Re: Hot Standby (or Failover) support.

In article <3atped$1unp_at_steamer.clam.com>, Robert Metcalf <bobmet_at_steamer.clam.com> wrote:
>In article <89E152B.0389001DA2.uuout_at_compudata.com>,
>BARRY ROOMBERG <barry.roomberg_at_compudata.com> wrote:
>>-> Can anyone tell me what Oracle offers for "Hot Standby" (normally
>>-> automatic) or "Failover" (normally manual) support? I'm particularly
>>-> interested in Sun, HP, DEC(Alpha), IBM RS/6000 and NT machines.
>>
>>I don't know what they "OFFER" for this, but I've created it
>>by putting a backup machine in recovery mode and applying
>>the archived logs as they show up. I did it under SCO Unix,
>>but it should be easy in networked environment.
>
>HACMP/6000 is an IBM LPP for the RS/6000 that enables a
>cluster (2-4 RS/6000s) to provide highly available services to clients,
>with minimal interruption in the event of system or component failure.
>
>You can configure an HACMP cluster and ORACLE in one of two ways:
>concurrent (Oracle Parallel Server) or non-concurrent.
>
>In concurrent mode, an instance of ORACLE runs on all cluster nodes.
>Each node has access to the raw logical volumes. An HACMP daemon
>(cllockd) provides locking services to the ORACLE server, to arbitrate
>requests for the shared, concurrent volume. (File systems can not be
>used in concurrent mode.) If a node fails, client requests are directed
>to the surviving node.
>
>In non-concurrent mode, one node is active and one is a standby.
>When the active node fails, it's applications, disk resources,
>and IP Address can be automatically taken over by a "standby" node.
>Clients experience a brief interruption in service, and then can continue
>processing. If a component (physical volume, communication adapter) fails,
>mirror or standby components step in for the failed original.
>
>--
>Bob Metcalf CLaM Associates 617-621-2542 (v)
> 101 Main Street, 5th Floor 617-252-0820 (f)
>bobmet_at_clam.com Cambridge, MA 02142

Now ask him if anyone in the real world is using this in a production environment -- not yet. The one customer that is scheduled to be using this is still waiting for all the vaporware to flesh out. If they (IBM & CLAM) ever get all the details worked out, this could be pretty slick, but so far all the current guinea pig has gotten is a lot of promises and missed deadlines.

just mho.

-- 
replies to:  hack_at_netcom.com
Disclaimer:  the opinions expressed here do not necessarily
reflect any known standards of thought and may not make any
sense whatsoever...