New to SRDF

We have a DMX4 and Unisys mainframe in Ca and another DMX4 and Unisys mainframe in Tx for DR. Currently there is a Unisys DMS2 database. We cut the transaction log every 15 minutes and ship it from Ca to be applied to the DR in Tx (database log shipping).

We will start using SRDF/A to replicate the database soon. We have never used SRDF until now. If the link between Ca and Tx goes down then the database updates will occur in Ca but there will be no way to replicate the tracks to Tx until the link is back up. I'm guessisng the SRDF changes will queue up in cache on the Ca array until they can be sent once the link is back up.

Is this correct? Would this impact performance on the Ca array? Could all the cache be used up in Ca to hold SRDF changes? What is the best way to handle extended SRDF outages due to WAN down issue? Hopefully this will never happen but we need to plan for it.

Thanks,

Brad

Responses(3)

Allen Ward

2.1K Posts

0

February 25th, 2010 11:00

If you are concerned about significant outages on the link between sites you should consider implementing DSE as well as enabling Transmit Idle for your SRDF/A sessions. The TI setting will handle shorter outages by allowing the sessions to remain in async mode even through a link outage. This means that when the link comes back up (after a relatively short outage) everything will still be consistent (as opposed to catching up in Adaptive Copy mode). DSE provides a disk based extension to TI that can allow you to maintain async mode through extended outages.

The other thing you are going to want to seriously consider though is some kind of "Gold Copy" clone of the R2 devices at a known good point in time on a regular basis. The reason for this is that if your link outage does extend long enough to drop you out of Async mode, your R2 devices will not be in a consistent state while you are getting back to async.

For the other part of your question regarding filling the entire cache with backed up track change data during an outage, there are limits built in to prevent that. These can be modified, but the defaults will certainly prevent you from completely filling cache with track delta data.

RobertDudley

448 Posts

1

February 24th, 2010 16:00

If or when your SRDF links are down long enough to drop the srdf groups the array no longer maintains the data in cache. Data is destaged to disk and the array maintains a "tracks owed" to target device listing per device in the srdf groups. When the SRDF links are restored and the srdf devices are again brought ready on the link to transfer data the tracks owed are read from disk and transferred to the target array.

S

SKT136306

31 Posts

0

February 25th, 2010 06:00

also u may tune "SRDF/A Maximum Cache Usage (Percent)". Also "Max # of Device Write Pending Slots" which seems to have a default value set.

View All

No Events found!

Symmetrix

New to SRDF