Managing data storage is just as much of a task (or greater) as managing the servers themselves. It makes sense to centralise management in larger organisations wherever possible. Storage area network (SAN) technology enables a network of physical storage devices to be managed as a single unit. They can then be readily divided into a number of logical drives.
Basic SANs use either Fibre Channel (FC) or iSCSI technology. iSCSI runs over standard gigabit Ethernet networks using software to support the SAN infrastructure, whereas FC devices are much faster but require specialised hardware such as optical fibre cabling and FC switches.
SAN vs. NAS
Storage area networks and network attached storage (NAS) are similar names and easily confused. The key differences are:
- The manner in which stored information is presented to servers. Data stored in a NAS will be seen on a non-local drive whereas data stored on a SAN appears to reside in a local drive.
- Physical hardware. A NAS is a single storage device, whereas a SAN is a centralised collection of networked storage devices. Multiple servers could require an individual NAS each, but a SAN allows sharing of a single storage area.
- Administration. SAN space can be allocated as needed to each server, often on-the-fly. Administrators can determine a total amount of physical drive space required without concerning themselves with how to physically allocate space to each server.
- A NAS is optimised for sharing while a SAN is optimised for speed and capacity.
Comparison of NAS and SAN (Credit: Enex TestLab/CBS Interactive)
NAS/SAN hybrids are also possible where one or more partitions in the SAN are provided with a NAS gateway.
Storage clustering is a method of combining networked storage devices to form a single virtual device. It allows the creation of large, redundant storage arrays which remain relatively easy to manage. Since clustered systems contain multiple disks, each with its own read-write head, performance can potentially be multiplied by simultaneously accessing multiple physical devices.
Available in a variety of flavours, Redundant Array of Independent (or Inexpensive) Disks (RAID) is a method of combining several physical drives into a single logical unit with varying levels of data redundancy. The most commonly seen varieties are 0, 1, 5, 6, 10, 50 and 60. RAID 0 (also called striping) connects several identically sized volumes together, increasing read and write speed but offering no redundancy. RAID 1 (also called mirroring) is a 1:1 copy of a disk, so if one disk fails, there is a redundant copy — it should increase read speeds, but has a penalty on writes and can be costly — for every disk you wish to protect, you must purchase an additional identical capacity disk.
RAID 5 and 6 are the compromise — both allow a collection of disks of identical capacity to be connected together to form a single logical unit, but rather than the expense of having to purchase a drive per drive you wish to protect, they spread the parity data across all the disks in the array. RAID 5 can survive a single drive failure, RAID 6 can survive two.
RAID 10, 50 and 60 are combinations of RAID 0 plus RAID 1, 5 or 6. Some SAN vendors support non-standard RAID varieties or less common standards such as RAID 3 or 4.
Consider an organisation where data storage is shared equally between two servers. If one server is discovered to use its virtual disk space faster than the other, it may be necessary to purchase more drives to satisfy the capacity needs of this space-hungry server even though there is still storage space available on the other server. Thin provisioning is a method of allocating available space to each virtual drive as needed. This has the potential to save organisations from purchasing vast amounts of hardware that may never be fully utilised.