DBMS Interview - November 1995
Of the "big four" database companies -- Oracle, Sybase, Informix, and CA -- Informix has the lowest profile. While the other companies are holding splashy press events and advertising in several different forums, Informix has been content to stay in the shadows, away from the aggressive cat fights and "me too" announcements. But Informix isn't going to be in the shadows for much longer. Last year, the company was third in Unix database sales, and this year it expects to take over the number two slot. And according to company officials, its Unix database license sales are growing at an annual rate double that of Sybase (nearly 50 percent compared to Sybase's 18 percent).
Despite its history of "stealth marketing," Informix is considered to be a top contender in the high-end parallel processing market. Built on an engine completely rearchitected four years ago for parallel processing, Informix's Dynamic Scalable Architecture provides three database servers that scale from the workgroup through SMP servers and distributed database environments to MPP servers and SMP clustered environments. (For in-depth discussions of Informix's parallel database technology, see DBMS's Parallel Database Special, March 1995.)
The key to competing with Oracle and Sybase is focus, says Informix VP of Marketing Steve Sommer. Informix intends to remain steadfastly focused on its core business, and avoid being distracted from the ancillary businesses of video servers (read: Oracle) or careless acquisitions (read: Sybase). Therefore, Informix has made good progress in the high-end applications and embedded database market, as well as in the application development tools market with Informix-NewEra. Strong partnerships with systems integrators, hardware, CASE, and other vendors fill in the gaps, providing Informix and its customers with a complete, integrated solution.
Features Editor Theresa Rigney recently sat down with Steve Sommer to discuss the company's plan to jump out of the shadows and take a more aggressive marketing approach. An edited transcript of their conversation follows.
DBMS: Why does Informix have such a low profile, despite being such a large company?
SOMMER: Stealth marketing. Historically, Informix has been very successful at selling indirectly to customers, so the main goal was to ensure that all the channels knew about us -- VARs, the hardware vendors who sold our database, distributors, and so on. Those constituencies knew Informix very well, and they helped us grow very quickly in the '80s and early '90s.
The profile we've got in the end-user market, however, is relatively low. Typically, we haven't marketed heavily to these users, but most companies in the Fortune 1000 do have Informix somewhere, it's just buried within an application (a client/server application, a warehouse application, or logistics, or a time and attendance package, a legal billing package, and so on). So although Informix is there somewhere, we haven't needed to market to the end users; the database decision just wasn't strategic at that level.
We have many distributors, such as Siemens-Nixdorf in Germany. Every SNI computer sold in Germany has an Informix database in it, but it doesn't say "Informix database." In effect, we have the number one market share in Germany -- ahead of Oracle -- but not many people know it.
Recently, we shifted our technology to the high end of the market -- high performance, scalability, and parallel processing. These are the multimillion dollar deals and the high strategic decisions. We've ramped up our service and support to 24 by 7, around-the-world geographic coverage, and we've added financing programs to allow things such as leasing terms. And we've had to adjust the marketing message as well, to high-end strategic marketing.
There are also some secondary reasons for this low profile. For several years --from 1990 to 1993 -- there wasn't a centralized marketing department at Informix. The company was marketing in each of the geographies, so pieces were all over the place. Plus, we hadn't talked to the industry analysts in three years; no one was dedicated to keeping them up to date on what was going on at Informix. But our competitors were talking to the press and analysts every week. So I guess it's not surprising that we did have a low profile for quite a while.
The visibility, however, has gone up dramatically in the last 18 months. The analysts know about us and recommend us as the best database. We also have seminars that are attended by thousands of people a quarter. We're also doing more hard-hitting comparative advertising, and we've had very focused messages about Informix scalability, parallel processing, and the paradigm shift from uniprocessor to multiprocessor computing.
We've seen more of an Informix presence in the VLDB and MPP space, but do you have any plans to tackle the desktop markets dominated by products like Watcom?
Historically, most of our sales were low-end database -- the two- to 32-user market. We offer a product called SE (Standard Engine) that competes with a Watcom-type product. Performance is not key, but it's for users who want ease of use. That database actually got us into the highly deployed sites; it is really big in the retail in-store processing markets, as well as bank branch automation and property management markets -- all of the major hotels (Sheraton, Hyatt, Marriott, and so on) have Informix installed at every property. With this database, you don't need a DBA, a Ph.D, or an MIS-type person on site, because it is very easy to configure and load, modify, and manage remotely or locally.
In the last two years, however, we have focused more on the higher end. This gives us the halo effect that Informix has a scalable, leading-edge technology. We're doing more of an enterprise sale, which focuses on mainframe-caliber applications, as well as departmental and workgroup applications. To ensure that we do well in the workgroup space, we're coming out with the OnLine Workgroup Server product, which is due at the end of this year. This product will be upward-compatible (based on our OnLine code) with our other products for the midrange and high end. It will be priced very competitively with Oracle Workgroup Server and Microsoft SQL Server, and will provide high performance with ease-of-use functions for NT and Unix.
