Re: Tom Kyte Availability Option Questions / OPS newbie

From: Christopher M. Day <christopher.day_at_rdbms.freeserve.co.uk>
Date: Wed, 28 Feb 2001 21:01:04 +0000
Message-ID: <3A9D6710.56353761@rdbms.freeserve.co.uk>

kg,
good points, in warehousing lots of NOLOGGING operations are used.

OPS provides protection from CPU and memory failures. For protection from disk subsystem failures, players such as HDS provide architectures to mitigate the risk of subsystem failure. (Mirrors over distance, multi-path fibre etc)

The use of replication only shines if you implement something like Quest's Shareplex solution...

Closing I would suggest you price each option up first, then find the business requirements for availability and the MTTR (Mean Time To Recover). I think your money would go on mission critical vendor support contracts and StorageTek silos !

Chris

kg wrote:
>
> But remember that a standby is not a complete copy of your original
> database. There are a few (and only a few!) operations that do not get
> recorded in the redo logs (no logging bulk loads, create table... as
> select..., etc )
>
> The dml of these types of operations are not fully recorded in the redo and
> hence are not replicated to the standaby - they also are not recoverable
> from your archived redo logs if you do a restoration.
>
> "Mark Townsend" <markbtownsend_at_home.com> wrote in message
> news:B6B71A4D.3A12%markbtownsend_at_home.com...
> > Oracle Parallel Server allows you to run your single database across two
 or
> > more machines in a cluster. Hence if one machine fails, the other
 machine(s)
> > are still available for use (albeit with twice the workload). I also think
> > that the standby database capability in Oracle8i is a better redundancy
> > solution than replication - replicating a complete database for redundancy
> > is difficult and can introduce significant overhead into your production
> > environment.
> >
> >
> >
> > > From: Doug C <dcowles_at_i84.net>
> > > Organization: Posted via Supernews, http://www.supernews.com
> > > Newsgroups: comp.databases.oracle.server
> > > Date: Mon, 19 Feb 2001 21:27:52 -0500
> > > Subject: Tom Kyte Availability Option Questions / OPS newbie
> > >
> > > I was looking at an article by Thomas Kyte about the various
 availability
> > > options for Oracle, which is at :
> > > http://www.oracle.com/oramag/oracle/00-May/o30tom.html
> > > but i will sumarize the options.
> > > There are 5
> > > 1) OPS
> > > 2) Replication
> > > 3) Standby Database
> > > 4) OS- Oriented solutions - such as dual ported RAID devices.
> > > 5) Architecture
> > >
> > > My question revolves exclusively around this comment he makes:
> >
> > --------------------------------------------------------------------------

---

> > > ----------------------
> > > "In terms of OPS compared with replication:
> > > If failover and availability are the key issues, using OPS wins out in
> > > my opinion. If redundancy is the main issue, then using replication is
 the
> > > answer (because using OPS is not a redundant solution-you have only one
> > > database)."
> >
> > --------------------------------------------------------------------------

---

> > > ----------------------------------
> > > Since we have only one database, and that box could crash, how does this
 come
> > > in
> > > the lead for availability? I can see how one appserver fails over to
> > > another,
> > > but if there is something wrong with you db server, seems like
 availability is
> > > down the drain.
> > >
> > > Comments, insights?
> > > I would pose this to Tom directly, but he is pretty swamped.
> > >
> > > - D
> >
Received on Wed Feb 28 2001 - 15:01:04 CST