What you will learn from this tip: The cost of an iSCSI backup system can range from practically nothing to as much as a Fibre Channel SAN. Here's what you need to know to pick the right price-performance point for your enterprise.
One of the most attractive things about
The basic iSCSI protocol, as defined by the Internet Engineering Task Force (IETF) in 2003, is simple: Just wrap SCSI commands and data in TCP/IP packets and send it over Ethernet. There's one issue though: All that wrapping and unwrapping requires CPU cycles and other overhead. While most operating systems today, including Windows, Linux, Solaris and Netware, come with server-side support (called 'initiators' in iSCSI-speak) for iSCSI, a number of companies also offer much faster hardware-based products using specialized processors called TCP/IP offload engines (TOEs) to handle the packetizing and unpacketizing. If you need a really elaborate setup, companies like SANRAD offer iSCSI SAN switches that let you configure full-blown SAN networks.
If you can get away with using your existing Ethernet LAN and software initiators, you can have an enterprise-wide backup system for the cost of the backup device, such as a tape drive, and the cost of integration. If you need a high-speed, high-performance system with maximum throughput and minimal interference with your existing network, you can put together an elaborate disk-to-disk-to-tape (D2D2T) system using TOEs and a separate 10 gigabyte Ethernet network, for example. More significantly for most storage administrators, you can also get just about anything in between.
The performance of an iSCSI backup system varies pretty much as a function of the price. That means for storage administrators considering iSCSI backup, the big question is: How much performance do you need to support your backup needs?
Fortunately, that's a quantifiable question. Unfortunately, the answer is likely to be different for almost every user.
The major variables are network load and available CPU cycles on the server. Backup involves moving large amounts of data from storage to the backup device as quickly as practical. This puts a concentrated load on the network. The packet handling overhead in iSCSI requires a lot of processing power from somewhere. If you have a server with horsepower to spare (typical) and a network with a good deal of excess capacity (not so typical) you can add iSCSI backup fairly inexpensively. The less spare processing power and network bandwidth you have -- or the more performance you need out of your backup product -- the more expensive iSCSI backup will be.
Both server and network loads are easy to determine with basic management tools. Just be sure to collect statistics over a long-enough period to accurately cover spikes in usage, such as the one most companies experience at the end of a billing cycle.
If you don't have enough bandwidth to handle backups without slowing down your LAN unacceptably, there are two major ways to handle the situation. The obvious one is to use a separate network for some or all of the devices being backed up. Since Ethernet is cheap, this is a popular approach. The other common way to go is to beef up your backup's throughput by using a D2D2T backup built around inexpensive SATA disks. The D2D2T approach stretches your backup window and allows you to reduce the load backup puts on the network.
Lack of processing power to handle the iSCSI protocol is probably best dealt with by using iSCSI devices containing TOEs. A number of companies offer such products, including Alacritech Inc. (controllers) and Qlogic Corp (HBAs).
For more information:
Solving the riddle of data storage with iSCSI
About the author: Rick Cook has been writing about mass storage since the days when the term
meant an 80 K floppy disk. The computers he learned on used ferrite cores and magnetic drums. For
the last 20 years, he has been a freelance writer specializing in storage and other computer
This was first published in January 2006