NEW YORK — The two popular misnomers with the storage segment of IT is that network-attached storage (NAS) is a networking technology based on Ethernet while storage area networks (SAN) rely on fibre channels. Storage networking can even run over “scuzzy” — small computer system interface (SCSI) — albeit cannot be deployed as easily or cost-effectively, according to a leading expert.
“SAN really is just a network of SCSI devices. Just like IP, it can go over different data transports,” said Jacob Farmer, senior consultant at Boston-based Cambridge Computer Services Inc. Farmer was the lead-off speaker of SAN Summit, which on Monday marked the start of the TechXNY 2001 trade show at the Javits Center in Manhattan.
SAN Summit kicked off with two well-attended conferences on storage technology. The crowds directly reflected the amount of interest in the IT segment, given all the research that projects storage as the single largest IT investment that many companies will make this year.
To differentiate between NAS and SAN, Farmer explained during his “Introduction to SANs” seminar, NAS is a system that simply transfers files between two or more host computers, which he referred to as “initiators.” A SAN, meanwhile, transfer information in blocks rather than files and need to rely on a file system to organize the blocks into files.
But what many IT managers, directors and chief information officerss fail to understand is that storage networking isn’t as simple as plugging two or more CPUs to a hard-drive or DAT. For example, if two Windows NT machines are plugged into a one tera-byte hard-drive, the OS on both machines will each send its own respective signature files and repeatedly try to re-format.
In order to effectively deploy a SAN, you need to partition your SCSI device and create redundant connections to back-up that I/O processing, Farmer said. For that reason, SCSI-based networks fall short when compared with fibre.
“It doesn’t matter what you connect…what matters is what you run over that connection,” Farmer said.
A hard drive or DAT can be partitioned so that certain portions of the drive can be assigned to a specific CPU. However, the best way to partition is to assign a portion to a device based on its Logical Unit Number (LUN), which is just a subset of the SCSI device.
“If a SCSI ID is a street address, then the LUN is like an apartment number,” Farmer explained.
Unfortunately, the switch at the center of a SAN directing the I/O processing doesn’t recognize LUN, meaning the partitioning needs to be created at the host level or on the device.
Anyone who missed Monday’s seminar can register for a free, three-hour course explaining the basics of SAN technology here on the Cambridge Web site.
