1

If I have one server with RAID 50 or 10, then data is distributed among hard drives in case one fails.

Can the same be done across multiple servers. Can data be distributed among multiple servers, so that if one server fails, data would not be lost. I would assume I need at least 3 servers?

I am novice to this, so any direction you can give me would be greatly appreciated.

Running windows server 2012.

SJ Johnson
  • 13
  • 2
  • What kind of data? Databases? Sure. Files? Sure. You need to give us a lot more detail. – joeqwerty Sep 26 '16 at 18:28
  • Both, database (sql server) and files. – SJ Johnson Sep 26 '16 at 18:41
  • 1
    OK, again with the sparse information. We need lots of details. Knowing what version and edition of SQL Server would be helpful. At any rate, take a look at this: https://msdn.microsoft.com/en-us/library/ms190202.aspx. As for the file data, take a look at DFS. – joeqwerty Sep 26 '16 at 18:45

5 Answers5

3

You can do S2D if you upgrade to WS2016 Datacenter. Be prepared to purchase a full SCVMM license or run all management through PowerShell.

If you decide to stay with WS2012R2 then StarWind provides the most robust storage Fault Tolerance functionality.

Stuka
  • 5,445
  • 14
  • 13
2

Can data be distributed among multiple servers, so that if one server fails, data would not be lost.

Yes.

I would assume I need at least 3 servers?

Two servers is typically the minimum to get your database mirrored or to replicate your files.

HBruijn
  • 77,029
  • 24
  • 135
  • 201
2

Yes, there is a relatively simple way, it's called "Hyper-Convergence".

Software-defined storage solutions usually do the job of replicating data across nodes in order to achieve high availability.

For example: StarWind Virtual SAN, HPE VSA, Microsoft Storage Spaces Direct, VMware VSAN. They all provide a different "flavors" of storage redundancy. Such as: double or triple replication in real time, parity-based redundancy, etc.

Using one of such SDS solutions you can achieve cluster running without a downtime and data loss even if one node fails. The whole idea behind this is to replicate data on the block level between DAS of each cluster node which gives an ability to automatically fail-over production VMs to another node in case of failure.

Depending on specific vendor requirements and architecture, you may need different amount of servers to start with.

  • StarWind Virtual SAN requires minimum 2 nodes.
  • HPE VSA requires 3 but 3rd host can be a VM.
  • VMware VSAN starts from 3 nodes
  • Microsoft S2D requires 4 physical nodes.
Strepsils
  • 5,000
  • 10
  • 14
  • 1
    Only the latter option is viable for this user - they specified they use W2012R2 – Chopper3 Sep 27 '16 at 15:46
  • 2
    StarWind runs on bare-metal Windows. HPE VSA is an option too as it runs inside virtual machine. Microsoft S2D is the least attractive option actually due to significantly higher pricing. – Strepsils Oct 04 '16 at 13:01
1

Since you are on server 2012, you could set up a replica server that replicates data and use that, or you could set up a hyper-v environment with shared SAN storage. Either option will be costly, but when done right the hyper-v option works very nicely in a cluster configuration. Minimum of 2 hosts and shared storage is necessary for this, but once it is set up it is easy to add more hardware to the cluster for more vms or processing power. It also really helps with minimizing down time due to hardware failure or maintenance, as vms can be migrated between hosts without shutting them down.

milkman
  • 43
  • 4
-4

There isn't really a way to do that. Typically you would have a backup solution on each server that points to a separate storage location. You can restore just a file or the entire server if needed.

There is a fairly new concept making moves in IT that utilizes software-defined block storage. This would allow you to have a bank of servers (say 3) that all have resources, including local disk space, being used together. You run the software (scale.io for example) and you can have redundant, smart storage.

MoX
  • 1