Network management Toolkit

The FAQs on SANs

Tags:
Fibre Channel,
NAS,
EMC,
SAN

Rupert Goodwins ZDNet.co.uk

Published: 05 Dec 2002 09:55 GMT

What is a SAN?
A SAN is a Storage Area Network, a collection of hard disks in a dedicated very fast network connected to servers via multiple switches. The SAN network is called a fabric, and connects any server to any storage device.

Why would I want one?
SANs are designed to be very reliable, very scalable and very flexible. You can add more storage without disrupting operations, use a number of back-up schemes that don't load your main network, and you can expect the SAN to manage very large data collections that span multiple disks. There are a wide variety of SAN switches, routers and arrays available from a number of vendors, with most activity in the mid-range £5,000 to £200,000 segment. SANs can also be physically spread over whatever distance their network allows -- kilometres with fibre-based systems -- which can be a highly robust configuration.

Why wouldn't I want one?
They are expensive, although prices are dropping, and while the concept is proven there is still much work to be done in interoperability between different manufacturers. Of late, they've taken to suing each other for patent infringements in preference to designing open interfaces, although there are signs that this madness is passing -- check the Bluefin initiative from the Storage Network Industry Association (www.snia.org). One of the benefits of SANs is that they let you manage disparate storage units in a central place, however, this software is often proprietary and not easy to integrate with the rest of your installation.

Bluefin?
Bluefin is a common interface for SAN Management, according to the SNIA. It's based on the Common Information Model (CIM) from the Web-Based Enterprise Management (WBEM) initiative, sanctified by the Distributed Management Task Force (DMTF). The whole thing is based on clients and agents swapping information and commands through messaging, and includes resource locking and automated discovery. It also generates more acronyms per square inch than any other concept in known space.

What sort of network does a SAN use?
The most popular is Fibre Channel, although iSCSI, Gigabit Ethernet and IP over Fibre Channel (FCIP) are alternatives. Infiniband had some support, but is now widely seen as the deadest of ducks; iSCSI was heavily promoted by big names such as Cisco, but met with indifference. Typically, switches run at 2 Gbps. Regardless of the physical layer, it's important to remember that a SAN is a network in its own right and has the same management and security issues as any business critical LAN. In particular, structuring the internal SAN network into separate zones with different access privileges and physical configurations can enhance security and reliability.

What kinds of switches are appropriate for SAN use?
There are two main classes of switch, director and fabric. Director-class switches are the dreadnoughts, with high port counts, hot-swappable redundant core components, and other reliability enhancing architectural features that provide 99.999 percent uptime. They have a price to match. Fabric switches are less highly specified. McData and InRange Technologies are the biggest names in director switches; they also make fabric switches alongside Brocade, Vixel, Qlogic, Gadzoox and the one that makes everyone else nervous, Cisco. The usual suspects, IBM, EMC, HP/Compaq et al will all sell you complete SANs based on other people's equipment.

Can I use my existing network to build a SAN?
In theory, yes. In practice, many of the benefits of a SAN come from having its own dedicated bandwidth, and if the network has to carry other, non-storage traffic reliability and response times would suffer. It's best to consider the SAN as an independent entity. It also benefits your existing LAN, as data-intensive tasks such as backups or replication can take place away from your mainstream LAN environment.

What's a SAN Appliance?
In a plain SAN, the storage network acts mostly as a communications medium, and the allocation of data to storage devices happens on a server. A SAN Appliance sits in the storage network and mediates each operation, centralising the control over data placement. It can also handle redundancy for the disk controllers. While this offloads some of the processing requirements from the servers and also helps simplify management, the appliance can become a bottleneck or source or unreliability itself.

What's NAS?
NAS -- network attached storage -- connects storage devices to servers over a network, but leaves the storage management as a one on one relationship between server and storage. You can amalgamate NAS systems with SANs by using a NAS head, and the SNIA is keen to get everyone to move on from the SAN/NAS dichotomy and into a world of more flexible architectures.

Have your say instantly in the Tech Update forum.

Find out what's where in the new Tech Update with our Guided Tour.

Let the editors know what you think in the Mailroom.