From Software Futures, a sister publication Sybase used to be known and feared for its technical expertise. The reputation got dented by the non-scalability of release ten of its database. Now that scalable System 11 is out on the marketplace, the company is finally free to turn its attention towards the next nirvana – object […]
From Software Futures, a sister publication
Sybase used to be known and feared for its technical expertise. The reputation got dented by the non-scalability of release ten of its database. Now that scalable System 11 is out on the marketplace, the company is finally free to turn its attention towards the next nirvana – object technology.
By Clare Haney
Is imitation the sincerest form of flattery or what? That was a key question posed at Sybase’s rather downbeat International User Group’s 1996 North American Conference held in San Diego in May when the database vendor finally took the wraps off its much-delayed object strategy. And, were there lots of oohs and aahs of excitement? Well no. In fact, the unveiling begged the question at the start of this piece. Wasn’t the Emeryville, California-based company just playing catchup with its relational rivals? Even Dennis McEvoy, president of Sybase’s enterprise business group, standing in as proud parent for the company’s new object baby, felt constrained to comment, with a self-deprecating smile, Boy, it looks just like Informix’s Data Blades. And so it did. At least on a cursory look. But as we know all too well in the software business, appearances can be deceptive. We’ll take a look at the differences between the Sybase approach and what McEvoy dubbed Illformix, the combination of Informix and its $450m object/relational acquisition Illustra purchased last December. It’s worth considering at this point that although Sybase is certainly last to the object/relational ball, it hasn’t really missed that much. In the words of David Hsieh, the company’s vice president, enterprise product marketing, recently recruited from LBMS, all that Oracle and Informix have been offering so far is duelling slideware. And now Sybase has joined in these overhead projector battles of vaporware, with McEvoy stressing, as if we hadn’t already guessed, that all the company has to offer at the moment is its object vision and product roadmap.
However, whether the Sybase approach differs from the other relational players or note, it doesn’t only talk up the database. Sybase terms this not being database myopic. In San Diego, all company personnel underlined the importance of having an end-to-end solution to the enigma that is object technology. So not only are we talking a new object/relational database, a variant of SQL Server 11, known as Adaptive Server (see diagram, p17), but also middleware with Sybase’s ObjectConnect software and application development tools with the company’s Powersoft family of tools (PowerBuilder 5.0, Optima++ and Watcom C++). McEvoy sums up the current situation in today’s market as Sybase sees it. There’s universal confusion about objects – what they are and more importantly, what they aren’t. They aren’t complex data management, Web data and they’re certainly not universal servers. Wouldn’t it be great if you could just share objects across the enterprise? We’ve done a pretty good job in the industry of providing objects for clients with VBXs and OCXs, but we have neglected the application layer and the database. And Sybase will be quick to tell you that it knows that object/relational hybrids are the way to go, precisely because it has fully investigated the pure object-oriented database alternative. In last month’s lead about DB/Expo, we mentioned the company’s abortive Brahms project, abandoned late last year. In fact, chief executive Mark Hoffman admitted to us that his company has spent (or wasted, depending on your point of view) between $20m and $30m on associated object-oriented projects. These include Brahms, as well as Build Momentum, the company’s failed attempt to create a graphical OO application development tool, since cannibalized and reheated to form ObjectCycle, an object management repository for the latest release of PowerBuilder, and Enterprise Momentum, an OO repository and application modeling framework. Ever the pragmatist, Hoffman says prosaically, We’re going to have some failures and we’ll have some more as we go forward… I’m not losing sleep about our products and product cycles any more. That wasn’t the case a year ago. Quite. A year ago, Sybase was deep in the SQL Server can’t scale and PowerBuilder can’t really scale either slough of despond.
Could look good
Actually not all of the aforementioned object research ended up filed purely under could look good on the resum experience. You may well have heard of another Sybase object project called Lego. The company has always been rather fond of the building brick approach encapsulated in this codename. In his keynote speech in San Diego, Hoffman sang the praises of that way of looking at the world, telling users, Whether the world will go OLE, CORBA or SQL3 in the future, we don’t know and we shouldn’t have to care. No one should have to make decisions like that now. Why should you be forced to stick your neck out on issues like this? You’ve got enough to worry about. Lego will be productized as one of the flavors of the company’s OO middleware ObjectConnect. Lego is the OLE variety, due to debut in the third quarter of this year, while the C++ release, already out in the market, is derived directly from San Mateo, California-based object/relational database vendor Persistence Software’s object-to-relational mapping and code generation technology. Next on the horizon for ObjectConnect is more CORBA ORB functionality and more support for ODMG-compliant OO databases. Sybase says that in the very near future we can also expect to see some remote three-tier user-defined object methods with Distributed ObjectConnect, the GUI front-end to ObjectConnect. The database vendor is also considering Smalltalk and Java language binding and supporting more object/relational mapping styles. It may well decide to use an Interface Definition Language instead of ObjectConnect’s existing GUI repository language. Back to Adaptive Server. Well, we guess choosing this moniker is better than battling out the rights to the Universal Server name, la Oracle and Informix. The Sybase database apparently represents the object/relational evolution of SQL Server 11 and will incorporate complex structured data types, such as text search, time series, geospatial and HTML, either within the body of the database or as specialty snap-ins. Sybase claims it doesn’t believe in dumping every kind of data type into Adaptive Server which can compromise performance, explaining that while the object/relational database will support video and audio clips, it will not handle unstructured data such as video and audio streams which will continue to be dealt with by Sybase’s Intermedia Server already available in the market.
A Frankenstein approach?