How does this vision differ from Oracle's and Sybase's?
The first factor is that we have a very focused strategy. We are a database and tools company, and we are happy being a database and tools company. That way, we will stay ahead of Oracle and Sybase in terms of technology is that we will continue to focus on our database and tools. We are not in the applications business -- we don't know about MRP, ERP, CRP, or any of those manufacturing terms. We are also not interested in competing with our VARs or partners. We draw the line at having the best database and tools, and all the standards-based connectivity you need, but we have partners who build all the applications.
We don't compete with our system integrators either. If a company wants basic consulting on database optimization, modeling, and deployment, we'll send a high-level consultant over there for a short time; but if the company wants 18 people to come in and reengineer the entire business and develop and code new applications, we'll leave that to our system integrators.
We're also not doing a lot of set-top Beavis-and-Butthead-on-demand type stuff. Consumer TV may be a great market, but it's not our business. We're very happy to sell the database to manage the assets --the video content, the billing, or subscriber management -- but we're not going to get into set-top TV boxes or any of that stuff.
That's why Oracle is about four times bigger than us in terms of revenue, but if you compare Unix databases to Unix databases, they're only twice as big. Sybase is trying to do the same as Oracle. Every time Oracle makes an announcement, Sybase is like a little kid yelling, "Me too!"
In 1994 we were number three in the Unix database market (we had 19 percent share, Oracle had 20). In 1995 we will take over the number two slot. Sybase will still be bigger, because it's done about six or seven acquisitions in the last few years, but in terms of Unix database core business, we'll overtake them this year. Sybase database license sales are growing about 18 percent; ours are growing 50 percent. That's one of the biggest differentiators.
When you get to the next level down from corporate strategy to database strategy, guys like [Robert] Epstein at Sybase fundamentally believed that the future of computing was going to be networks of uniprocessor systems. This was a good model of computing in the '80s, but every hardware vendor is now saying that it's going to be networks of systems, some of which will be uni, some will be two-way, four-way, eight-way, SMP, some will be clusters, and some will be MPP. We realized back in 1990-91 that this was the way the industry was going (we weren't visionaries but we listened to our hardware partners), so we set up a development lab and totally rearchitected our core database (with 800,000 lines of new code) specifically to address parallel processing. That commitment to parallel processing is also something that sets us apart from Sybase and Oracle.
Other vendors have followed Informix's lead with parallel query processing. How will Informix stay ahead of the game?
A couple of ways. One is in terms of basic performance. We have an architecture that should give us a three- to five-year edge over the competition. It's not any one feature or function; rather, it's the fact that the core architecture is designed for parallel processing. And this architecture will be extended to all sorts of nontraditional data types, not just alphanumeric data, but image, audio, video, multimedia, documents, and workflow. So taking the highly parallel architecture we've got and extending it to handle more nontraditional data types in parallel will be one way to stay ahead.
Another way is basic functionality. Companies like Microsoft and Sybase are saying, "Well, by 1997 we'll support parallel query and row-level locking," but those are things that we and Oracle do now. And we're already coming out with 64-bit operating system versions; I don't know when those other vendors are going to come out with that.
We're seeing a trend away from commoditization. A lot of different databases can handle the low end, but when you talk about high-end computing, performance and scalability really matter because you can save millions of dollars in hardware costs. Faster and faster technology is coming out from the hardware vendors, and we're trying to ensure that we stay ahead of the competition by optimizing for it. We have huge joint development labs (one at Hewlett-Packard, one at Intel -- a lot of engineers from AT&T GIS, Sun, Digital, and others are here in our development labs), so we can be involved earlier in how these vendors are designing their SMP and clustered technology. They benefit because we understand parallel database a lot better than they do. We helped one major vendor rearchitect its approach to parallel processing. The advantage to us is that we know what they're coming out with three years from now.
How will Informix incorporate object technology into its database strategy?
The approach we're taking is toward object-relational DBMS [ORDBMS]. For a while (late '80s, early '90s), there were two camps: either relational databases were going to extend with object extensions, or relational databases were going to dead-end like hierarchical systems did, and object databases would take over. Therefore, all these VCs [venture capitalists] were pumping money into Ontos, Servio, Versant, and others. However, the consensus seems to be now that object databases are very limited; they have poor performance and scalability, they're not very interoperable, and so on.
I think our major competitors are going with the same approach: taking the core relational product -- which offers great performance, scalability, and interoperability -- and adding object extensions. Our plan over the next 18 months is to add an object API, through which you can add specific object extensions to the database. It will be an extremely efficient way to handle all sorts of nontraditional data: geographical data, 2D and 3D spatial data, fingerprint data, and so on. We'll have our own extensions and work with third parties.
