Exchange Server Forums

How high-available do we need to go?

Users viewing this topic: none

Logged in as: Guest

Printable Version

All Forums >> [Microsoft Exchange 2010] >> High Availability >> How high-available do we need to go?

Page: [1]

Message

<< Older Topic Newer Topic >>

How high-available do we need to go? - 20.Nov.2012 4:07:29 PM

dutchad

Posts: 1
Joined: 20.Nov.2012
Status: offline

OK, this is not so much a question, more trying to get a discussion going.

I have learned, as probably the most of you out there, that�s it�s best practice to create �High Availability� for my Exchange 2010 environment.
It consists of a CAS array, 2 HT servers and 2 MBX servers, all running Exchange 2010 SP2.
So, I�ve created a second copy of my MBX in a DAG, running my servers on mirrors for the OS and RAID10 for the databases, separated logfile disks and database disks.
For some time now we also run MS Data Protection Manager to �protect� (we still say �backup�, but OK...) everything:
DPM backups the entire servers every day and every 15 minutes takes a snapshot of all the MBX databases.
So, when one of the mailservers crashes, I keep on running on the second DAG member and have enough time to rebuild the primary.
In this rebuilding we have the choice between restoring the server and let DPM to the rebuild of the databases or just let the DAG resync again.
So far so good, but now I�m realizing that we have build 300% of resilience, where I think 100% should be enough (this is what we were all dreaming about a couple of years ago?)
On the other hand all the users constantly want more storage and more performance and want to pay less�.

That triggered me to a naughty thought:
If my protection is that good, why wouldn�t I create a primary mailbox server that has high capacity and high performance?
Let�s say: I�d build a new server with a RAID-1 for the OS, a single disk for the logfiles and a RAID-0 (high performance stripe set) for the database?
I hear you all say: that�s not done, it�s out of the question, one disk defect and your server is gone!
OK, I agree. But then again: what�s the problem if my server is gone?
I always will have my secondary DAG member that will take over seamlessly and I will be able to recover the database quickly using Microsoft DPM.
After this recovery there will be a data loss of something like half an hour which the still running other DAG member will have solved quickly.

So, my question is: am I overlooking something or was I just stuck in �old� best-practices which from time to time should be reconsidered?

Like to hear from you all!

Post #: 1

Featured Links*

RE: How high-available do we need to go? - 21.Nov.2012 3:02:05 AM

zbnet

Posts: 1019
Joined: 25.Sep.2003
From: York, UK
Status: offline

It's good that you've started to understand and appreciate the HA features of Exchange 2010!

With Ex2010, Microsoft have delivered a very capable, flexible product, with a level of HA that was not possible with previous versions of Exchange. Clearly how you utilise that capability and flexibility to meet the needs of your users is up to you. Personally, I'd rather have flexibility and options than not have them.

But really only you know the needs of your users, and what is acceptable to them. A DAG allows you to have anywhere from 1 to 16 copies of each database (given enough servers). So how many copies is the 'right' number to keep? Only you and your users can answer that.

Obviously there are some common requirements, and so MS produce recommendations and guidelines to try and help people who don't want to bother with a full understanding their user requirements or the full extent of the capabilities and flexibility of their product. So MS say 2 copies of a DB are good, but three are better. That recommendation manifests itself in the recommendations of JBOD - don't use it (use RAID 1 or 1+0 instead) for your DBs if you only have 2 DAG nodes; but when you have 3 or more DAG nodes then JBOD for DB drives becomes feasible, without too much risk to your data.

Do you still need a backup if you have multiple redundant copies of your DBs? Only you and your users can answer. Some do need backups, for off-site recoverability, or for compliance purposes; some don't - MS themselves saved millions of dollars by removing the requirement to back up their own internal DAGs.

The key to answering all these questions for yourself is a good understanding of the capabilities of the product, and a discussion with your users about what risks they are willing to tolerate to the data, and how much loss of it (either temporarily or permanently) costs them and their business.

It sounds like you're along way along that road - well done, and keep engaged with this community, we are better together.

(in reply to dutchad)

Post #: 2