Enterprise storage - it's SAN-tastic
- By John McCormick
- Jun 22, 2005
Network Appliance FAS900
Whether you're part of a large agency that has long dealt with massive storage requirements, or a smaller agency just beginning to buckle under the weight of accumulated data, the storage area network'in all its increasingly varied flavors'remains a popular solution. These days, however, the kind of SAN you install depends on your needs, budget and expertise.
A SAN makes large amounts of data (into the hundreds of terabytes) available to network users. Traditionally, SANs have been expensive and complicated to set up, but that's changing.
Two recent major developments have been the arrival of turnkey SAN systems and the popularity of 10-Gbps Ethernet networks'and 40-Gbs Ethernet is on the horizon. The first is important because SANs, particularly Fibre Channel SANs, often present interoperability issues that make them hard to assemble and manage. This fact can be a deal breaker for small to mid-size organizations that would like a better way of handling all their data but don't have extensive dedicated IT resources. Entry-level turnkey systems make it possible for these agencies and departments to know in advance how much the SAN will cost to implement, that it will work out of the box, and that it will cover their storage needs now and in the future.
In addition, the increasing popularity and relatively low cost of 10-Gbps Ethernet has made it possible to build inexpensive SANs based on IP and SCSI protocols, specifically iSCSI [see 'NASA, we have cheap storage,' GCN, May 23, page 27]. So-called IP SANs use mostly off-the-shelf commodity parts and can be designed to achieve performance comparable to'or in some cases, greater than'expensive and difficult-to-implement Fibre Channel SANs.
New disk drive technology is another reason SANs are becoming less expensive and more attractive to smaller enterprises. SCSI hard drives are the traditional way of providing high-performance storage, but today's SANs use less-expensive Serial-ATA (SATA) hard drive arrays.
Besides offering higher performance, SCSI drives have historically been popular because they're more reliable. But that criterion is losing importance as SATA achieves better'albeit not SCSI-level'mean-time-between-failure ratings. SATA drives, many of which carry a 400,000-hour MTBF rating, could still run for 40 years. SCSI drives are rated for 1 million hours or more'better than twice SATA'but how long do you plan to have your new SAN? By opting for SATA drives, you could double or triple your capacity for the same price, and given that storage is usually RAID-protected, cost is probably a more significant consideration than the reliability of SAT versus SCSI.Fibre Channel versus iSCSI
At the heart of today's SAN decisions is the choice between Fibre Channel and iSCSI. Both are block storage protocols. The main difference is that Fibre Channel combines SCSI commands, status information and data in Fibre Channel frames, while iSCSI does the same thing over standard IP packet networks.
Fibre Channel can be implemented over either fiber optics or copper wire, but it requires a special switch. Particularly in smaller organizations, the additional $2,000 and up required to connect a server to a Fibre Channel SAN can make the technology less attractive. Connecting storage via iSCSI, however, can be as simple as adding a free iSCSI device driver to a server's Ethernet card. In fact, some Fibre Channel SAN islands are now being interconnected using IP networks.
As for the performance advantage of Fiber Channel, it's true the technology is faster than most of today's iSCSI-based SANs. But remember that iSCSI can run over 10-gig Ethernet. If you've got it (for now, a big 'if'), you could achieve better performance from iSCSI than from 1-Gbps or even 4-Gbps Fibre Channel.
Here are some other considerations for the iSCSI versus Fibre Channel decision:
- If throughput demands are low enough to permit use of standard network cards, iSCSI can be one-third the cost of Fibre Channel.
- Fibre Channel is yet another new technology workers must learn, while iSCSI is Ethernet- and IP-based.
- Fibre Channel experts are few and far between, while IP network technology is common.
- Fibre Channel requires a centralized location and is distance constrained, but iSCSI can be located anyplace that has Internet service.
- Fibre Channel has a higher capacity than iSCSI over basic Ethernet, but for 5 to 6 simultaneous users the difference isn't important.
- Fibre Channel components are usually more difficult to integrate into existing IP-based enterprise networks.
In short, iSCSI has many advantages for new installations, and new iSCSI-based SANs can interface with legacy Fibre Channel systems. That said, data centers already us- ing Fibre Channel have little reason to convert those SANs to iSCSI.When you're in the market for a SAN, ask yourself these questions:1.
Is the system easily scalable?2.
What is the average downtime per year?3.
How will I secure it?4.
How easy is it to integrate hardware and software from other vendors?5.
Can I use it with emerging technologies? In other words, can the management tools accommodate new standards and protocols?6.
Will I be able to find qualified technicians to install and support the network?7.
What are my actual throughput needs? Can iSCSI and SATA drives provide the necessary performance?
The Fibre Channel versus iSCSI debate will remain lively for the near term. The deciding factor may turn out to be the skill set (IP and/or Fibre Channel) available in your IT department or your contractor's. Expanding an iSCSI-based IP SAN will almost always be more affordable, and ever-faster Ethernet connections will close the performance gap between the leading SAN protocols.
John McCormick is a freelance writer and computer consultant. E-mail him at [email protected]