Linux: a good deal with drawbacks

As Linux servers gain acceptance, the question of how Linux plays on storage networks arises. Not so bad, is the short answer, but not as well as you might hope, either.

This article can also be found in the Premium Editorial Download: Storage magazine: Is it time for SAN/NAS convergence?:

Linux stars on the mainframe
Mainframes may ultimately drive Linux into corporate computing. Linux and the mainframe combine the robustness of mainframe computing with the flexibility of the open-systems environment.

"With Linux on the mainframe, you can quickly run new applications on the Linux instances without risk to your other applications," says John Eilert, co-author of Linux on the Mainframe. A Linux application can fail while everything else continues to hum along.

Colorado State University's computer science department in Boulder is conducting research on virtual Linux servers running on an IBM mainframe. The group put Linux on top of an IBM 9672 model YX6 running the VM operating system. The system is configured with 3TB to 5TB of storage to handle 500 Linux instances. A Linux instance can be a computer science student, a Web server or some other application. The Linux mainframe project is part of the department's ongoing research into DB2 and Linux on IBM's System 390, says Dan Turk, assistant professor.

One problem the researchers found running Linux on VM turned out to be backup, or, more specifically, recovery. "There was a nightly backup of the full VM system with all the Linux instances," says Gregory Wobermin, a student member of the research team. The backup process turned out to be slow and resource-intensive. Although the backup worked fine, restoring specific files on individual Linux instances proved difficult. "We couldn't select an individual file to restore. We would have to restore the entire system image before we could pull out the file we wanted," he says. This took an enormous amount of time and resources.

The solution turned out to be Tivoli Storage Manager. The Tivoli server software runs on the VM layer. The team installed Tivoli client software--approximately 2MB in size--on each Linux instance. With Tivoli installed and running, the researchers did a full initial backup once. After that, they only backed up file changes. "Restore was easy. There is a Web interface and we just select the files we want for each instance," Wobermin says. Installing the Tivoli software on each Linux client was easy as well: You set it up on one instance and then copy it to all the others.

The only hitch comes at the end of the semester when the team has to re-create all the Linux instances to accommodate new students. Few businesses, however, will face a situation of wholesale turnover every three or four months.

With Tivoli in place, the nightly backup now consumes far fewer system resources. The incremental backups are fast, and the restore process is simple and quick. The learning curve, however, was significant. "We knew almost nothing about VM and only a little about Linux," says Wobermin. He adds: "We had to do a lot of reading."
Linux and enterprise storage sounds like an oxymoron. For many Linux users, their storage is direct-attached, with disk capacity topping out at a few hundred gigabytes. But Linux is slowly securing a foothold in enterprise data centers as major vendors such as IBM Corp. and Oracle Corp. throw their weight behind the open-source operating system.

"Today, Linux is just entering the mainstream," says Dan Kusnetzky, vice president, systems software research, International Data Corp. (IDC), Framingham, MA. In the latest IDC Linux survey, 15% of respondents reported using Linux for database applications-- compared to just 6% a few years ago. However, when it comes to mission-critical production applications, Linux still receives little attention, maybe 3%, says Kusnetzky.

From the enterprise standpoint, there's nothing inherent in Linux that would complicate storage any more than any other new operating system. Like the various flavors of Unix and Windows servers already in the enterprise environment, Linux servers can connect through SCSI adapters and host bus adapters (HBAs) to storage resources. Linux servers also run as virtual instances on a mainframe, which handles the storage for Linux. But like any new element in the IT infrastructure, Linux brings its own complications that require new skills, new drivers and new storage management tools. Vendors have made progress, but users still face significant holes in Linux's storage compatibility as compared to other popular operating systems.

Lower costs and flexibility
The appeal of Linux is straightforward: lower costs and flexibility. "You can get Linux on a low-cost Intel box and not pay anything for the operating system," says Mike Karp, senior analyst, Enterprise Management Associates, Boulder, CO. As for flexibility, "you are not locked into any product."

The cost issue alone is a slam dunk for Linux. When BAE Systems Inc., in San Jose, CA, wanted to find a replacement for a Sun Solaris server running a key Oracle application that needed more horsepower, it found that "we could get an Intel box for 40 cents on the dollar compared to Sun," says Daniel Dodge, manager of software infrastructure for the aerospace and defense manufacturer. The company saved even more money opting for Red Hat Linux rather than Windows Server 2000.

But Linux isn't without its drawbacks, Karp concedes. "Nobody owns Linux. With Linux, things are a little ad hoc," he says, referring to the lack of a major vendor to take responsibility for the operating system. Many Linux vendors are small and young, although this is less of a concern with open-source products where you have full access to the source code. Many open-source tools--although free--offer only an informal support community rather than the comprehensive support infrastructure IT managers have come to expect from proprietary tools vendors.

The support of major vendors, however, is a stabilizing factor for Linux. "Oracle is a big proponent of Linux, and that was very important to us," says Dodge. BAE stuck with major vendors right down the line in its Linux implementation: Dell Computer Corp. (servers and storage), Oracle (database) and Veritas Software Corp. (storage management).