McEvoy was even more forthright in his condemnation of the opposition, dubbing Illformix’s idea of taking body parts from scalable and non-extensible Informix and extensible and non-scalable Illustra and trying to stitch them together a Frankenstein approach. He claims that both Informix and Oracle with its Oracle8 database will be forced to rewrite all levels of their databases at once in order to obtain their nirvana of a Universal Server. Incremental extensibility is the right approach, he says in justification. Our component-based approach to the problem is consistent with the way that object technology is supposed to work. Our competitors have been noticeably silent on how to connect objects to existing data sources. He points to the fact that Informix has already had to backtrack on its timeline for merging its own database engine with Illustra. Analysts now are thinking that it can only be achieved at the earliest by the second quarter of next year, six full months behind the database vendor’s early ambitious targets announced at the launch of the combined Informix and Illustra hydra. They said there’d be no code changes required in the DataBlades, now they’re into the discovery phase and have realized that there needs to be a 20% to 25% change, McEvoy gloats. Exactly a year ago, we gave the message that one size database doesn’t fit all. Oracle said Oracle7 was for everyone. We said you couldn’t do it with a single codeline and bought out SQL Anywhere [formerly the low-end Watcom database.] Six months later, Oracle gave up and responded with Personal Oracle Lite. They had to admit that one size doesn’t fit all. We’ve done the same thing with data warehousing and will do it with object technology too. Sybase also says it wants to ensure relational quality performance in objects and that users will be able to join specialist and relational data in the same query. It plans to incorporate the SQL3 standard into Adaptive Server over time, but doesn’t expect it to play a large role in defining reusable business objects, which it says will better be developed in existing programming languages. It hopes to debut the first group of its snap-in data products this summer.
Not merely DataBlades
OK, so Sybase isn’t purely homing in on the database. What else is there to stop us dubbing its specialty snap-ins (catchy? …. not!) merely DataBlades Mk II? Well, the company would argue that unlike Illustra, it doesn’t intend to treat the partners who will supply some of its snap-ins as contracted hired hands. Instead, they will be real long term partners. So, let’s talk to one of Sybase’s snap-in partners. Vision International is a division of Autometric Inc, based in Alexandria, Virginia. The company should close this financial year with $70m in sales, with 18-month year old Vision contributing around a seventh of that figure. You may have heard of the Vision division in connection with the NASA EOSDIS project. The company will be supplying a spatial snap-in for Sybase’s new object/relational database. I don’t believe Sybase’s Adaptive Server is like Informix’s Universal Server at all. So says Joe Cardinale, director of spatial data products at Vision International. The Informix DataBlades snap right into the database kernel itself, which means if you screw one of them up, you crash the server. With Sybase, all the specialty snap-ins are servers themselves, so if you screw one of them up, they don’t crash the database server. If you’re thinking that a single database server can handle all abstract data types – it’s nonsense. Sybase is on the right track. He contrasts the Vision approach to dealing with geospatial data with that of Illustra and Oracle. We model the earth as it’s meant to be modeled, as a sphere. Our competitors model it as a plane. They’re using a flat earth model. If you use that model around the Equator and the Poles, the answers you get aren’t 10 feet off, they’re just plain wrong. He also sees Sybase as offering a more open architecture than Oracle and Informix. We’ve built our Spatial Query Server as a Sybase Open Server application so it snaps into SQL Server System 11. To do that for Oracle, you’d have to write it in PL-SQL and for Informix you’d have to write it in Illustra’s DataBlade kit. They have no way for you to write a multi-threaded server snap-in. So has Vision started work on its specialty snap-in yet? No, we haven’t started yet. Since Sybase looks at the Adaptive Server database as just a migration of System 11, we’re just dealing with talking to another communications layer.
The users’ take
Well, that’s enough from vendors. What do users think to the Sybase object road map? We talked to Scott Barnes, senior manager of database marketing systems, at telecommunications giant MCI, based in Atlanta, Georgia. He has built and maintained a data warehouse containing some 40 million telephone numbers providing regeneration and decision support for the organization’s small business sales division. Although he admits that his company is not moving towards that particular strategy at the moment, he infinitely prefers it to the Oracle and Informix approaches. It fits in very well with Sybase’s bolt-on strategy. If you don’t need it, you don’t have to have it. Objects aren’t for everyone. Theirs is a smarter strategy than the other database vendors in the business world. Sybase had to come out and say ‘Yeah, we’re doing this too. We’re not leaving you [users] in the dark.’ I need to look into it more and understand it more. Gary Oehlerts, director of business systems, at the United States AutoClub Motoring Services Division of financial organization Associates Bancorp, based in Irving, Texas, agrees. He currently uses System 11 SQL Server to support the AutoClub’s customer base of 10 million people across the US who it provides with emergency and roadside services I need to find out more about what it really means to me in a business perspective. If Sybase’s direction gives me the opportunity to improve the performance of my business, translating into more profitability for the organization. With System 11 Sybase improved its performance and quality. Now they need to take the next step. From a preliminary look, it looks good. Sybase is making sure that its customers know what its strategy is. The overall feeling from the user conference was ‘It sounds OK, but we don’t know what it is yet.’
So Sybase has finally thrown its hat into the ring bidding for your object/relational business at a time when Java and the Web seem more likely to provide multi-platform portability for business objects. Certainly the Sybase approach seems the most sensible, sane and practicable. For the time being, all three database rivals stand poised on the starting blocks for the race to the object/relational prize. We reckon Sybase should be the snap-happy front-runner, but it needs to completely throw off any negative past baggage still weighing down its technical reputation to avoid losing out to Oracle and Informix in a photo finish.