Re: Replication vs Data Guard
- From: Leyi Kamus Zhang <kamusis@xxxxxxxxx>
- To: denis.sun@xxxxxxxxx
- Date: Thu, 12 Jan 2012 13:59:48 +0800
Hi Denis
1. If you just want to compare the network traffic, and if not
considering the redo compress function in Data Guard, then replication
will definitely beat Data Guard, since you can choose to only send one
table's modification when using replication, while you have no choice
but have to send the whole redo log when using Data Guard.
2. But ... when you mentioned latency, in my opinion, it is not just
only include the netwrok transfer time, but also include the REDO/SQL
apply time in the target (of course also include the RDO/SQL build
time in the source). The REDO apply technology used in physical data
guard is more efficient than SQL apply used in other replication
method. Try to imagine there's an application always update the same
row over and over again.
So, it's complicated, it's related to your application business.
--
Kamus <kamusis@xxxxxxxxx>
Visit my blog for more : http://www.dbform.com
Join ACOUG: http://www.acoug.org
On Thu, Jan 12, 2012 at 8:17 AM, Denis <denis.sun@xxxxxxxxx> wrote:
> Brandon,
>
> Thanks for sharing your data and all others replying me in the list or
> privately. I am wondering for how your "apply lag" varies depending on redo
> generate rate if any. Will a burst of redo change the "apply lag"? Also I am
> not sure how Oracle calculate this lag. For comparison, I share my latency
> data. In my replication set up we at soure insert a row into a table every
> min and using a trigger at target db to capture target insert time. Source is
> at CA and target is at NY, ie from west to east in US. , then compare source
> insert time and target insert time, in such way we get the latency in second.
> I looked half hour awr for 14:00-14;30, redo is 1,678,257.19 per seond. The
> latency is as follows:
>
>
> ID SRC_INS_TIME TGT_INS_TIME LATENCY
> ---------- ------------------- ------------------- ----------
> 3710400 2012-01-11 14:00:01 2012-01-11 14:00:06 5
> 3710401 2012-01-11 14:01:01 2012-01-11 14:01:03 2
> 3710402 2012-01-11 14:02:01 2012-01-11 14:02:03 2
> 3710403 2012-01-11 14:03:01 2012-01-11 14:03:06 5
> 3710404 2012-01-11 14:04:00 2012-01-11 14:04:03 3
> 3710405 2012-01-11 14:05:00 2012-01-11 14:05:06 6
> 3710406 2012-01-11 14:06:00 2012-01-11 14:06:06 6
> 3710407 2012-01-11 14:07:00 2012-01-11 14:07:04 4
> 3710408 2012-01-11 14:08:00 2012-01-11 14:08:03 3
> 3710409 2012-01-11 14:09:01 2012-01-11 14:09:06 5
> 3710410 2012-01-11 14:10:01 2012-01-11 14:10:05 4
> 3710411 2012-01-11 14:11:00 2012-01-11 14:11:04 4
> 3710412 2012-01-11 14:12:01 2012-01-11 14:12:11 10
> 3710413 2012-01-11 14:13:01 2012-01-11 14:13:06 5
> 3710414 2012-01-11 14:14:00 2012-01-11 14:14:03 3
> 3710415 2012-01-11 14:15:01 2012-01-11 14:15:03 2
> 3710416 2012-01-11 14:16:00 2012-01-11 14:16:03 3
> 3710417 2012-01-11 14:17:00 2012-01-11 14:17:05 5
> 3710418 2012-01-11 14:18:00 2012-01-11 14:18:10 10
> 3710419 2012-01-11 14:19:00 2012-01-11 14:19:05 5
> 3710420 2012-01-11 14:20:00 2012-01-11 14:20:04 4
> 3710421 2012-01-11 14:21:00 2012-01-11 14:21:04 4
> 3710422 2012-01-11 14:22:01 2012-01-11 14:22:04 3
> 3710423 2012-01-11 14:23:01 2012-01-11 14:23:06 5
> 3710424 2012-01-11 14:24:00 2012-01-11 14:24:05 5
> 3710425 2012-01-11 14:25:00 2012-01-11 14:25:03 3
> 3710426 2012-01-11 14:26:00 2012-01-11 14:26:04 4
> 3710427 2012-01-11 14:27:00 2012-01-11 14:27:05 5
> 3710428 2012-01-11 14:28:01 2012-01-11 14:28:03 2
> 3710429 2012-01-11 14:29:01 2012-01-11 14:29:05 4
>
> Again I understand different purpose of replication and dataguard, just
> curious about in the best case scenario, can replication technology beat
> dataguard in term of latency since amount of data transport across network
> seem different between them.
>
> ________________________________
> From: "Allen, Brandon" <Brandon.Allen@xxxxxxxxxxx>
> To: "denis.sun@xxxxxxxxx" <denis.sun@xxxxxxxxx>; "oracle-l@xxxxxxxxxxxxx"
> <oracle-l@xxxxxxxxxxxxx>
> Sent: Wednesday, January 11, 2012 11:34 AM
> Subject: RE: Replication vs Data Guard
>
> I'm getting about 4 seconds latency on my standby using Data Guard with
> Real-Time Apply, but this standby can't be used for anything unless I stop
> the redo apply and open it read-only (or do a switchover or failover). So,
> if you need to use your secondary/standby/destination database for something
> other than DR or intermittent read-only access then you need to consider
> other replication options, or Active Data Guard.
> --
> //www.freelists.org/webpage/oracle-l
>
>
--
//www.freelists.org/webpage/oracle-l
Other related posts:
