Exchange 2013 LUN/Database Configuration

[IT_Vision]

Hello,

Looking for some ideas on LUN / Mailbox Database setups for a 2-Node DAG Setup for Exchange 2013.

here is a quick outline of our current setup with Exchange 2010 in a DAG.

We currently have around 50 Mailbox Databases supporting 500 user mailboxes which are spread across 10 LUNs and another 10 LUNs for Logs. The Mailbox server is setup in a DAG across WAN links. We are utilizing VMware for the hosts and using iSCSI inside the guest OS with mount points. Since implementing Veeam I want to start utilizing it to back the new Exchange 2013 environment. Current space used for all mailbox databases is around 800GB.

My initial though was to create 20 LUNs on our backend storage (it’s a mix of SSD and spinning disks) and gets pretty good iops. 10 luns for the vmdk which then will house 5 Exchange Mailbox databases each. Then another 10 LUNs for the logs, and create Is this over kill?

My concern is if I present the drives to the Exchange mailbox server it is going to look like 20 mapped drives unless I utilize mount points again. The other thought was to consolidate some of the databases. The reason for so many databases was we are currently using Microsoft DPM and to perform a restore we had to restore an entire database so we wanted to keep the size down on each of the databases.

Also another thought was to create 1 TB LUN for all the mailbox databases and another smaller LUN for the logs. Not sure how Veeam or Exchange for that matter would like a single LUN like that during a snap shot process for backup.

Any thoughts or current ideas people have implemented would be greatly appreciated

Cheers!

[foggy]

The reason for so many databases was we are currently using Microsoft DPM and to perform a restore we had to restore an entire database so we wanted to keep the size down on each of the databases.

Definitely not a concern for Veeam Explorer for Exchange, which allows to perform granular item-level recovery directly from the backup file.

Also another thought was to create 1 TB LUN for all the mailbox databases and another smaller LUN for the logs. Not sure how Veeam or Exchange for that matter would like a single LUN like that during a snap shot process for backup.

Veeam B&R will like it as long as it has enough free space for the VM snapshot.

Also, please review this topic with some recommendations on backing up Exchange DAG, should be helpful.

[IT_Vision]

Hi Foggy,

Thank you for the response. My plan is to setup a separate LUNs for Databases and logs. Since this will be in a VMware environment each of the LUNs will house a single VMDK. Inside the guest OS I plan to utilize mount points to store the Mailbox Databases and Logs on those separate LUNs. Below is what I am thinking about doing to separate the DBs from the LOGs.

MDB-VMDK1 = MDB1, MDB2, MDB3
LOG-VMDK1 = MDB1Logs, MDB2Logs, MDB3Logs

MDB-VMDK2 = MDB4, MDB5, MDB6
LOG-VMDK2 = MDB4Logs, MDB5Logs, MDB6Logs

Would it be idea to create separate jobs to backup lets say MDB-VMDK1 first, then LOG-VMDK1, then MDB-VMDK2 then LOG-VMDK2 in that particular order? or backup both MDB-VMDK1 and LOG-VMDK1 at the same time and setup another job to backup both MDB-VMDK2 and LOG-VMDK2 when the previous job completes.

We are also looking at a doing a 2 node DAG across WAN links so I think doing a single 1TB VMDK on a single LUN might pose a problem so I want to try and split thing up a bit.

Any thoughts as to setting it up this way?

[foggy]

I don't think that creating separate jobs is a good idea, you will definitely have issues with several types of restore (full VM restore, for example). A recommended way is to backup the entire node in one job.

Btw, you can search for other existing threads discussing similar considerations for information on what others are doing, here's one of them.

[smartsys]

I would create a number of databases based on user quota's. Normally 3 or 4.
Each database would have a seperate (properly formated) Windows volume and a drive letter. On a seperate VMDK in a separate VMWare Datastore and on a separte LUN.
I often use a common LOG drive for all databases, also on a seperate VMDK, VMWare datastore and LAN.

When using a DAG there is exactly the same config on another server and databases are replicated between those.

Microsoft recommends about 200GB datastore size with 120% provisioned storage space for a normal Stand-Alone implementation. In a High-Availability implementation this goes up to 2TB datastore size.

I normally don't you DAG solutions.
Backup with Veeam is a two stage process. A multiple time per day backup or replication with a few days worth of restore points is for disaster recovery. Second is a Backup Job copy job to cheap, but reliable storage for archive (restore of single items) purpose.

I mainly use Netgear hardware for backup tasks. The bigger rackmounted ReadyNAS of even better the ReadyDATA products.
The primary backup goes to a RAID 1+0 (10) volume and the copy to a RAID 50 or 60. Best is to create a second copy of the archive to another storage box at another location or use tape as a backup of the backup archive. It is even possible to enable DEDUP on the archive storage to more efficiently use the available storage space.

Veeam works very efficiënt with the Veeam Explorer for Exchange.