Disaster Recovery: Dedicated Infrastructure vs. Shared

[Amy Katt]

Hello Everyone,

We are in the process of redesigning our Business Continuity/Disaster Recovery Plans. I am looking for some suggestions/comments as to the benefits/drawbacks of having dedicated infrastructure for recovery vs. shared infrastructure and reconfiguring it when needed. Any suggestions or experiences you can share, would be greatly appreciated. I have researched endless WebPages, but have not found what I am looking for.

Thank you in advance for your comments.

[Silver_1]

Hi,

Understood what is your aim but need more info regarding your current infrastructure. eg: Distance beteween the datacenters etc.

Regards,
N

[Geoff Wild · Accepted solution]

And yet again this is one of those "that depends" kind of questions.

If you have the money, and your RTO is very low (Like 1 hour) then dedicated is the way to go.

That said, one of the systems I support has a RTO of 2 hours - and we use a "shared infrastructure" so to speak. What we have is a Prod system (2 node cluster) and a QA system (2 node cluster across the country).

In the event of a DR, we refresh QA with latest prod data, and bring it up as production.

On smaller APPS, we have a longer RTO - and bring up more then 1 APP on the same DR server.

Also have some that are DR'ed to servers with less power - IE Prod is 5 cpu's but DR only 4 is required.

Rgds...Geoff

[Amy Katt]

We have two campuses 15 miles apart. In general most production servers are in one campus and our test equipment is in the other.

The majority of our applications have a one week recovery. We have a large number of critical applications that we want to bring up over the course of that week, but by the end of the week all critical applications need to be up.

[Deoncia Grayson_1]

Since you have a week to bring up your critical applications before it makes an impact, then I guess it becomes more cost effective to go for a shared infrastructure, but if you needed something in less then a hour then it would be more beneficial to have a dedicated infrastructure just for disaster recovery purposes.

I don't have that luxury, but I am in a clustered environment.

[DCE]

Amy,
as long as you have the seperate campuses and the luxury of time to restore your production systems, then the shared environment will work.

This assumes your development/QA servers have enough horsepower (cpu, memory,disk) to absorb the additional work load. Also, would you contiue the development/QA work in the event of a disaster? If so, can you run concurrently with production?


If you find that the restore window shrinking, you may want to investigate a ServiceGard metro cluster. That is basically a sevice guard cluster designed to work between data centers in the same region, and would cut your down time significently. Again though, the mtro cluster has point in the design where you have to decide to use dedicated (idle) servers or development servers for the failover.

Another option is to use the services of one of the Disaster Recovery/Business Contininuity comapnies (SunGard, ComDisco, IBM are three that come to mind). these companies provide everything from servers to mobile offices in the event of a disaster.

Also, what type of disaster are planning for?

Losing a server?
The server dies - replacement parts are not available. Data may or may not be salvaged from the system. backup tapes are good.

Losing a data center?
The fire system accidently goes off, a fire in the data center, etc. The networking may still be viable. Hopefully tapes were stored off site.

Losing a building?
Explosion, flood, fire. Networking is gone as well as equipment. People have no office to work from.

Losing a region?
tornado, earthquake, hurricane, regional power outage due a power plant going off line. Will both data centers be affected because of how close they are? Can the backup tapes be transported to the DR site, or are not available due to the nature of the disaster?

[Jan van den Ende]

Amy,

the (I would say important) info that has not been given, nor asked yet:
What kind of servers are you dealing with?
It makes A LOT of difference concerning possibilities, cost, effort, impact, and risk, what you are running. Microsoft, (high risk, very high effort, moderate cost, moderate-high service interruption) HPUX (I am no expert on that, but AFAIK low-medium risk, medium effort, medium cost, moderate (low?) service interruption), Tru64 (low-medium risk, low-medium effort, low-dedium cost, low-medium service interruption) or VMS (low risk, very low effort, medium effort, low service interruption). If you are running VMS in cluster, the effort will be even lower, and the service interruption can be down to seconds.

So, IT DEPENDS!

fwiw,

Proost.

Have one on me.

jpe