Implementing DR between two US & UK sites (system replication)

[Khalil Ahmed]

If I may I would like to describe first what I am trying to achieve.

We have two sites one in the US and one in the UK. Both sites have the same environment i.e. run the SAP R/3 ERP software on HP9000 servers (i.e. HP-UX 11 O/S) against an Oracle 9i database. Both sites also utilise HP’s EVA5000 disk array to house the production data.

What we would like to do is to allow each site to be a Disaster Recovery (DR) site for the other. To do this we have allocated money for additional hardware (i.e. servers, disk capacity) for both sites and for the network connection (i.e. “pipe” ) between the two sites. The idea being that if the UK went down (say), a limited number of UK users would still be able to continue working by logging on to the replicated UK system on the US site, at least until the UK site had been fully recovered. Similarly if the US went down, a limited number of US users would also be able to continue working by logging on to the replicated US system on the UK side.

Has anyone done anything like this before?… any recommendations on the best method and more importantly how to do it. Some methods I can think of are overnight EVA5K disk-to-disk copy, Oracle Parallel-Server and Oracle Stand-By system (although the Oracle methods may require some specialised knowledge/training).

In fact if you have any documents or know of any links to documents that would help me I would be very grateful.

Many thanks

[Sunil Sharma_1]

I am not sure how SAP will integrte in this but for oracle you can setup a standby database using cold backup of production database and apply archive logs from production environment to DR site. When disaster will happen shoutdown standby datbase and open the database in normal mode.

Sunil

[melvyn burnard]

You might be interested in reading through the Designing Disaster Tolerant Clusters manual at:
http://docs.hp.com/hpux/pdf/B7660-90015.pdf

And particularly the Continental Cluster portion.

This allows you to have the two clusters, one in UK, one in USA, monitoring each other, and provided you have the data replication in place, each site could act as a the DR site for the other, wiht automated failover and alerts in the event of a failure.

[Alzhy]

Yep.. if an ALL HP solution -- you're best bet is a Continental Cluster.

If not, you may want to look at a Veritas Solution.

How large is your pipe between the UK and the US?

[Khalil Ahmed]

Melvyn/Nelson, I've had a brief look at the PDF document, this ALL HP solution involving this continental cluster approach seems a bit complicated, both to implement and to maintain afterwards. I assume implemented the ALL HP solution would involve buying additional licenses software? (by the way we recently removed Service Guard from our local site because it was a pain to maintain).

Isn't there a simpler method?... For example could we not do a nightly mirroring of the disks onto the remote site (putting the database into hot backup first). Then in the event of a disaster as long as we have all the relevant database and SAP files we could startup the database and simply "roll forward" (ie apply the DB redo logs).

I am also considering the Oracle standby database approach. I need to look into this in a little bit more detail, but as far as I understand it involves applying archived redo logs to the remote standby DB as and when they are created (simplistically speaking!).

As far how big the "pipe" is between the two sites that still needs to be determined and may be dictated by the approach adopted.

For Example if we decide on the Standby DB approach, we need to cater for 40 MB archive logs that during the busy period are being created every 5mins. So every 5mins we would be pusing 40MB down the pipe to the remote site.

Our DB is currently at 400GB so for a nightly disk-to-disk copy we would need to push 400GB+ down the pipe (I guess?).

Regards

Khalil

[Alzhy]

The Veritas as wellas the ALL HP approaches both have each strengths and weaknesses. Successful implementation will largely depend on

1.The expertise and experience you have in house.
2. Storage Architecture chosen
3. Bandwidth between the sites.

The Oracle Standby approach is not totally a seamless failover solution .. you can still have some data loss.

You mentioned EVAs as the storage involved ... I do not know if Continous Access (The EVA's replication software) works accross IP networks .. I think it only works for FC links...

[Alzhy]

Here's the link to the available Veritas Solution:

http://veritas.com/Products/www?c=category&refId=149

Am not sure but I think some components of HP Wide Are/Long Disntance Clusters/HA use Veritas Products like VxVM, Filesystem (VxFS) and Veritas Storage/Volume Replicator.

[Geoff Wild]

For our DR, we use EMC SRDF - which is like Business Copy in HP/Hitachi...

All that does is copy changed tracks on the disk frames from one site to another...

I don't know if the EVA has a tool like that or not...

It works really well - we just passed out DRP audit....

We were able to bring up at our DR site, our 1.5TB SAP system in less then 48 hours...

Rgds...Geoff

[Alzhy]

Geoff, Yes the EVA has that tool .. called Continous Access -- but I think it works accross relatively "short" distances. The distance in question is accross the Atlantic.

Some word of advice though.. what ever DR/BC/Replication/Failover strategy you implement -- be sure there is no vendor lock in - so your scheme works no matter what storage infrastructure you'll choose in the future.