Workshops

Break Down The Barriers Of Large Volume Backup

So the question remains: "How do I back up all of this data?" Network managers face the challenge of not only backing up one 100-GB server, but multiple servers scattered across an enterprise. To make matters worse, the window of opportunity to back up these servers has been reduced because of worldwide operations and increased working hours. How can you back up more data stores in a limited window of time?

Through our partnership with an international energy company in the San Francisco Bay area, we evaluated a number of strategies and techniques to break down the barriers o f large volume backup. Although most of what we learned is common sense and good management practice, we found that there are a n umber of ways to streamline the problems associated with enterprisewide backup processes.

The Static Choice The task of securely backing up large amounts of data can be staggering. Before you dive into your backup, you must determine the makeup of that data. In many cases, data is static and rarely changes. Such data includes online manuals; reference materials and images, including scanned documents; and graphic elements.

Additionally, be on the lookout for users who place large amounts of data on the server so that it's easier to access. We've been guilty of this and have been known to back up an entire client hard drive to the server. Yes, it's a great way ensure you have a backup of your laptop, but such procedures could cause havoc for your systems if everyone did this. Additionally, the contents of an entire CD-ROM could easily be copied to a server to improve access time. Ideally, static data should be moved to a CD-ROM server or an optical jukebox or, if speed is still an issue, to a separate disk farm or a separate disk volume on the server.

To create these CDs, we've used a number of products including units from Hewlett-Packard Co., Philips Electronics N.V. and Pioneer New Media Technologies. They're available for just under $1,000 and let you burn your own CD-ROMs. In addition, CD-ROM servers, such as Axis Communications' Axis StorPoint CD, allow the CDs to be available on the network as a standalone device (see "StorPoint Doesn't Put All Its Eggs in One Basket," December 1, page 52). StorPoint supports Novell NetWare, Microsoft Windows NT and Unix (via the Network File System) networks. StorPoint also can be used as a Hypertext Transport Protocol (HTTP) server, allowing CDs to be accessed via a Web browser. By removing your static data from the server, the amount of data to back up can be significantly reduced.

Another way to reduce the quantity of data to back up is to use hierarchical storage management (HSM) software. HSM solutions are available for most network operatin g systems, including Banyan VINES, IBM OS/2, NetWare, NT and Unix. HSM takes advantage of the strengths of various media types, such as the high capacity of tape and the low cost and high performance of hard disks. However, if the software isn't tightly integrated into your backup process, this option may be difficult to accept.

HSM systems migrate infrequently accessed data to less expensive media types. By placing infrequently accessed data on secondary and tertiary storage mediums, such as optical and tape, you can cut down on backing up static data. Migrating this data comes at a price, however, as users will experience slight performance hits when accessing data from tape or optical media.

Also, beware that while HSM is attractive on paper, it's complicated to install and configure. If used correctly, however, HSM can reduce the amount of data being backed up.

Tuning Up Your Backups You can take several steps to increase the performance of data backups. First, if your backups are pe rformed over the network, consider creating a separate backup backbone. Although placing file and application servers near users makes sense, it does not facilitate the backup process (see "Un-Optimized Enterprise Network Configuration" on page xxx).

A second backbone dedicated to backups separates the everyday traffic from that generated by backups, eliminating the potential for backups to slow the production network (and vice versa) and enabling backups to be performed during normal working hours (see "Optimized Enterprise Network Configuration".

Technologies such as FDDI are popular when used in this capacity and can easily support high-bandwidth backups. Since FDDI is a token-passing mechanism, it can support a high sustained throughput. On the other hand, technologies such as Ethernet's Carrier Sense Multiple Access with Collision Detection (CSMA/CD) begin to fade at about 40 percent utilization. FDDI and other token-based networks perform well under load. In our tests, we've seen token-based te chnologies move data with utilization as high as 65 percent to 70 percent. Additionally, many companies have turned to proprietary solutions, such as the Thomas-Conrad Networking System (TCNS), and to more traditional fast network technologies.