Re: block chg tracking
Date: Mon, 16 Feb 2015 10:16:44 -0500
Message-ID: <54E209DC.9020709_at_yahoo.com>
On 02/16/2015 09:00 AM, Jeremy Schneider wrote:
> Obviously, reducing reads does matter (and RTO matters too). Running
> full backups every four to six hours on production won't fly at many
> places - and increasingly,
Every four to six hours? Full backup is usually run on daily basis. And even every 4 to 6 hours is possible, if the backup is SAN snapshot.
> the "wee hours of the morning" are another
> global customer's business day.
That is true. That's why a lot of attention needs to be devoted for scheduling, unless the full backup is actually a snapshot.
> In fact I've heard CIO's mandate that
> no backups are allowed from the primary production database at all
> (even weekends), after very messy performance-related support cases
> with their customers (internal or external). Sometimes it's
> worthwhile to remove this factor so people can't blame it, even when
> you know it's not the root problem. The common solution I've seen for
> that situation is offloading backups to a physical standby - as Kenny
> discussed. And in that situation, you can still leverage pretty much
> any strategy under the sun including a backup system that supports
> dedupe.
That is true, but what Kenny described is not a typical solution. Kenny described the snapshot-only solution which doesn't include long term data preservation, like the one mandated by SOX or HIPAA. Those solutions can be combined. Also, please note that snapshots are fast. Running a full DB backup as snapshot every 4 to 6 hours is not a problem. Note that once you have snapshot, copying the files to tape can be done using file system utilities like tar, cpio or rm (if 100% compression is needed). The only remaining problem is the one of cataloging these backups and modern backup utilities do that for you.
-- Mladen Gogala Oracle DBA http://mgogala.freehostia.com -- http://www.freelists.org/webpage/oracle-lReceived on Mon Feb 16 2015 - 16:16:44 CET