Are you following the SQL3 standard?
We do plan to support SQL3, but the standard won't go far enough in what the industry is going to require in terms of object extensions. Actually, we try to follow the standards completely, and not break a lot of new ground in proprietary things. That's one of Sybase's big problems: It did stored procedures and triggers before there were any standards. Eventually, SQL-89 and SQL-92 came out with standards that we and Oracle followed, and now any application built with Sybase is quite difficult to port to other databases.
How much of the Informix business is OLTP versus data warehousing?
It depends on how you define the data warehouse. We don't track our business by how much is decision support or batch/OLTP, but I would guess that about 65 percent of our business is OLTP and 35 percent is decision support. However, decision support is not necessarily synonymous with data warehousing. A data warehouse denotes a very large database (50GB, 500GB, or more) and it usually denotes an aggregate database; that is, you're taking data from several production systems and structuring a huge database for queries. A majority of our customers, however, do decision support based on one database that is used for OLTP and decision support, or who have replicated the production database to another system for decision support.
These scenarios don't fit the traditional definition of data warehousing. The short answer is that the majority of our business is still OLTP, but a fast-growing part of our business is decision support and data warehousing.
But because Oracle recently acquired the Express product family from IRI, and Sybase acquired Expressway last year to aid its warehousing strategy, does Informix see a need for an in-house OLAP/data warehousing piece in its product line?
We looked at Expressway, which offers a way to build an index quickly for querying, but decided not to buy it (actually, several of the database vendors looked at Expressway). There are different reasons for why we didn't buy it. It does help to have fast index-build capabilities, and while we are going to do that, fundamentally we saw more value in building an index quickly through parallel processing. Again, we understand where the industry is going. Instead of trying to figure out some way to make software run on a 386, we say, "The hell with that: Let's look at the Pentiums coming out and make sure we can take full advantage of the current chip, operating systems, and all that." Sybase relies heavily on specialized index builds because it can't build and scan indexes in parallel. We can do that.
Regarding the Oracle acquisition of IRI, as far as I can tell it did it because of its applications business. It's a completely proprietary form of OLAP -- you don't set up a data warehouse in Oracle and then have Express read the data. Rather, IRI has a proprietary format for data, and it extracts the data and reformats it into this proprietary data format. So, with Express, you're actually querying a data mart. IRI and one or two other companies do it that way, but every analyst I know, including Gartner and Meta Group, says that this approach will go away -- you don't want to have all these proprietary OLAP/data marts around. But the IRI technology worked with the Oracle applications, so I guess the acquisition made sense.
Our approach is to focus more on nonproprietary OLAP, and to ensure that a variety of third-party tools work with our database. Then you can map the data any way you want for drill-down analysis, multidimensional queries, and so forth. So we're working with Stanford Technology Group, Information Advantage, MicroStrategy, Business Objects, and others to make sure we work with their software.
NewEra 2.0 made its big splash at Informix's user conference in July. Would you characterize this tool as meant for the "mass markets" of developers, like Visual Basic or PowerBuilder, or for existing Informix customers only?
The answer is C -- none of the above. It is definitely meant for the installed base, and we have all sorts of functionality in 2.0, including support for terminals and character-based applications, so it's going to be very easy for our Informix-4GL customers to migrate to NewEra. These customers want to be able to take their old applications and go to an object-oriented world if they want to, and also add graphical front ends to them (to make their applications look like 1996 applications, instead of 1986 applications). And we've made it very easy for them to do that.
NewEra is also designed for customers who are new to Informix, and saying, "I want a great tool and great database that work together, but I have no experience with Informix." It's also open, so it works with Sybase and Oracle databases.
In terms of overall positioning, NewEra is not Visual Basic, so when you say "mass market," you're looking at 1.5 million copies of Visual Basic out there at $99 -- that's definitely not the market we're in. We're in more of a midrange, or departmental and enterprise application development market. We do overlap with PowerBuilder, but its $249 desktop version product is below where we are in terms of price, performance, functionality, and scalability. When you start getting into the PowerBuilder Enterprise, which is $3500 or so, as well as the high-end tools such as Forte and Dynasty, we're competing head on with them.
If you want to develop fairly complex applications that need scalability and partitioning -- where you can split part of an application onto clients and servers (PowerBuilder runs only on the client) -- as well as large development teams, we're set up for that. If you're looking at larger applications with hundreds of screens and more concurrent users, we're set up for that. Where the low-end versions of PowerBuilder and Visual Basic do well are with the very cheap, simple, easy-to-use decision-support applications with a single user or a couple of workgroup users. They can do those types of applications very, very well.
Has NewEra attracted third-party support?
We've got a lot of partners already. There are two types of partners. One is the complementary third-party software vendors that do things like upper-CASE, analysis, and design, all the way to testing -- companies like SQA and Mercury -- they all work with NewEra now. These types of partnerships give you a life cycle solution. Again, Informix is focused on the development tool and development environment, but it's not focused on upper-CASE or testing, or anything that third parties do well.
This approach is in contrast to Sybase, which bought SDP [and its S-Designor product] -- a purchase that is now upsetting every Powersoft/CASE partner relationship. Rather, we like to define our business and get close to partners, and they know that they can work with us for the long term.
The other types of partners we have are class libraries. Those are object-oriented modules that extend our product. So, if you want to add imaging or multimedia, for example, you can buy NewEra and buy a module called Imaging from TMS Inc. (Stillwater, Oklahoma), add a couple lines of code, and link it together. There is also a Lotus Notes class library, as well as dozens of others.
You mentioned Forte and Dynasty earlier, and application partitioning is a major new feature in Informix-NewEra 2.0. What is Informix's competitive positioning in this area?
I don't want to sound arrogant, but long-term viability is an issue for companies like Dynasty. Forte has already burned through millions of dollars in money. In Forte's case, the company came out with complex, very high-priced technology. You can develop a very complex application, but managing and modifying it are two very important parts of the client/server world, and I heard that these tasks are very hard to do with these products. We've tried to come up with something that is not a $75,000 entry point (like Forte), and where it is very straightforward to develop the application, including for small and mid-size applications, because you'd like to have one tool that you can grow with. And once you develop and partition it, if you decide to upgrade it to another platform or computer, you can easily repartition the application.
So I think that we've got more flexible technology that is easier to learn and use and less expensive. Other issues are worldwide training and support. A lot of the companies we're selling to now are the Global 2000, and if you're not in Germany, Korea, and Australia, it's really hard to close a deal. This means an office, support, service, and executives -- whatever it takes. Forte and Dynasty don't have these things, so I don't see them as long term, global players.
And what's in store for the next era of NewEra?
Version 2.0 is a huge release for us because it provides a large amount of ease-of-use templates, and ease of use is important to all tool users. The partitioning was really a big deal -- it supports the second-generation scalability issue. It's got global language support, so it's very easy to localize into Korean, Chinese, and so on (the majority of our business is outside the U.S.). And it has character terminal support -- no other tool in the world can support concurrent processing on dumb terminals and PCs. That way, customers who are migrating to PCs don't have to throw out 3000 terminals and buy 3000 PCs; instead, they can add them slowly over time.
The next release will be 3.0, which is due out the first quarter next year. The key things there are: One, it's going to be a 32-bit version of the application development environment, specifically for Windows 95 (and supporting Windows NT); two, it will have a 32-bit object-oriented repository built into it, so you can more easily store and browse through all the objects you develop, and other team developers can browse through it. So it will facilitate code reuse. Those are the two biggest features in 3.0.
On the server side, how is Informix's Dynamic Scalable Architecture (DSA) product line positioned?
In the simple world, where every computer had one processor in it, it was more a one-database-fits-all strategy. Our solution in this world was called Informix-OnLine. When we realized that the world was going to multiprocessor technology and very different architectures -- some shared memory, some shared disk, some shared buses, some shared nothing --we realized that we also had to come up with a whole new architecture. We called that architecture Dynamic Scalable Architecture: Dynamic because you can change the database and how it works based on the platforms, processors, hardware architectures, and so on. Plus, you can manage it dynamically depending on the number of users, how you allocate resources for OLTP and decision support, and so forth. Businesses need more and more flexibility.
So we came out with DSA, and the first version was named OnLine Dynamic Server. We also see a huge market at the workgroup level (two to 32 users) where some of that functionality is not as important --users need good performance but, more important, they need ease of use. So we decided to come out with a special product with less functionality but more ease of use at a much lower price; we call that our OnLine Workgroup Server product.
The third product extends OnLine Dynamic Server to the high-end, massively parallel and clustered systems. Significant new code is necessary to optimize these two types of architectures, and again, a lot of system management is built into it. Our OnLine XPS (Extended Parallel Server) addresses this market.
Essentially, we have a segmented market: We have three databases that are all based on DSA -- there's very high overlap with the code -- but they're positioned differently for different market needs. They're priced differently, they have different features, but they're all based on the same basic code, and all migratable and fully scalable.
What are the main things that DBMS readers should know about Informix?
The users of our technology don't have to worry about us losing focus. When I say focus, I mean on our database and tools. We're going to stay ahead of the competition because we are so focused. I expect that this year we will be taking over the number two spot worldwide in Unix database revenue. Anyone serious about making a database decision for Unix or NT should be looking at Informix.