What's best practice for Domain Server setup

[tonyoung]

Hi all,

I'm planning to migrate my existing Windows 2008 servers to Windows 2012. Also had funding for new hardware too =)

Here's what I got:
3 - Proliant DL370 with 14 GB RAM and ~4.2 TB of hard drive
3 - Proliant DL380 with 20 GB RAM and ~1.2 TB of hard drive
1 - HP D2D backup 4 TB

I will be setting up:
1 - Domain Controller / DNS / Printer Server on a DL380
1 - Sql server on DL380
1 - Application server on DL380
1 - File server on DL370 running DFS replication (replicating with below server)
1 - File server on DL370 running DFS reflication (replicating with above server)
1 - Backup server running Backup Exec on DL370

The 3 Proliant DL380 servers I plan to make only one partition (C drive) with the OS and associated software going on that. Running RAID 6.

The 2 Proliant DL370 will be setup with 2 - 2 TB partitions both on different controllers. 1st partition will be divided into 2 logical drives. 1 for OS (200 GB) and 1 for shadow copy (1800 GB). And the other partition will be 2 TB for data. The Second DL370 will be mirror of this server also running DFS replication for fault tolerance. Both servers setup with RAID 6. Users will map to 1 server for personal folder and the other server for share folders. If 1 server fails all users can remap to the other server.

The final Proliant 370 will be the backup server running backup exec connect to HP D2D via iSCSI. This server will also house a removable hard drive cradle for offsite monthend storage. This server will backup all data on the file servers and any additional data stored on the shares. This will able be seperated into 2 - 2 TB partitions. One for OS and the other for additional storage.

OK... what you guys think? Is it a good plan? Disk space good? Configuration good? Any better ideas? Thanks all for your input.

Tony

 

[cajuntank]

Here's a possible scenario I would look at. I personally use HP's Lefthand P4000 SAN equipment, but I know 2012 now comes with the storage role that will turn the box into a iSCSI storage box (of course you can pretty much do this now with Microsoft's free iSCSI target software) , but I do not know yet if you can create a redundant pool of storage by using multitple boxes using 2012. So assuming you can (which of course would be better), I would take the DL370 machines, create a SAN storage pool for your DL380 machines. Bump the RAM and processor, unless you have dual now on those boxes, and create a 3 node Hyper-V cluster. Create all of your servers here on out as virtual machines on the Hyper-V cluster. Your DL380 machines just need to have two physical hard drives in RAID1 for the host OS and iSCSI off to the storage pool for the virtual machines for all the other servers you need to install. You don't make mention on the generation of DL servers, so again, assuming they are newer, you should have enough processing power due to the multiple cores and processors and enough RAM to host literlly a dozen or more servers, if not dozens depending on RAM and processor requirements of your servers and applications.

In this scenario, all of your systems are redundant and hardware agnostic.

 

[tonyoung]

Thanks cajuntank for the suggestions. What I'm trying to achive here with the hardware is 2 seperate file servers containing the same info in real-time. Any failure on either hardware or software on one will not affect the other. I will also be using these server to load-balance as users will got to one server for one share location and the other for another. Is windows capable of replicating the actual data with Clustering? thanks.

 

[cajuntank]

In clustering, (as of 2008 R2), the only allowance for clustering in an Active/Active scenario (where two or more servers are accessing the same data location) is Hyper-V. All other clustering will be Active/Passive where one will be active and the other(s) will remain passive until some event causes them to switch. This might be different in 2012 as there was talk about increasing support for other applications to take advantage of Cluster Shared Volumes. I make this statement because there is a changes that 2012 will bring to the table that I'm not up on yet.

In regards to users accessing shares and load balancing, you could always have done this with DFS. DFS would allow you to have multiple servers (don't even need to be in a cluster) and they have shares that are sync'd together via DFS by creating shares on each server, then having those shares be part of the DFS. You then control how replication is handled from DFS management, no clustering needed.