Client Server/Wars

As they struggle to cope with the longest economic recession in 45 years, Japanese companies are targeting their Information Technology (IT) departments for major surgery. More painful restructurings, such as mass redundancies, are being postponed as main frame computers go under the knife.

Some companies plan to take extreme measures. "Certain customers I have spoken to have a corporate mandate that, in less than three years, they will not have any mainframes running," says Unify Japan president Phil Rosenbaum." One customer has 30 months to get rid of 38 mainframes. The company will only retain 5 machines."

The future at stake

Industry statistics emphasize just how much is at stake. The domestic mainframe industry was worth 1.6 trillion yen in 1992, with sales of over 4,900 mainframes. Both by value and volume, though, 1992 was a disastrous year for makers; shipments were down by more than 20% over 1991. Reliable final figures for 1993 will not be available until mid-June, but a continued decline seems evident; first-half 1993 results showed a further drop in shipment value of 10% over 1992.

Japan is a nation of big companies and big computers. Names like Fujitsu, Hitachi, IBM, and NEC can be found on computers in virtually every office of 100 employees or more. These mainframe vendors are worried about the surge of interest in downsizing. Not only do the prospects for future sales look dismal, but their lucrative secondary market for customized programming, maintenance, and other support services is coming under threat.

Such vendors realize that they will be unable to hold back the oncoming storm. Instead, they have decided to ride the downsizing wave, setting up special project teams to learn and implement interconnectivity and open systems. NEC is a case in point. In July 1992, it established a special team of 50 engineers within its Tokyo head office to promote use of mixed NEC and non-NEC solutions, something that would have been unthinkable in the 1980s.

A surge of interest in downsizing

When discussing downsizing, it is important to remember that Japanese corporations do not have the same kind of love affair with computer technology as those in the West. Japanese companies have traditionally spent most of their equipment budget on mainframes or office computer systems for essential database and accounting functions. Their remaining funds have been trickled out for the purchase of dedicated word processors to prettify corporate correspondence. Such Western enhancements as e-mail, business graphics, and personal productivity packages are virtually unknown in the typical Japanese corporation, seen as unnecessary frills to the essential activities of business.

Viewed from this context, the true magnitude of the changes now happening in Japan is dramatic. Since the onset of the recession, attitudes toward small systems have changed. Suddenly, IT managers are discovering such "radical" concepts as small system scalability and distributed processing. And in no area has the interest been greater than in Relational Database Management Systems (RDBMS). That is where the real action is right now.

Japan has more than a dozen major monthly computer magazines, and almost every one of them now carries regular articles explaining downsizing and profiling case studies. This is in stark contrast to only a year ago, when the issue was largely ignored. IT staff at major companies clearly understand the risk to their jobs if they don't move soon and offer plans to drastically reduce costs.

"The recession is postponing big project deployments:" says Allen Miner, Oracle Japan's director of technology, "but we are seeing a huge volume of experimental systems. Many companies are doing work groups, and everyone has a project to learn what downsizing can do for them and how far they can go."

One sign of the dramatically increased interest in RDBMS implemented on small systems -- especially UNIX servers with PC clients -- can be seen in the surge of people attending new product releases and technical seminars. At a half-day Sybase Japan seminar held at the Tokyo Hilton Hotel at the end of 1993, for example, over 1,000 corporate IT people showed up, making it standing room only.

Three main players

The major players in downsizing (and, more specifically, the client/server arena) can be divided into three main groups, each offering different expertise and benefits. First, the traditional mainframe vendors, who are suppliers to the biggest Japanese corporations, are branching into client/server systems and developing staff with the necessary engineering expertise. Second, the software houses -- dominated by leading-edge American firms-are channeling products through both large Japanese manufacturers and mid-tier systems integrators.

Third, the independent systems integrators and value-added retailers are making their own significant contributions to development of the client/server market. Whatever their expertise and background, however, all three groups share a common desire: to migrate corporate business systems off mainframe hardware and custom-made data bases, and onto an Open Systems UNIX-PC platform and client/server RDBMS.

Traditional mainframe vendors branch out

The first group of maior players consists of the traditional mainframe vendors, who want to be all things to all people. Typically, these are the only companies that can reach the largest Japanese corporate users, due to carefully maintained, decades long business relationships and the credibility that thousands of enlployees can bring. Most of these companies--which include DEC, Fuiitsu, Hitachi, IBM, NCR, NEC, Toshiba, and Unisys--have been scrambling to roll out both UNIX- and DOS-based client/server systems and the engineering teams with which to implement them.

Fujitsu is a good example of a company keeping all its bases covered. In its UNIX lineup, Fujitsu sells the full range of Sun products, the DS/90 range of 1CL-sourced SPARC workstations, and its own recently announced S4 family of SPARC workstations. For DOS/V clients, Fujitsu sells both its own brand of PCs and Deli machines (through its subsidiary, PFU). On the software front, Fujitsu supports ORACLE and Informix, plus its own native RDBII and C-RDB databases.

NEC is another aggressive competitor. Not only is it maintaining leadership (still over 50%) in the PC market--even in the face of concerted challenges by Fujitsu, IBM, and other DOS/V manufacturers--NEC is also gaining share in the UNIX market. According to NEC, although in both 1992 and 1993 it remained back in third place as a UNIX workstation supplier, after Nihon Sun and Hewlett Packard (HP), its market share nevertheless moved up to 15%. This relentless growth nudges HP at 18%, and must have HP and Nihon Sun wondering how to deal with the situation. Nihon Sun's distributors--Fujitsu, Toshiba, and Itochu (formerly C. Itoh) in particular--are vulnerable to attack from a newly aggressive NEC, since they carry large amounts of inventory and are relying on their continuing dominance in the fast-growing UNIX market to prop up ailing balance sheets.

Oracle is paying special attention to NEC's push into UNIX. Just two weeks after it released a recent update of its ORACLE7 RDBMS in the US, ORACLE launched a Japanized version for the NEC 4800 series. This NEC port preceded by several weeks the English-language versions for most mainstream machines in the US, and it was a couple of months ahead of ports to other Japanese systems.

American firms dominate the RDBMS field

Software houses make up the second group of players. Typically, these are American firms creating leading-edge technology. They have proven particularly adept at channeling large volumes of both branded and OEM products through large Japanese manufacturers and mid-tier systems integrators. The use of existing Japanese vendors has proved a smart strategy. It sidesteps potentially damaging competition by allowing the manufacturers and integrators to continue selling hardware which they understand best, while still being able to reach their considerable client bases.

The current leader among the RDBMS software houses is Oracle, which accounted for over 30% of the Japanese market in 1992. Second place goes to Informix, followed by Unify, Sybase, Empress, and Ingress. Dataquest Japan estimates that last year these companies had created an annual market worth about Y38 billion, up from 24.7 billion yen in 1992. Further, six of the top seven RDBMS companies reported strong sales growth last year.

As is happening elsewhere in the world, the RDBMS software market is separating out into three growth areas: market leader Oracle, feisty challenger Sybase, and the rest. Oracle's successes in Japan swamp the achievements of the other players, probably because Oracle understands the current market the best--especially the point that customers are locked into their hardware and services vendors. These vendors are, in turn, typically committed to a single hardware manufacturer. To exploit this and further fuel the downsizing momentum, Oracle has ported ORACLE7 and its predecessor ORACLE V6 to over 30 hardware platforms in Japan.

Current number two in the Japanese marketplace is Informix, but the company to keep an eye on is Sybase. Since its establishment as a wholly owned subsidiary in 1992, Sybase Japan has grown from a staff of 5 to 90. Its sales doubled during 1993, and in 1994 they are expected to double again. Sybase positions itself on the basis of its architecture. To undecided systems integrators, it cites such differences as lower memory overhead, OLTP-specific design, and better host system connectivity. Sybase is targeting companies with large volumes of data. A full implementation for Japanese-character handling became available with the release of System 10, and the software is available in OS/2, NetWare, and Windows NT versions. Further, Sybase's commitment to the client side of the equation is being fulfilled by middleware such as Japanesecapable ODBC drivers (actually made by Q&E of the US) and the product's easy integration with the popular Japanese versions of Visual Basic and Access. Sybase's client base in Japan is similar to that elsewhere in Asia in that it is dominated by world-class Japanese banks and securities companies. Also, unlike Oracle's "all things to all people" multi-platform support policy, Sybase intends to port to only a few selected hardware platforms, such as NEC's 4800.

Hardware giants move in on systems integrators

The third group of companies significant in the development of the client/server market are the systems integrators and value added retailers (VARs). Japan's marketplace has traditionally been characterized by long-term relationships between these companies and the end-users. As Uni fy Japan's Rosenbaum says "Whereas Western companies like to take ownership for change within their computer operations, Japanese companies like to delegate this ownership to close vendors, preferring instead to have their IT staff do the overall coordination and management."

This delegation of responsibility comprises an important difference of the Japanese market. Although the essential need fo the change is communicated by the customer, most of the decision-making downstream of that initial request is left to the vendor. Companies like Oracle understand this well and have committed significant resources to teaching vendors how to implement client/server technology. Oracle has eight traininig centers around Japan, offering courses in RDBMS implementation and programming to over 400 people a month.

Although systems integrators have close relationships with their clients, they are now under threat from the maior hardware manufacturers, such as Fujitsu and NEC. The giants are rapidly moving into the integration business in order to find positions for their younger skilled workers exiting the main frame business.

Prominent systems integrators in the client/server market (and providers of downsizing services in general) include Sumitomo Metal Industries Ltd., through its SIS Co., Ltd. joint venture with IBM Japan and its SSD Co., Ltd. joint venture with NEC; Nomura Research Institute; NTT Data; Hitachi Information Services; CSK Ltd.; and Itochu's CTC. Also in this arena are the software localization houses such as KK Ashisuto and Software AG. American players like Anderson Consulting and EDS have met with little success to date, although EDS has become more aggressive since its takeover last September of the mid-size software publisher Japan Systems Company Ltd.

Client/server, or just downsizing?

The question remains as to whether Japanese companies are using client/server computing to re-engineer their business or are simply downsizing for the short term in order to cut costs. The important question holds the key to just which RDBMS companies will be successful over the next 18 months.

Steve Furney-Howe, strategy promtions manager for Nihon Sun Microsystems, says, "People are looking at doing one of two things. Either they want technology that saves money by automatic functions, or they want to redesign processes to help them make more money. The second option is where most people want to be." Certainly, business re-engineering has become all the rage in Japan. Looking through any bookstore would make the casual passerby think that there was a business concept revolution afoot. And maybe there is...

Anecdotal evidence, however, suggests that most Japanese corporate IT departments are still very much stuck in the old, centralized mainframe ways. Transaction monitoring products, such as Encino from Transarc Corp. of the US and Tuxedo from Novell's UNIX Systems Group, are finding ready markets Japan, even though the small client/server solutions being built by most early adopters should not yet have a need for such high-end management tools.

Another reason to believe the markets long love affair with centralized computing remains strong is the popularity of Oracle's multiprocessor data servers. Tl implies a demand for heavy-duty, sing: point data handling. Some of Oracles competitors, at least, like to disparage these products as a host-slave solution rather than a true client/server one.

One can argue that it is precisely because Oracle understands the markets lingering infatuation with a centralized model that the company is doing so well in Japan. Besides, the Oracle products work well, quite apart from the fact that they have been ported to such a wide range of machines.

Incentives to shift to a more distributed client/server scheme are not yet apparent. It is after companies downsize and start looking for ways to further reduce traffic loads and cut communications costs that other RDBMS solutions will become an attractive option.

A further issue contributing to the success of companies having a full and integrated range of products is the inability of third-party integrators to freely mix and match hardware and software as they would in a true Open Systems environment. As Brian Rogers of Comshare Japan says, "The lower levels of systems services are still missing in Japan. For example, drivers to connect the Japanese version of OS/2 into a TCP/IP network are not available. In the USA we could buy them for around $300 to $500 per client." With the scores of mainframe, workstation, and PC platforms in Japan--and an equally prolific number of UNIX versions to support each-there are substantial interoperability problems yet to be overcome.

The Japanese approach Japanese companies are remarkably similar in their approach to implementing a client/server solution. First, the IT staff attend seminars and collect information on software solutions. They don't usually focus on the merits of different hardware vendors, since changing the hardware brand name that the company uses is usually a major political decision they just don't want to deal with. This means that it is common for a company that is downsizing to buy PCs from the same supplier as their previous mainframe or mid-size system--making it hard for outsiders to break in to the market.

Next, a typical company will request its vendor to provide from a choice of RDBMS software a specific package that runs on a suitable UNIX, Windows NT. or DOS platform. If the company is a financial institution, SYBASE is the most popular choice. Other industries prefer ORACLE or Informix.

After building a small network, the IT staff will experiment by batching and downloading mainframe data by using a product such as Intercom's Falcon. They then massage the data for MIS use. Typical early applications might include e-mail, scheduling, performance reports, and dissemination of company PR information. As the IT staff become more confident and start making demonstrations and further implementation proposals to management, they are given the green light to add more equipment and grow the network.

Eventually, typically over 12-18 months, all the non-core business computing is done on the client/server systems, and from there the experimentation phase ends. The next step is where the real work begins--migrating the core systems to a small systems environment. It is this scalability that appeals to the incrementalist, low-risk thinking of so many Japanese companies. And it is the reason why a growing number of client/server specialists are betting the technology will become successful.

Only the first skirmishes The Japanese market is going through the same first flush of downsizing that the US and Europe experienced 3 years ago. Japanese RDBMS users face a similar learning curve, one that they must negotiate before they can effectively implement client/server solutions.

Given that the Japanese system integrators are locked into long-term relationships with the large hardware manufacturers, foreign entrants are generally seeking to leverage their chances by offering their experience as consultants in maintaining system and data integrity. This creates a window of opportunity that switched-on newcomers will follow.

The action so far has been merely the initial skirmishes. Although Oracle has snared most of the market for now, major opportunities for the other RDBMS makers will arise when the second wave of system enhancements takes place, in about 12 to 18 months.

And that is when the real Japanese client/server wars will begin.