Still, users who want to deploy Linux and enterprise storage do encounter some problems. For example, leading enterprise storage vendors have been slow to provide the latest drivers for Linux. "There are still no EMC drivers for the latest version of SuSE," says Stevan Townsend, manager of database administration, Tractor Supply Co., Nashville, a retail farm and ranch store with more than 400 locations in 30 states. At its central office, the company runs an Oracle database for its SAP applications, a sales and audit application and a data warehouse on a Linux cluster running the PolyServe Linux clustering software. It stores the data on a 14TB EMC Symmetrix storage array. EMC is not the only slowpoke when it comes to keeping up with the latest Linux releases. "QLogic and Veritas are slow to deliver agents," says Townsend, but he adds that they all eventually come through. Tractor Supply just ordered an EMC Clariion array for its Linux cluster.

Another potential problem for enterprise storage users is the 2TB logical volume limit of the Linux kernel. "This was an issue early on, but it will go away," says Ranajit Nevatia, director of Linux strategy, Veritas, referring to the next major Linux kernel release, V 2.6, which will eliminate the 2TB limit. A test version of the new kernel, 2.6test, has already been released to developers. The general release could happen later this year or early next year. But even without the next kernel, "there are a lot of applications that fit within 2TB," he says.

True enough, but DreamWorks SKG, Glendale, CA, a film production and distribution company running Linux clusters in a large mixed environment along with several enterprise storage area networks (SANs) built around IBM ESS and Hewlett-Packard storage arrays, can look ahead and see the 2TB limitation becoming a problem soon. "Our largest file today is 600GB, but our digital asset management system will hit 2TB next year," says Rick Garcia, senior system administrator. The immediate fix is to split the file, but that creates more management work. "Linux 2.6 will solve the problem," he says. The only question is whether Linux 2.6 will arrive soon enough.

Pros and cons of Linux storage
  • Low commodity pricing
  • Readily available free tools
  • Freedom of choice, no vendor lock-in
  • Supportive Linux community
  • Not many storage tools available
  • Slow to get new drivers
  • 2TB logical volume limit
  • Another operating system to learn

Limited storage support
Maybe the biggest problem facing enterprise storage managers is today's limited tool selection for Linux storage. For example, Veritas offers a clustering file system for other platforms, but the Linux version isn't expected until early 2004, Veritas' Nevatia says.

Many of Computer Associates' (CA) BrightStor enterprise storage management products have been ported to Linux. However, CA's SAN Manager and SAN Designer have not. But there is a workaround, says David Liff, vice president, BrightStor brand marketing: "Most companies don't have Linux-only SANs," but companies can manage Linux on mixed SANs using Windows or Unix versions of the CA tools along with its Linux agents.

The Linux storage tool picture is improving seemingly weekly. For example, PyX Technologies, El Cerrito, CA, has developed both client and server software based on Linux for iSCSI-enabled storage. The product--which made its debut in August--fully conforms to the iSCSI standard as ratified by the Internet Engineering Task Force (IETF), according to PyX. Its iSCSI software is being bundled into products from ASA Computers Inc., Database Appliances and Pogo Linux Inc.

Instead of waiting for tools from major storage vendors, many Linux administrators manage Linux storage with tools--often free--from the open-source community. Administrators simply download tools such as MRTG, which tracks and graphically displays network conditions, and Samba, which provides file and print services to server message block (SMB) clients, right off the Internet. "There is a lot of freeware out there," says Eric Wright, an independent Linux consultant, Annapolis, MD, and creator of The Linux Spot Web site. Other online sources for tools are: and

The Linux Spot runs on an IBM Netfinity Pentium server with an IBM ServeRAID Ultra 160 SCSI RAID 5 controller and a Compaq ProLiant external RAID array. At one point, The Linux Spot ran almost 1TB of direct-attached SCSI storage, although Wright has cut it back as he shifted the site to primarily FTP serving.

Many small Linux sites still get by with internal disk storage or a small directly attached array. Typical is, which provides online divination, readings and astrology. "We just use internal storage, no NAS, no SAN," says Eric O'Dell, senior database administrator.

The company keeps its two million record customer file on a MySQL database running on Red Hat Linux. It has a generic Intel server with dual Zeon processors. The company is only now focusing on storage with the acquisition of its first external RAID array. Two database servers will share a 45GB partition, leaving 90GB of unused capacity for growth, O'Dell explains. The biggest decision: which shared file system to use, Sistina's GFS or ext3. After trying both, the company settled on ext3, a popular journaling file system. "We were experiencing a [database] consistency problem and ext3 resolved it," O'Dell says.

Few companies have converted their large enterprise storage infrastructures to Linux or are likely to do so, at least anytime soon, suggests IDC's Kusnetzky.

Increasingly, however, organizations are adding Linux systems into their environments, often starting with a DNS or Apache server and expanding from there. In the process, they are incorporating Linux servers into their storage strategies. Storage vendors are closely monitoring the trend. HBA manufacturers, management software providers and other storage vendors are slowly supporting Linux. It may not be the operating system they address first, but Linux is definitely on their to-do list.

This was first published in September 2003
This Content Component encountered an error



Enjoy the benefits of Pro+ membership, learn more and join.



Forgot Password?

No problem! Submit your e-mail address below. We'll send you an email containing your password.

Your password has been sent to: