File Server Storage Planning

Question

I am planning a large file server and I would like your advice and suggestions.  We currently have 4TB of data and I anticipate, based on growth history, we will grow to 25TB within the next 5 years. I am planning our storage infrastructure for this.
 Our files are 95% .jpg files and 5% word documents, excel docs and other misc. documents.  We have 2 servers.  One in building A and one in Building B (DR Site) that are connected via GB LAN.  The plan is to replicate the data from serverA
to ServerB using DFSR.  This will provide high availability in the event of a failure on ServerA.  We also have a 3rd server in buildingB that is a DPM2010 server which will backup this data.
I am trying to decide between to options and would like to know your input and advise.
Option 1:  Split our data into 4 volumes and then grow those volumes over time by adding storage and spanning each disk as necessary.
Option 2:  Put all data on the same volume and grow that one volume over time by adding storage and spanning the disk.
Taking into account our scenario, which decision do you recommend and why?

Answer 1

Hi!
You have quite a few questions inside that question. :-)
To evaluate the impact of using DFS-Replication, you need to look into the rate of change of the files, not exactly the total capacity or the number of files. DFS-R has some known limitations (for instance, you can't replicate open files). Also note that,
when using DFS, both server are running and available to users. Users in building A will go to the server in building A and users in building B will go to the server in building B (if you configure your sites and subnets correctly in Active Direction). That
could potentially lead to a replication conflict if a user in building A and a user in building B edit the same file at the same time. In that case, the last writer will win the conflict.
For true high availability, you need to use a File Server Failover Cluster with a SAN for shared storage. Since you mention different builds, you need to have two SAN solutions (one in each building) and a SAN-based solution to replicate the storage
between them. With a failover cluster, your specific file share will be running in only one of the buildings. If the cluster resource is online on building A, users from both buildings will access tthe share on that server and the server in building B will
only take over that resource in case of a failure of the server in building A.
In regards to using one or multiple volumes, it's really your call. Having multiple volumes will require multiple replication groups (if using DFS-R) or multiple cluster disks (if using Failover Cluster). This added configuration might pay off in giving
you more flexibility. For instance, you could have different cluster groups that could be onlined on different cluster nodes. It also could influence how you plan your backups. You can mask the complexity of having multiple volumes and even multiple file servers
from the end users by using DFS-Namespaces.
DPM 2010 would work with either solution.
Jose