Re: Backup Recovery Strategy for Large Database
Date: 8 Feb 1995 17:12:12 -0500
Message-ID: <3hbfjt$4ti_at_news2.delphi.com>
dvick_at_lanier.com (Don Vick) writes:
>In article <3gp6ua$ads_at_yoda.syntex.com>,  <rjen_at_rvax.syntex.com> wrote:
>>
>>In article <D2yuwt.7pB_at_lanier.com>, dvick_at_lanier.com (Don Vick) writes:
>>|>In article <Pine.SV4.3.91.950124081125.1063A-100000_at_gateway.ctg.com>,
>>|>P.E. Detzel <ctg27516_at_ctg.com> wrote:
>>|>>I need to develop and implement a backup/recovery strategy for a large 
>>|>>Oracle application.  This will be a global system and is required to be 
>>|>>   ...
>>|>
>>|>We use this technique for backing up a 20+ gigabyte database that has 
>>|>stringent uptime requirements (though not 24x7 yet).  At an opportune 
>>|>moment, we mirror the disks containing the database, then bring down the 
>>|>database, separate the mirrored drives, and restart the database.  The 
>>|>database is down for less than five minutes, and we have a complete static 
>>|>image that can be backed up to tape later.  Maybe you can adapt this to a
>>|>hot backup strategy.
>>|>
>>
>>I'd be interested to know of any performance degradation when you mirror the
>>disks containing the database.  We have thought of this, but it takes 4 hours
>>or so to fully update the mirror disks, and performance suffers during this
>>period.  We're on VAX/VMS here.
>>
 
>The systems people tell me there is a noticeable performance degradation
>during mirroring.  Operations schedules it during a light-load period, so
>it has not been a problem for us.  This on a Sequent with multiple Pentiums.
 
>Don
>--------------------------------------------------------
>Donald E. Vick  (dvick_at_lanier.com, dvick_at_crl.com)
>Voice: (404) 493-2194    Fax: (404) 493-2399
An expensive but worthwhile strategy for 24x7 shops is to deploy three of mirrors. Therefore, you are always protected with a mirrored pair. The 3rd set can be unsync'd, backed up (perhaps using AT&T's ComVault or similar product) and resync'd with your production mirrored pair. (See ... it really is all done with mirrors !)
George A. Polisner Received on Wed Feb 08 1995 - 23:12:12 CET
