By Gael Core
IS personnel dream of the
day when they'll be able to centrally manage heterogeneous
operating systems, but until that day they have to handle them one at a time.
How many network managers wish that their organization's network were comprised of uniform systems? It might not make good business sense to be tied to one architecture or set of systems, but it would be a reasonable task to manage one operating system across a large enterprise. Unfortunately, that is not the reality today, so businesses grapple with the administrative headaches of backing up files, managing user accounts, scheduling jobs and providing security across many different hardware platforms and several operating systems. In reality, it's a segregated world of systems, and it's going to remain that way for the near future. According to International Data Corp. of Framingham, MA, 36 percent of businesses surveyed recently have two or more midrange (or larger) server platforms and operating systems. That requires two or more efforts to manage them.
Underlying this issue is a core set of problems that have to be addressed and solved before multiple operating systems can be managed from a single platform. While recent World Wide Web management efforts offer some glimmer of hope, there have been more obstacles than progress toward solutions for several years. Among the problems are a general lack of standards for interoperability, tools that offer only point management solutions and the inability of management solutions to scale upward and meet the demands of large networks. "The number one item that people trip up on is the issue of scale," says Asa Lanum, a principal with the CTO Group of consultants based in San Francisco. "The scale gets out of control rapidly and has a dramatic impact."
Complicating the ability to set up automated management routines are the very natures of the different systems being managed, Lanum adds. Among problems that make it difficult if not impossible to manage multiple systems in large networks are three key issues.
This last point figures prominently in most computing environments. The growing diversity of information systems is driven by a computer industry that is feverishly competitive; every few months new technology is introduced, and companies buy into the technology usually to gain a competitive advantage. That frenetic pace seriously impacts enterprise management work as more vendors' technology gets introduced into networks.
At a fundamental level, it's difficult to get operating systems to interoperate with one another. Largely that's due to each OS having its own unique design characteristics, from underlying kernel architectures to system calls. "You don't get operating systems to interoperate. What you do is get them to coexist," says Judith Hurwitz, president of Hurwitz Consulting Group in Watertown, MA. "You're not going to manage different operating systems as though they were one. End-to-end systems management across all operating systems just doesn't exist."
For years, vendors have paid at least lip service to efforts to develop standards that could improve interoperability between their systems. Most notable was the Open Software Foundation's work to create standards-based systems management frameworks, called the Distributed Management Environment (DME). But vendors' insistence upon their own particular technologies derailed that effort. Still, users and analysts recognize that standards are the basis for making disparate platforms communicate and share information.
"The fact is we're not even close to having all our functionality in a single operating system environment integrated, let alone doing this in a mixed environment," says Michael Goulde, an analyst for the Patricia Seybold Group in Boston. When it comes to trying to manage the operating system environment, there are some preliminary attempts to provide consistency, but "it's still primitive," Goulde adds. That is especially true when users are trying to manage several operating systems, such as Unix and Windows NT, across multiple hardware platforms. But coming up with management standards is difficult, because there has to be tight integration between the operating system and the hardware.
With management standards in limbo, it's difficult for systems administrators to keep many different server platforms maintained and running. Tasks that have to be done on a regular basis include backing up or migrating files, deleting temporary files and cycling log files. Also, files have to backed up, security maintained and jobs scheduled across these different platforms.
Today each separate task is accomplished using point solutions. In Unix environments, system administrators write Unix shell scripts or Perl scripts to automate these tasks, but that approach has its limitations. Scripts have to be rewritten when new systems are added to the network or when new management routines have to be added to the mix. Writing these scripts also requires experienced programmers, who may be hard to find and costly to employ.
This search for the holy grail of managing multiple platforms has led many vendors full circle back to mainframe environments, which they are trying to imitate in a client/server world. Centralized mainframes offer automated management that is yearned for by users trying to manage distributed server operating environments, such as Unix and Windows NT. As a rule, the more distributed the environment, the greater the cost. It took many years for mainframe tools to evolve and become refined. But those high-end resources are often an order of magnitude more expensive than, for example, those same resources offered on Unix systems--if they are offered.
Most currently available tools have had little success in overcoming the problems of trying to manage multiple operating systems. That's because they've been designed to address specific kinds of management tasks, such as backup or job scheduling. "There are products out there that address these issues, but as far as having a consistent environment with something as simple as user accounts, forget it," Goulde says.
One approach to the challenge is to use a management framework to bridge these problems. CA-Unicenter from Computer Associates International, Hewlett-Packard's OpenView and IBM's Tivoli Management Environment have made strides in this area. Most of their success has been in managing multiple versions of the Unix operating system.
A strategy users and analysts suggest to deal with shortfalls in management functions is to use a framework and then write programming scripts to automate specific tasks. "The best you're going to do is find scripting languages that run cross-platform, and you're going to have to select vendors that are committed to providing systems management products cross-platform," Goulde says. Again, that strategy takes expert programmers to write the scripts and change them when tasks or systems in the network change.
A pure Unix network provides distinct advantages not realized in a more diverse environment. Repetitive tasks of running workload scheduling, tape management and event processing on multiple servers can be accomplished. Management frameworks can make these tasks manageable. "Without CA-Unicenter, it would be a monumental task to manage those jobs," says Bob Luti, systems manager for operations at Parametric Technology Corp., a developer of CAD/CAM software in Waltham, MA. "We would have to go into each individual server to make sure that the backups have been successful and that our work on that processing has been completed." He says his company saves valuable administrative time and money using the automated facilities within the framework. But Luti anticipates that, like many other IS shops, his also will be working with Windows NT.
