Learning the ABCs for Network Storage - Part Two

SAN Overview

A SAN or Storage Area Network is simply your storage and backup devices grouped together in their own network and separated/abstracted from your server/servers via a switch. The basic idea is to use a single degree of separation so you can pool storage resources, allow access from multiple servers, and eliminate the downtime associated with DAS. SAN is a simple concept, but in the real world it's also joined at the hip with Fibre Channel.

Fibre channel is a high-speed technology/communications protocol that allows flexible connection topologies and cable runs of several kilometers. One common misconception concerning Fibre Channel is that it's all fiber optical (notice the difference in spellings). Fibre channel cable runs of 20 meters or less are actually copper; only on longer runs do things go optical.

A typical SAN consists of a farm of disk arrays, CD-ROM carousels, tape libraries, etc., interconnected with Fibre Channel and connected to the outside world via a Fibre Channel switch or bridge. The connections inside the boxes can be Fibre Channel, SCSI, ATA, or any other technology. Wiring all the storage devices into a network requires a plethora of hubs, bridges, routers, multiplexors, copper to optical converters, and other devices just as with a normal communications network.

SAN pros:

SANs are almost infinitely scalable: There are no meaningful limits on the number of devices you can include in a SAN. You could easily scale one to hundreds of terabytes.

SANs mean less down time: Unlike a DAS box, you can upgrade your SAN, replace defective drives, or backup your data without taking any servers offline. A properly configured SAN with mirroring and redundant servers should be capable of zero downtime.

SANs can be shared: Because they're not directly attached to any one server or network, a SAN can be shared by all.

SAN means long distance: With Fibre Channel capable of runs up to 10 kilometers you can keep your data in a remote, physically secure location. As the WTC tragedy has proven, keeping data down the hall is little guarantee of safety. Of course, there's nothing to prevent you from moving a Fibre Channel NAS such as Network Applications' high-end boxes offsite either. Still, most NAS boxes use more mundane Ethernet technology. Fibre Channel switching also makes it very easy to establish private connections with other SANs for mirroring, backup, or maintenance.

SAN is versatile: A SAN can be single entity, a master grouping of several SANs, and even include SANs in remote locations.

SAN cons:

SANs are expensive: The cost of a SAN is high compared to NAS and DAS technology for several reasons: Fibre Channel technology tends to be pricier, SANs serve the high-end customer where high margins are more acceptable, and bringing IT up to speed on SAN can be pricey since maintenance requires a higher degree of skill than DAS or NAS.

Leveraging existing technology can be difficult: Allowing you to employ your company's legacy storage was supposed to be one of the SAN’s strengths. However, lack of SAN-building skills has hindered deployment of homegrown SANs. Pre-packaged SANs based entirely upon Fibre Channel are far more common than scratch-builts and probably won't support your existing devices.

SAN choices are more limited: the number of SAN vendors is miniscule compared to NAS and DAS vendors for the simple reason that developing a SAN product is far more expensive. With the high price and large scale of SANs, buyers demand timely, rock-solid service and support--something large vendors have an easier time providing.

SANs aren't that smart: A SAN relies on the servers accessing it for its file systems. A SAN appears as a blank disc to the servers and is then divvied up, or partitioned between them. A Unix server can't see the Windows portion of the SAN and vice versa.

SANs can be wasteful: Because SANs are relatively dumb and each server requires its own space, you may wind up with multiple copies of files--one on every different section of the SAN. If you're streaming a lot of content this could lead to a colossal waste of space.

SANs are more difficult to manage: because of the large number of components and colossal amounts of storage that can be involved, efficient management of a SAN can be a real chore. Virtualization (see the virtues of Virtualization) of SANs is one of the ways to get a handle on their complexity.

Note: If the roll-your-own-SAN method appeals to you, or you simply want to get your IT staff up to speed on SAN maintenance a training center like Infinity I/O (www.infinityio.com) might be in order.

Virtues of Virtualization

Handling large, complex storage systems

One of the buzzwords you'll hear surrounding SAN and storage setups in general is virtualization--or abstracted management of huge amounts of equipment and data. The best way to explain virtualization is by example. Say you have a SAN with 100 disks or RAID arrays of varying sizes being utilized by a Unix server, a Linux server, and a Windows server. Would you want to divvy up all those disks between the servers, unit by unit? Of course not. You'd simply want to know the total amount of storage you have available, then divide it between the servers by need. Virtualization takes all your diverse storage devices and makes them appear as one huge volume.

Three distinct approaches are currently applied to virtualizing a SAN:

The software on the host initiator method typified by Veritas' SANPoint Control

The inside-the-SAN-storage-box utilized by vendors like EMC

And a separate virtualization box between the Fibre Channel switch and the storage network. Vicom is the leading purveyor of this last method.

The software method is referred to as host-based virtualization and requires a control program on every server that accesses the SAN. The advantages are ease of installation and access. However, if you're using servers running different operating systems (a heterogeneous environment in SAN-speak) the software must support them all. If you're planning to add NAS boxes to your network as well as SAN, the software route is appealing since programs that virtualize both your SAN and NAS servers are in the works.

Storage-based virtualization such as that employed by EMC is efficient, however it depends on homogenous hardware--that is, everything from the same vendor. Integrate another vendor's equipment into the SAN and you have to start looking at host-based or network-based virtualization.

Network-based virtualization is the Vicom hardware method, which is in many pre-configured SANs. It's simply another component of the SAN that sits between the Fibre Channel switch connected to the host servers and another Fibre Channel switch connected to the SAN. The biggest advantage of the hardware method is that it could care less what kind of operating system the host servers in front of it are running or what hardware sits behind it. If you have multiple servers running different operating systems accessing a heterogeneous SAN, the hardware method might be your best bet.

Conclusions

How to choose between DAS, SAN or NAS

Know what your needs are: Before buying and implementing a new DAS, NAS or SAN system it's absolutely necessary to accurately gauge your future storage requirements. You don't want to buy a 747 if you're jumping puddles, and a Cessna won't handle a transatlantic passenger service. Take a long look at the amount of data you're currently handling, expect to handle in the future, and then build in a large safety margin. If you're already using DAS and the capacity of your present server hasn't been maxed out, adding a few drives is the simplest, albeit a short-term solution. Adding NAS to DAS is less expensive and far easier to implement than SAN, so it's your best bet until you reach the enterprise level. Then SAN or a combination of SAN and NAS will be more manageable in the long run.

General Tips:

No matter what direction your storage strategy takes you, follow these basic tactics:

Consider vendor reputation and support: Ask the vendor if they have a client you could talk to who has a similar installation. Check around the Internet message boards and newsgroups to see what kind of problems people have had with the products you’re looking at. In other words, kick the tires with a vengeance. With management of storage estimated to be anywhere from 3 to 5 times the cost of acquisition, it pays to do some preliminary groundwork.

Keep your eye on the future: While DAS, NAS, and SAN duke it out, technologies like iSCSI that seek to leverage Ethernet and even the Internet for data transfer are emerging. However, don't let envy of the future dissuade you from purchasing now. Customer pressure is making companies look at the big picture, and integration of NAS and SAN into a common storage 'fabric' is already taking place so you'll most likely be able to leverage your current technology into future solutions.

Factor in upkeep: The sticker price of your storage, whether it be DAS, NAS, or SAN is just the beginning of your costs. As we hinted at above, the Microsoft ad about storage farm humming along quietly without a hint of trouble or human intervention is pure fantasy. Of course that's everyone's goal, but in real life you have to figure human training and supervision into your financial plan or you're in for rude surprise.