By R. Todd Thomas
When I started at Austin Radiological Association (ARA) about five years ago, its server “farm” consisted of just one RS/6000 server running AIX that ran the company’s mission-critical application, along with a Windows box that was set up to run MS Mail (although no one ever used MS Mail).
Backing up at that time was simple—there was an 8mm tape running from a drive internal to the AIX server. The backup routine wasn’t kicked off from within AIX but from within the application itself. Even then, backing up the application, data files, and OS took roughly seven hours.
|Sponsored by Compaq|
After we started deploying NT, we needed a cross-platform backup solution that could back up multiple servers. Using our relationship with a large systems-management vendor, we acquired and deployed a tape-based solution that employed host “agents” that communicated backup data across the network to the backup server.
When we first deployed this solution in year two of my employment, it worked quite well: one tape, multiple servers, done by morning. Open-file agents allowed us to start the process earlier in the evening, and even though backups were now taking 11 hours to complete, the centralization of backups probably saved us operational overhead in the way of support and supplies.
Eventually, we outgrew this solution, and we needed a storage area network. We opted for a SAN instead of the less costly option of network attached storage devices. In this article, I’ll describe how my organization benefited from a SAN and why I believe it was the right choice.
When problems began
Things are much different today from the way they were five years ago. We now run the following:
- Two RS/6000s
- Nine Netfinitys running Windows NT (two of which are Windows 2000 servers)
- Five Proliant Windows NT servers
In addition to the mission-critical application, IT now supports other “critical” applications, including:
- An Exchange environment
- An intranet site
- Accounting and finance
- Systems and network management
- Asset and policy management
In this mix are included SQL 2000 databases, a SQL 6.5 database, and a SQL 7 database. Backups—using the same solution noted above—were taking anywhere from two to four days, depending on the day of the week. (We used a GFS—grandfather/father/son—rotation scheme.)
Production was starting to be affected because host CPU cycles were now being used during the workday to facilitate the agent-based backup. On the horizon was the deployment of our Internet site and the installation of an Oracle-based application. More servers, more data, more time.
Our company is owned by the radiologists, who interpret and report on various types of radiological exams. We have easily grown 20 percent per year since I started there. One of our biggest problems is the management of X-ray film. In seeking a solution to this ever-growing problem, healthcare systems are turning to PACS (picture archiving and communication system).
Turning film images into “soft copies” requires space—lots of space. Federal and state laws require some films to be stored between 7 and 20 years. Add even a terabyte’s worth of image data to our application data mix, and our standard backup routine starts to look woefully inadequate. The idea of restoring a terabyte’s worth of information in the event of a disaster from a tape-based backup solution scared me.
Add to that the fact that my company was looking at becoming an SP, or service provider. Since our company would be responsible for providing specific services to healthcare partners, IT would have to factor in uptime guarantees, 99.999 percent reliability, and quick recovery times.
SANs to the rescue
When the industry started to talk about SANs a few years ago, I knew I had the solution to my problem. A central data repository that could be backed up to tape at leisure without affecting production. While the idea of long backup times still bothered me, I could see the potential of SANs.
With this solution in mind in preparation for PACS, I began a yearlong sales cycle in 1999 with the largest storage vendor in the country. While I looked at other SAN offerings, this company had the hardware reliability and the software to make the idea of tape backups obsolete. Utilizing a combination of this company’s hardware, software, and professional services, we deployed our SAN with full backup capabilities to disk in six months.
We now back up our entire server farm in 45 minutes. You read that right—45 minutes. While this does not account for the Oracle deployment or any image data at this time, we certainly have room to grow. Restore operations take a few minutes, and recently, we did a complete Windows 2000 install to a server in less than 10 seconds.
With every silver lining, there has to be a cloud—this solution was not cheap. Yet the ability to back up in less than an hour, the data explosion we were about to face, and the need for five-nine reliability—both in storage hardware and in server downtime—drove my company’s decision to move forward with a Cadillac solution.
Fortunately for the IT professionals at the company, the board also viewed this as an investment in growing its business by pursuing other means of income in the healthcare arena: ARA has begun thinking about becoming a vertical service provider.
Reviewing the decision
Deciding on a multimillion-dollar solution was not easy—especially with the ever-lower costs of network attached storage devices. Our considerations in designing a SAN included all of the following—some of which factored in favor of a SAN, some against.
- Robustness of the storage software
- Reliability of the storage hardware
- Vendor market strength
- Business objectives
- Backup and restore times
- Management of available storage pools
- More complex to use
- Staff time and training
Other questions to ask
Along with the pros and cons listed above, you should ask yourself other questions when considering a SAN. In terms of the OS and applications installed, ask yourself the following:
- Will you run your applications directly from the SAN or continue to run them from host-based storage?
- Will your servers boot off the SAN or internal disk?
- Is there application functionality such as database replication that could benefit from a storage network?
You should also consider the frequency of backups:
- How often are backups needed and how long does each one take?
- What length of time for backups would be convenient?
Finally, give thought to the type of data being stored (images, video, text, etc.):
- How big and how many are the individual files?
- Would the offload of the file format onto the network benefit from a high-speed storage network?
- How quickly do you need to restore data in the event of data loss?
The quick results seen with some of the functionality of the SAN have helped our administration realize the value of its purchase. We initially invested in a SAN to decrease backup time. But we discovered that this solution means much more. It may have a tremendous positive impact on business development because it has given the company the ability to become a vertical service provider. What started as the solution to an IT problem became an important investment in the future of the business.
R. Todd Thomas is the Director of Information Services for the Austin Radiological Association.