Architecture type question.

[MSwift]

Greetings!!

We all know both DR and HA.

From the architect point of view, what would be the difference between Disaster Recovery (DR) and High Availability (HA)?

Thanks

Mike.

[Patrick Wallek]

HA - Address single points of failure. This could include: multiple power supplies in a server, multiple NICs teamed, multiple paths to storage, utilization of RAID1, RAID10 or RAID5 in storage (to prevent data loss due to loss of a disk), clusters in case the entire server goes down

DR - What you do in a disaster. At its most drastic -- What do you do if you lose your entire data center? How do you recover? At its least drastic - How do you recover a server that completely failed?

[Jorge Pons]

Hi MSwift

In DR you must support any disaster can ocurr and the system must be up and running.
For example, to have BRS in other location.

HA, support one single point of failure, only one.

Regards, Jorge

[Rita C Workman]

HA is generally involves local failure, resulting in the use of failover technology. That might be failover from one nic card to another, one fabric to another, one disk to another - or one server to another.
HA generally, at worst, has the potential for a locally recoverable interuption.

DR, by it's name, denotes disaster. Disaster recovery implies far more serious disruptions, potentially leading to relocation. Your site is down, but your staff can work remotely. Or your site is down and your staff is unable to work remotely at their current location.
So disaster recovery requires making your systems available, ensuring your staff is in place. DR involves handling all this kind of restoration technology and machine and people logistics.

Kindest regards,
Rita

[James R. Ferguson]

Hi Mike:

I think of high-availability (HA) as a subset (albeit a required one) of DR (disaster recovery).

If you don't have HA you may find that a lack of a resource is a nascent "disaster" that grows.

HA, in one sense, is server-centric. DR, in one sense is server-wide insofar as in a DR situation you have to address establishing inter-interoperability among many pieces.

You could envision Business Continuity (BC) as a superset of DR.

Regards!

...JRF...

[Ismail Azad]

Hi MSwift,

"High availability" as the term suggests refers to acheiving maximum amount of uptime on your server as even 10 minutes of downtime can cause huge losses in terms of money and when you think about "HA" on HPUX, first thing that comes into mind is HP serviceguard..

HA as mentioned addresses or takes away the point of having an SPOF be it in terms of disks, cables, controllers, for which you can mirror, provide PV links and multiple controllers. However , you can say disaster recovery addresses "multiple points of failure" or one "huge SPOF" talking in terms of sometimes a "metro" and "continental" cluster. Yes, a "subset" as mentioned by JRF is a very nice way of putting it...

What happens in the case of a volcanic eruption, earthquakes and floods? When data is critical it is critical and "planning" for it has to be done.

Regards
Ismail Azad

[Torsten.]

>> What happens in the case of a volcanic eruption, earthquakes and floods?

You can expand this list, e.g. total power loss for the building, city ..., building burned down, etc.


You need to think about if you need to keep your business up if something happens or if you are lost anyway.

If you need to keep it up, you need a solution, this could mean a data center in another building/city/continent/planet ...

If such bad thing happens, you need to continue from another place, which could be a mixture from HA (clusters) and DR (start the backup data center with current data).

[Jose Mosquera]

Hi,

Please take a look of this article, no images are included but is useful:
http://www.disaster-resource.com/articles/06p_098.shtml

Rgds

[Manix]

In simple terms as we have application fail overs between cluster nodes for high availability , in DR we may have entire data center replicated to a different geography and applications can run even when there is a complete outage at one end.

There are some features like VVR ,CVR where data is replicated across the site to enable complete fail overs & recovery.

Thanks
Manix