Manage Learn to apply best practices and optimize your operations.

iSCSI storage: How to leverage RAID and ensure server uptime

Learn how RAID can be used with iSCSI storage, as well as how iSCSI storage systems can be configured for high availability to eliminate single points of failure.

As a technician at a community college, I'm looking to upgrade our IT infrastructure. I want to get rid of our standalone servers and push toward virtualization with Microsoft Hyper-V. Right now I have a cheap 4 TB NAS system, which is iSCSI-capable and good for testing purposes. I'm planning on clustering two servers running Windows Server 2008 R2 to iSCSI, but I have a few questions.

First, does iSCSI have the ability to be partitioned as RAID? The iSCSI NAS I currently have does not allow iSCSI...

storage to be on RAID. So if I create a 3 TB RAID partition, only 1 TB is left for iSCSI. Am I right to assume that this isn't how storage is partitioned on a more expensive iSCSI storage system?

Also, with my current infrastructure, the single point of failure is on the iSCSI storage. In a network that requires certain servers to be up 99% of the time, this worries me. Is it possible to cluster two iSCSI storage systems? If not, how I can I achieve quick server uptime in the event of a hardware failure within iSCSI?

First, while there are storage systems that support both iSCSI and NAS (NFS and CIFS), there is no such thing as iSCSI NAS. You're essentially using an iSCSI (block)- or NAS (file)-based access of storage. For example, applications such as Microsoft Exchange might use iSCSI for block access to a LUN on a storage system that houses NFS and CIFS data, even though that storage exists in a file system that uses other portions of the storage system.

Likewise, the underlying storage for a unified storage system relies on RAID protection where specific configuration options will vary by vendor. Some solutions enable some storage to be configured as RAID 1 (mirror), some storage as RAID 5 (stripe with parity) and other storage as RAID 6 (dual parity).

The underlying storage is then allocated as LUNs or volumes and made available to servers through block interfaces such as SAS, iSCSI, Fibre Channel (FC) or Fibre Channel over Ethernet (FCoE), or via a file system.

Assuming that your environment is housing a 12 drive (512 GB disks) storage system, you could create a 3 TB RAID 5 (6 x 512 GB + 1 Parity disk) group and map that as a single LUN to the NAS file system. You would still have a 1 TB (4 x 512 GB) RAID 1 plus a hot spare disk. For high performance, you can create the iSCSI block LUN using faster 15 K SAS disk drives and then use 7.2 K SAS or SATA high-capacity drives for the NAS shares.

As to the single point of failure being at the iSCSI storage system, that will depend on what you use, how it is configured and where it is deployed. Most small business storage systems provide SAS, iSCSI, FC and or NAS capabilities, and have dual redundant hot swappable controllers, power supplies, cooling fans, spare disk drive options, snapshots and multiple RAID levels. Thus, they can be configured for high availability to eliminate single points of failure. If you are concerned that the entire storage system is a single point of failure, then you can mirror or replicate the data between a pair of storage system for an even higher availability and cost.

About the author: Greg Schulz is founder and senior analyst with the IT infrastructure analyst and consulting firm StorageIO Group.

Dig Deeper on SAN technology and arrays