the system needs to provide exclusive access to the log
 buffer to
> > an individual process (or processes) and this is done with a latch (it's
> > essentially a sephamore).
>
> True, but it's not an LRU latch.
>
> > > Could someone explain LRU latches and the relationship of these to the
> > > LOG_SIMULTANEOUS_COPIES and LOG_SMALL_ENTRY_MAX_SIZE parms? Thanks...
>
> Back to the original post (and if my information is still correct), the
> LRU latch corresponds to the buffer cache which has nothing to do with
> the log buffer. Therefore, the LRU latch and the LOG_* parameters have
> nothing to do with each other.
>

The words "got", "one", "it" and "in" spring to mind. Not necessarily in that order.

Regards
HJR
> HTH,
> Brian
>
>
> --
> ========================================
> Brian Peasland
> Raytheons Systems at
> USGS EROS Data Center
> These opinions are my own and do not
> necessarily reflect the opinions of my
> company!
> ========================================
Received on Tue Feb 27 2001 - 03:17:45 CST