Building Bridges Between Islands Of Data Acquisition

In the sophisticated research laboratory of today, scientists have become increasingly dependent on computers as a means of enhancing and accelerating their investigations. A growing number of lab instruments now rely on a computer for their efficient operation; and the modern lab commonly has numerous other computers scattered about to handle such day-to-day chores as data analysis and report writing. While the benefits of computers in the lab are clear, incompatibility among devices from a wide array of vendors can make it difficult to take full advantage of their capabilities. Fortunately, some leading hardware and software suppliers-- primarily the manufacturers of workstations--are moving to address the problem. Linking computers in a lab can definitely boost productivity, but even on the best of networks there's a connectivity hurdle yet to be solved. Let's say, for example, that you've managed to connect the PCs that control two gas chromatographs (GCs) from different manufacturers. Even though you can send data from one PC to another, or to a third computer for storage or analysis, the data that come from one chromatograph are meaningless to the other. In the past, if you wanted to use data from a Hewlett-Packard Co. GC in a GC from Varian Analytical Instruments of Sunnyvale, Calif., for instance, someone had to write a special software interface that would convert Hewlett-Packard data to the Varian data format. "But an enormous amount of energy goes into writing software that takes apart the data and gets it into a format some other instrument can use, " says Vince Dauciunas, product line manager for the Scientific Instruments Division of Hewlett- Packard. The Analytical Instrument Association (AIA), an industry trade association whose members manufacture and supply analytical instruments, is working to remedy this. The organization has a Communication Standards Committee that recently has issued a communication standard for the transfer of chromatography data between instrument data systems. "We brought together 12 of the large chromatography companies, investigated different types of solutions, and settled on one communication standard," says Michael Duff, executive director of AIA. "We tested it by exchanging data between our various members. The data made a lot of round trips--a Dionex [Corp., Sunnyvale, Calif.] GC would write a file to a Hewlett-Packard GC. It would read the file and write it back to the Dionex, and so on. Now we're getting pressure from the user community to do the same thing for mass spectrometry." The communication standard eliminates the need to write custom interfaces between different GCs when transferring data from an older model to a new one, for instance. For more information on AIA's GC communication standard, contact Analytical Instrument Association, 225 Reinekers Lane, Suite 625, Alexandria, Va. 22314-2875; (703) 836-1360. --C.P. In most labs, computers function in stand-alone fashion, creating what might be called "islands of acquisition," analogous to the so-called islands of automation that have bedeviled industry's efforts to implement computer-assisted design and manufacturing technology. While a lab full of nonintegrated computers is surely better than one with no computers at all, the drawbacks to this nonintegrated arrangement soon become evident. Many computers are underutilized, draining dollars from already tight budgets. Also, data must be reentered--manually or from a disk--each time the information is used in a computer different from the one that acquired it. This process is inefficient and error-prone. "Resources for to-day's labs have, at best, stayed constant," says Dave Callender, scientific marketing man- ager at Sun Microsys-tems Inc., Mountain View, Calif. "Labs have to find more efficiency." Problems such as these are being aggressively addressed by a group of computer makers who really know what they are doing when it comes to creating computer networks. Notable among these, judging from recent interviews with researchers throughout the United States, are workstation manufacturers, companies such as Digital Equipment Corp. (DEC) of Maynard, Mass.; Hewlett-Packard Co. (HP) in Palo Alto, Calif.; and Sun Microsystems. In industrial and business environments, workstation manufacturers have faced and, for the most part, met the challenge of connecting dissimilar computers on an electronic network. Now they are eager to bring this expertise to the lab. Why Connect Computers? Linking laboratory computers on a network eliminates duplication of effort and brings a wealth of other benefits, as well. For example, a computer network brings the ability to store data on a computer other than the one that collected it. At the Scripps Research Institute in La Jolla, Calif., this has been "a lifesaver," according to Philip Cohen, director of research computing. "When five or six people share one instrument, its computer disk can fill up quickly," Cohen says. "Having computers linked on a network lets us retarget data to a central file server." Data can also be transferred automatically to a laboratory information management system (LIMS). A network also permits fast and reliable data exchange between labs and processes; no more data re-entry or "sneakernet" (walking disks from one computer to another). Convenience is another plus. With a network you can access any instrument's data using the computer on your desk. A laboratory network makes it possible to combine data from different areas into one computer to produce a final report. For example, at United Engineers and Constructors Inc., a Raytheon Co. subsidiary based in Philadelphia, analysts used to have to go from one computer to another to find certain variables that figured into their calculations. "For an organic analysis on a soil, they had to look up percent moisture and when the soil was extracted and enter that information into their own computers," says Bill Longaker, systems manager. "Having everything in one place is more efficient, and it prevents errors that come when people type in wrong values." Digital Equipment Corp. 146 Main St. Maynard, Mass. 01754 (800) DEC-INFO Solution: Open Advantage/Network Application Support Hewlett-Packard Co. 3495 Deer Creek Rd. Palo Alto, Calif. 94304 (415) 857-5027 Fax: (415) 857-5837 Solution: ChemLAN Sun Microsystems Inc. 2550 Garcia Ave. Mountain View, Calif. 94043 (800) 874-7869 Fax: (800) 873-7869 Solution: PC-NFS Computers that come with instruments are basically single-tasking systems that get bogged down when asked to collect and analyze a high volume of samples. "We use robotics on our instruments so that samples are being injected all night," explains Longaker. "What we're finding is that a Unix [operating system] workstation is much faster than the instrument computer at cranking out calculations. It's better to have files transferred from the instrument computer to the workstation and have the calculations done there. An instrument's computer is good at collecting data, but not good at doing calculations while collecting more data." A network can also save bench space and money. A typical laboratory configuration is one PC per instrument, but PCs cost money and take up space. Some Unix workstations can control as many as 20 analytical instruments. "In our experiments, you have to collect data when it's ready," says Cohen of Scripps. "If you don't have space to store it, it's lost. The network gives us the flexibility to send data to a place where there's room for it. Another problem with PCs is that people don't back them up. If someone has a disk failure, they can lose a year's worth of work. We tell people to put a copy of everything that's important on the central file server that we back up once a week." Some companies are using a lab network to make data available to clients electronically. Clients can log into the United Engineers system, for instance, and download their results. If it sounds to good to be true, it probably is, or so goes the adage. Regarding networking in a lab, even when an outside vendor handles the nuts and bolts of creating the network, someone in- house has to keep it running. Says Longaker of United Engineers, "We wouldn't have done it if we had not had somebody like me to be the full-time overseer. I wouldn't recommend networking to a lab that has just a few analysts and no one who knows anything about computers. But in most cases, there is someone who's interested who takes on the task of managing the system. If a company can afford someone to do that, networking a lab is well worth the effort." Making The Connection The problem with networking, especially in an environment with as much equipment diversity as a lab, is that the different computer systems controlling various instruments--as well as other computers in the lab--have their own data formats and operating protocols. If you have a lab filled with only one type of computer--especially if the computers are Apple products, which have built-in networking capabilities--linking them is as simple as making the physical connection. In the typical lab, however, a network must include computers from many vendors, which increases the complexity of the task considerably. Although it is possible to acquire the necessary hardware and software to create your own network, the benefit of using one of the workstation vendors' solutions is that someone has done the research and made the decisions involved in networking. You don't have to choose among different network services' software or hardware connections, for instance. All you really have to do is tell them the kinds of computers you wish to connect and let them provide the appropriate products. Using a workstation vendor's lab networking solution does not mean you have to replace existing computers with workstations. Workstations are Unix-based, multitasking computers. They are faster than PCs, have better graphics, and are designed to be linked on a network. But makers of workstations know that their products will usually coexist with personal computers and other types of computers. Consequently, the networking solutions they have developed can incorporate existing equipment. Digital, for instance, has a program for laboratories called "Open Advantage." The foundation of this program is Network Application Support (NAS), a set of standards-based software designed to integrate systems and applications from dozens of different vendors. Hewlett-Packard calls its program "The Unified Lab," the heart of which is a package called ChemLAN. ChemLAN includes applications software, services, and hardware to link HP ChemStation operations. (ChemStation is HP's name for its laboratory instrument/computer combinations.) ChemLan is the networking system used to link United Engineers' lab. At this time, Sun Microsystems does not have a specific networking program for labs, but it does have a generic networking solution that is so widely supported that it provides similar capabilities to the lab-specific solutions offered by DEC and HP. Sun's product, Network File System (NFS), has become a de facto standard for file-sharing between computers made by different vendors. "The thing that's really important about NFS is that even though it is something Sun invented, it is an industry-wide standard," say Cohen of the Scripps Research Institute. He equipped his lab's DOS-based PCs with PC-NFS add-in cards and related software to connect them to a Sun workstation that functions as a file server. Although Cohen uses a workstation manufactured by Sun, he could have chosen almost any other Unix workstation since PC-NFS is widely supported. Laboratory networking solutions such as these are relatively new. They go a long way toward simplifying the task of linking diverse computer equipment, but for computers to truly fulfill their promise in lab environments, networking must become even easier. In the meantime, these vendors' programs are building in the right direction. Caren D. Potter is a freelance writer based in Trinidad, Calif.

Building Bridges Between Islands Of Data Acquisition

Interested in reading more?

Become a Member of