Automating network backup in data-intensive environments
Here are some guidelines for planning and executing a solid backup and recovery strategy.
By Allen Herder
Organizations of every size and description depend heavily on computerized information. Server applications, such as e-mail, data mining, and intra/internet, are growing at unprecedented rates. For example, PC and Unix application server markets are expected to grow at a 23% a year clip through 2002, while the file server market will grow 9% a year through the period.
In this information-hungry culture, backup serves three basic needs:
- To protect data against physical loss due to a major facility disaster, resulting from a fire, flood, bomb, or earthquake. Backups are typically stored off-site in secure locations.
- To recover a server that has suffered a hardware or, more commonly, a software failure. In many cases, the priority is to get the server back on-line quickly. Therefore, the backed up data must be readily available onsite. A dedicated or local backup device is usually required to restore the data at maximum speed.
- To restore select files or databases that may have been corrupted by a virus, a hacker, etc. Unless the backup can find the correct tape and restore the specific item, the productivity of the entire business must be reset to the last backup.
A complete backup solution must meet all three of these backup/restore needs. Therefore, selecting a backup plan is a lot more complex than it used to be, when "protecting data from a fire" was the only requirement. Adding to the complexity are on-line capacity requirements, which are increasing at record pace, and an ever-shrinking backup window.
Industry analysts estimate that it takes about four hours each workday to manually back up a medium-sized network using a single tape drive. And many IT managers admit they aren`t sure backups are completely successful. Problems include backing up data to the wrong tape, forgetting to perform backups, and logging entries incorrectly.
In comparison, if the same medium-sized network were automatically backed up to an autoloader or library, the IT staff gains almost a half day to devote to other critical tasks. No matter how busy the department becomes, an autoloader or library never forgets to back up critical user data or to clean tape heads. Data is backed up on a predetermined schedule, whether or not an operator is available.
Further, sophisticated backup software applications and autoloader or library robotics technologies ensure the right tape is loaded for the right backup--preventing a failed backup or valuable data from being erased.
Automated backup means backups are initiated, conducted, and monitored without human intervention. As a result, backups can be performed any time of the day, any day of the week. And should one backup span more than a single cartridge, the automatic cartridge swap relinquishes the need for operator monitoring and intervention, which would otherwise result in a manual tape swap.
The net result: peace of mind and more efficient use of resources.
Taking the right steps
Knowing how to restore backed up data is just as important as backing up data on a regular basis. The purpose of disaster recovery plans (also known as business continuity plans or contingency planning) is to avoid or reduce the effects of failure by creating a plan of action in the event of a disaster. There are four basic steps for creating a backup and recovery plan.
Step 1: Determine the exact composition of your system
List each server that needs to be protected. Show each server`s processor(s), speed, SCSI cards or available slots, LAN HBA and speed, hubs and switches connecting to the WAN, as well as the type of WAN. Don`t forget to list the capacity of the data to be managed and its makeup (see Table 1).
In this example, a company has eight servers at three different sites within the same city. Site 3 is the central site. Each site has a nightly back- up window, Monday through Friday, which lasts four hours, and a Saturday and Sunday window of 6 hours.
Step 2: Define data protection and recovery goals
Using the above example, the goal is to conduct full backups of all data every night to simplify tape management and maximize data protection.
The business averages $50,000 per hour in sales transaction per site, and any server needs to be recovered within 30 minutes to minimize lost business. E-mail must be protected daily from physical disaster, as does the sales transaction database on each server. Remaining data can be protected weekly, if necessary.
Step 3: Determine your hardware constraints
Table 2 lists typical performance factors for server components in a system. Additionally, the sum of the data rates of the standard hubs cannot exceed the WAN speed, and each switched port can run at the maximum WAN speed.
Based on the performance factors listed above and the amount of data to be backed up, the available backup windows for the sample sites are listed in Table 3.
From the previous data, the best solutions for the various sites are as follows:
Site 1: DLT autoloader (DLT7000 drive, 1 drive per library, and 8-tape capacity), with a backup window of 3.3 hours and the use of 1.2 tapes.
Site 2: DLT library, 15-slots (DLT7000 drive, 2 drives per library, and 15-tape capacity), with a backup window of 3.6 hours and the use of 3.75 tapes.
Site 3: DLT library, 15-slots (DLT7000 drive, 2 drives per library, and 15-tape capacity), with a backup window of 3.6 hours and the use of 3.75 tapes.
The use of auto-loaders and libraries allow you to perform unattended backups at night and to maintain on-line recovery access to a full week`s worth of databases and files for selective recoveries. The magazines are rotated off-site every week for data protection.
Step 4: Evaluate current setup
There are also ways of upgrading the WAN to increase the speed of the site-to-site WAN, or you could locate all the servers at the central site and connect them with T1 or greater connections to their clients. But these options can be expensive.
One way to accomplish your backup goals is to place backup devices locally in each site and manage them with remote management software.
Another approach is LAN-free back-up via a storage area network (SAN) architecture. Storage requirements are increasing dramatically, customers need uninterrupted access to their data, and IT managers want to centralize storage. The concept of a SAN is an emerging solution for solving today`s growing storage management issues.
Fibre Channel environments will become fully heterogeneous and less expensive over the next 12 to 18 months. Therefore, the best strategy is to identify backup products that meet your current requirements using existing topology and that will also help you move to Fibre Channel when it evolves to meet your growing needs.
The key benefits of a Fibre Channel SAN architecture include:
- Shared storage: The ability to share storage devices among many computer systems and programs.
- Higher bandwidth: 100MBps; 400MBps in the near future. SCSI is only scheduled to increase from 80MBps to 160MBps, and SCSI is not designed to quickly distribute large files over long distances.
- Centralized storage management: Tools to make sure the backup system is functioning properly, to improve performance, and to allow secure access to certain files.
- Scalable architecture: The ability to easily add storage without shutting down and reconfiguring new storage devices. Fibre Channel connections support up to 126 devices on a loop; SCSI connections are limited to 16.
- Longer distances: Up to 10 kilometers between devices, versus 25 meters for SCSI devices.
- Heterogeneous environments: A loop and storage run on any operating system or machine.
There are many methods to create an effective backup strategy. With a little effort, you can determine which one best meets you storage needs, budget, and organization.
Allen Herder is regional market development manager for Hewlett- Packard`s Information Storage Group, in Loveland, CO.