"Interoperability between Unix systems has always been there," says Wayne Fowler, an independent consultant in Markham, Ontario, Canada, who previously worked for the Toronto Stock Exchange. "Now there is a new player in the mix, and that is Windows NT." Today, users struggle with managing both Unix and Windows NT server environments, as NT has increased its share of the departmental server market.
There are no products or tools that let a system administrator manage both of these operating environments. And to further complicate the situation, Windows NT doesn't offer the breadth of management features found in a Unix environment. "NT is much less open and significantly less mature," Fowler says. "Vendors will have to gain more experience" to make NT compatible with Unix systems.
Systems managers keen on automating management tasks across Unix and NT shouldn't hold their breath. It will be some time before this problem gets resolved. "We haven't found any tools that cover all the environments we need to cover," says a large retail user who declined to be identified. When it comes to managing NT systems and Unix, his group is pursuing a strategy that formulates a comprehensive distributed system, which is built partly around a mainframe.
One management option is to make a Unix server appear to users as an NT server on the network. In fact, that approach has been used by AT&T in its Advanced Server for Unix. AT&T has designed its Advanced Server for Unix management functions to closely mirror those found on Windows NT Server. By doing so, IS personnel can move between the two management environments without being jarred by differences in pull-down menus and associated functionality. But when it comes to running Unix applications on an NT platform, IS shops have to port them to the other platform.
Microsoft's Systems Management Server (SMS) is designed to manage only Microsoft platforms. There are, however, ways of interfacing with SMS. It is structured around an SQL Server database, which means that other applications can interface with SMS and access configuration information collected by the management agents embedded in Windows and NT. Yet that doesn't help systems managers trying to automate tasks across Unix and NT.
Another promising strategy being used in NT networks is clustering. Users are optimistic that clustering technology will greatly aid management tasks, because systems administrators can view clustered systems from a common user interface. Best of all, such systems appear as a single system to managers. Microsoft is expected to exploit clustering technology not only to gain computing benefits but also the inherent management capabilities. "Clusters are making it simpler to manage lots of systems," Goulde says. The problem, he adds, is that they all have to be running the same operating system.
Part of the problem of past standards efforts has been that they've tried to create a standard specification that addresses all present and future needs. In the case of DME, that led to including both a procedural and an object-oriented set of platform interfaces, which produced further fractious debates. Unlike these past attempts, recent Web-based management efforts have some analysts cautiously optimistic. That's mainly because of the Web's graphical browser interface, which runs on most operating systems and platforms.
Recently, a group of hardware and software vendors banded together into the Web-Based Enterprise Management (WBEM) alliance to create standard specifications intended to make management tasks easier across different platforms. The specifications are based upon existing Web standards, such as Hypertext Transfer Protocol (HTTP), Hypertext Markup Language (HTML), Simple Network Management Protocol (SNMP) and the Desktop Management Task Force Interface (DMI; the DMTF is a cooperative industry effort aimed at making the management of desktop PCs and servers easier). WBEM members, which include BMC Software, Cisco Systems, Compaq Computer, Intel and Microsoft, claim they will succeed where efforts like DME failed, because instead of introducing new standards, they will leverage existing Internet technologies.
Specifically, one proposed data model, called HyperMedia Management Schema (HMMS), which runs over HTTP, could make management functionality consistent across different implementations. HMMS is a hierarchical, object-oriented data model for framing information. Within that model, for example, any device or application--as well as standards such as SNMP and DMI--can be represented in a consistent format.
Management applications built with HMMS should be able to tap into common data repositories and share information. Vendors that comply with this specification will have a consistent model by which to share management data across different platforms and operating systems. Also, users will be able to take advantage of integrated management tools that span applications, systems and networks.
With these Web management standards in place, system administrators are expected to gain a more flexible way of managing networks, systems and applications. Network administrators may be able to save time by providing a common communications exchange that works with any Web browser. A single browser interface will ease access to management data on several different networks, including Unix, MVS, VMS, Windows NT and Novell NetWare. HTML applications will link end users on browsers to the varied resources residing on the Internet. Web browsers will also provide systems administrators with a method by which they can check the accounting on the Web server to determine the best time to schedule downloads of data. "This idea of coming up with some standard interfaces so that a browser can go against a server is a real glimmer of hope," Goulde says. "It is a low level of cross-platform management, but it's a start."
Products based on HMMS are expected to be available by the end of next year. Most observers expect it to be a few years before a comprehensive management solution based on HMMS will be available. The first wave of Web-based management products will permit managers to view reports but not make changes to devices and servers on the network.
However, some industry observers are skeptical of HMMS and its promise to resolve cross-platform management issues. One of the problems is that HMMS is based on HTML, which may limit interactivity and intelligence of management applications built upon HMMS. Some analysts contend that other technologies, such as SunSoft's Java, may be able to overcome those intelligence and interactivity shortcomings. SunSoft's management efforts overlap with some of the work of the WBEM alliance.
In any case, Web-based management technologies have a long way to go before they're proven. Initial products are about a year off, but most expect it won't be until 1998 that mature products are delivered.
Operators and administrators are already taxed with trying to manage more systems than they're capable of dealing with. In terms of managing NT servers along with Unix platforms, few companies will be eager to have yet another desktop client system to manage those servers. Each additional desktop management station requires additional personnel to run and maintain it.
In the meantime, companies are looking to vendors to solve their pressing interoperability issues. "The burden to making a mixed environment work falls to the people selling it," Fowler says. Of course, many users will not be comfortable waiting for new solutions, and so they'll make do as they can.
Gael Core is a free-lance writer based in San Francisco. She can be reached at firstname.lastname@example.org.