RE: Bandwidth management for DC migration
- From: "CRISLER, JON A" <JC1706@xxxxxxx>
- To: "hemantkchitale@xxxxxxxxx" <hemantkchitale@xxxxxxxxx>, ORACLE-L <oracle-l@xxxxxxxxxxxxx>
- Date: Fri, 3 Feb 2012 23:54:51 +0000
If you are on 11gR2, there is new feature, sometimes considered undocumented,
to allow for dynamic compression of archive logs.
This will save some bandwidth for Dataguard configurations, but I don't think
it helps for the standby redo logs and Real Time apply.
But it can certainly save some bandwidth.
If you run a local RMAN, using compressed backupsets, the resulting backup will
give you an idea of what size has to be moved to the new data center.
Just make sure your backup clone uses compressed backupsets. Based on past
experience I would guess the resulting rman backup to be in the 200-300gb
range, but this is entirely data dependent: some data compresses better than
others. I once had a db that was 95% mpeg and jpeg and the resulting
compressed RMAN backup was more or less the size of the raw datafiles, as mpeg
and jpeg are already compressed and rman could not compressed them any further.
Assuming that you are using real-time apply, which (I am guessing) doubles the
amount of data transferred, you are now looking at 60gb over 4 hours, or 15gb
per hour, or 4.16 megabytes per second (15,000 mb / 3600 ).
Based on 4.16 MB per sec you are looking for at least OC-1 size (which is
roughly 5 MBps), and at that you will be running about 100% capacity at peak
loads. OC-3 is going to give you around 15 MBps rates so that might be
overkill. You might have some other offerings in the MPLS or older DS-3
available. There are also latency issues to consider- this is definitely not
a case where you want to run DataGuard in Max Protection mode :) You don't say
where in Asia but I would not be surprised to see 100ms + latency. Also, keep
in mind that there are long lead times to install this sort of bandwidth,
unless you already have a datacenter where you are sharing a high-capacity line
like OC-12, OC-48 etc. And the monthly cost for this sort of link - wow, it
will be expensive.
-----Original Message-----
From: oracle-l-bounce@xxxxxxxxxxxxx [mailto:oracle-l-bounce@xxxxxxxxxxxxx] On
Behalf Of Hemant K Chitale
Sent: Friday, February 03, 2012 6:16 AM
To: ORACLE-L
Subject: Bandwidth management for DC migration
I am in a Data Centre Migration project where we are moving ebiz data and
apps from a US data centre to Asia. We have DataGuard configured but the
bandwith is a constraint when handling the daily spike in redo volumes.
What options (software / hardware / business / users / concurrent requests
) do you explore? Typically what bandwidth is used to clone a 600GB
database and daily volumes of 65GB of redo (spike if 30GB in a 4-hour peak
window) ?
Hemant K Chitale
sent from my smartphone
--
//www.freelists.org/webpage/oracle-l
--
//www.freelists.org/webpage/oracle-l
Other related posts:
