Manshadi,
I'm a bit confused. You mention the HDS 9980 and XRC in the caption, the HDS
USP and asynch in the content.
There are three asynchronous remote copy methods supported in the 9980 and
USP:
- XRC
- TrueCopy Asynch
- Universal Replicator (HUR)
It's hard to answer your question without knowing which program product you
are using, and what other symptoms and performance data you have collected
from the storage Performance Monitor.
The one thing that all three of these have in common is that when you are
writing to the primary controller faster than it can send data to the
secondary the journal or sidefile will start to fill in the primary
controller. XRC and TrueCopy asynchronous will be have in as similar manner
using flow control to slow down writes and try and let the sidefile clear
out. For TrueCopy Asynch there are system options that can be set to make
this flow control less aggressive, and reduce the impact on response time.
For HUR the common practice is to disable flow control, as the unremitted
writes in the journal cache can be destaged to disk, so the risk of a
sidefile puncture is significantly reduced. You may want to check that you
have this disabled.
The answer is likely in RMF and the performance monitor data. Identifying
what component of response time was elongated should give you some guidance
as to what Performance Monitor data you need to examine. If you are running
the XRC monitor you will find some helpful data as well. Don't forget to
look at performance data on the secondary controller. A slowdown in the
secondary controller may cause data to be held longer in the primary
controller.
Ron
> -----Original Message-----
> From: IBM Mainframe Discussion List [mailto:[email protected]]
> On Behalf Of Mehrshad Manshadi
> Sent: Sunday, June 16, 2013 2:13 AM
> To: [email protected]
> Subject: [IBM-MAIN] XRC Problem HDS (9980)
>
> Hello all,
>
> does anyone have a "rule of thumb" as to how much Replication for
disaster
> recovery slows down overall disk I/O?
>
> We are replicating about 45 terabytes from HDS USP using asynch. We
> recently had issues that slowed our mainframe to a crawl. The work around
> was to stop replication. During the time replication was shut down our
batch
> ran twice as fast as it does with replication up.
>
> Regards
> Manshadi
>
>
> ----------------------------------------------------------------------
> For IBM-MAIN subscribe / signoff / archive access instructions, send email
to
> [email protected] with the message: INFO IBM-MAIN
----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to [email protected] with the message: INFO IBM-MAIN
