KESELAMATAN
Despite stormy conditions, Sybase feels it has the right strategy -- from laptop to enterprise -- for any shop’s specialized requirements.
Though Sybase Inc. is one of the "big five" database companies, executives are quick to point out that it is "more than just a database company." In fact, Sybase is positioning itself as the leading vendor of open and distributed databases, middleware, and tools -- each component of which can integrate with any platform, any operating system, and any application.
This task is not as easy as it sounds, however. In fact, Sybase has had some stormy times in recent months -- the most notable being its first-quarter 1995 revenue shortfall (stock plunged more than 16 points). The company also has been fighting accusations of "me-too" product announcements against its main competitor Oracle Corp., as well as integration issues that are the result of recent acquisitions.
Still, Sybase is the sixth largest software company worldwide, and its 1995 revenue was $957 million. Sybase officials feel that they've put together a product and technology strategy that will carry the company into the next century. It's stated mission is simple: "To simplify and integrate the distributed enterprise, and help customers successfully respond to rapid business change." In this article I take a look at Sybase's existing product infrastructure and future strategies, and focus on its ability not only to offer plug-and-play components, but whether Sybase offers an integrated solution for the entire enterprise.
When Sybase was founded in 1984, the dramatic increase in the use of local area networks (LANs) was enabling a new way of computing -- client/server. Sybase recognized this shift, and in 1987 released its SQL Server RDBMS to accommodate this new paradigm. As client/server and the industry have evolved, new -- and separate -- computing models have emerged, including OLTP/mixed workload environments (environments in which OLTP and decision-support functions are performed against the same database), data warehouse and decision-support environments, and mass deployment (mobile) environments. Each of these categories has specific, very focused business requirements. For example, today's OLTP and mixed workload environments must reduce the cost of operations and integrate diverse departments. Data warehouse and decision-support systems (DSSs) let users leverage their data assets, monitor business trends, and design new products and services based on their analyses. Mobile users demand information anywhere at anytime.
Database, middleware, and tools crosscut each of the OLTP, data warehouse, and mass deployment categories. Databases manage data and transactions, middleware transforms and moves data across heterogeneous environments, and tools aid rapid application development. This is the basis of Sybase's product strategy: to provide databases, middleware, and tools for OLTP, data warehouse, and mass deployment environments. Sybase's "three-by-three product matrix" is shown in Figure 1. Underneath the matrix lies the electronic commerce (or, in Sybase's terms, the electronic marketplace) segment.
In the three-by-three matrix, every product is designed to interoperate through the same API, thereby providing scalability from the smallest to the largest environment in a single architecture. Theoretically, each component is also designed to plug into existing computing architectures, but Sybase has been criticized in the past for integration problems. Company officials insist that the new breed of Sybase products is now fully scalable and interoperable. Let's look now at the tools in each of the matrix boxes, as well as some of the peripheral products on which Sybase is working.
According to Sybase's Daniel Lahl, group director, enterprise marketing, Sybase is not buying into the "universal database" concept now being pushed by Informix and Oracle. Instead, he insists that a maturing market needs different types of databases for different requirements. Therefore, Sybase has segmented and optimized its System 11 products for these different needs.
Sybase System 11 consists of four main database engines: Sybase SQL Server, Sybase IQ (formerly IQ Accelerator), Sybase MPP (formerly Navigation Server), and Sybase SQL Anywhere (formerly Watcom SQL). All of these products are available on some platforms.
Sybase SQL Server 11 is the latest release of Sybase's flagship database server. In addition to being fully ISO 9000-compliant, it features a Logical Memory Manager that provides named caches to reserve space for different types of objects, variable block sizes and buffer algorithms for faster I/O, a cache-sensitive optimizer to automate the best search and replace strategy, and a cache analysis tool. The Logical Memory Manager is specifically designed for mixed workload environments -- you never have to swap out the OLTP application when you perform a table scan, and different I/O "chunks" are specialized for OLTP and DSS.
SQL Server 11 also provides dynamic load balancing via a symmetric networking and parallel lock manager, as well as accelerated data access via data partitions and in-place updates. Efficient logging is provided via multiple log caches and expanded group commits. (For a soup-to-nuts discussion of System 11, see the interview with Sybase's vice president of Product Development, Dennis McEvoy, DBMS, October 1995, page 38.)
Included with System 11 is Backup Server, which is a separate server that performs automated backup (at up to 60GB per hour) while your users are still online. Audit Server, which is marketed to government organizations, provides B2-level security and password auditing. B1-level security will come later this year with Secure Server. This product will provide authentication services and systems management functions.
Several options are available now for System 11 systems management. SQL Manager is included with System 11, or, for real-time monitoring, users can purchase SQL Monitor for SQL Server 11.
Data warehousing can be divided into two main operating approaches: a centralized data store or a data mart. For the centralized data store approach, which implies preplanned queries and detail reports against very large databases, Sybase provides SQL Server 11 with the Sybase MPP option. Sybase MPP is a subset of SQL Server optimized for massively parallel processing (MPP) environments that utilize a shared-nothing architecture. According to Joshua Bersin, group director for data warehouse solutions, Sybase MPP provides linear scalability to 128 processors. Plus, the 11.0 code line will be integrated into Sybase MPP later this year to provide cursor support and large block I/O. It is now on most of the major platforms, including IBM's SP2.
The data mart approach implies that users will be performing exploratory analysis against lower-cost, departmental servers. For this approach, Sybase provides SQL Server 11 with the Sybase IQ option. Sybase IQ implements the Bit-Wise indexing technology that Sybase acquired from Expressway Technologies in October 1994. The Bit-Wise technology is based on the use of bitmaps to enable all data and data types (including binary large objects, or BLOBs) to be represented as bits within indexes. Column-wise processing is then applied to only the required bits in a specific query. Response time is further shortened by patent-pending algorithms for optimized storage, compression, and access of aggregates and range queries. According to the Sybase product engineers, the typical IQ compression can reduce the size of traditional DSS databases by 75 percent.
However, Sybase officials are quick to point out that Sybase IQ is not an indexing product; rather it is a database that is optimized for data storage and ad hoc queries. It is ideal for data warehousing environments because it runs on standard commercial hardware (including Hewlett-Packard, Sun, and Digital), and it uses less disk space.
Sybase IQ also supports opportunistic parallelism. When loading a database, the product distributes the workload across all CPUs. The product can load up to 4GB an hour. In production environments, Sybase IQ can assign certain tasks, such as sorts and joins, to separate threads, as well as optimize the blocking and caching mechanism for faster response.
The mass deployment marketplace encompasses mobile, desktop, and PC-LAN environments. These environments must remain compatible with corporate data, be easy to install, use, and maintain, and be low-cost with a small footprint. For these computing environments, Sybase provides Sybase SQL Anywhere. According to Lahl, the newest version of SQL Anywhere, version 5.0, provides 95 percent compatibility with Sybase Transact-SQL, as well as enhanced interoperability via Sybase's client/server APIs (which makes it interoperable with SQL Server -- the lack of this interoperability often led to criticism of Sybase in the past). SQL Anywhere also provides replication to and from corporate databases, as well as to and from mobile databases via SQL Remote, Sybase's new connectivity software. The SQL Central GUI administration tool provides real-time database monitoring of SQL Anywhere databases.
The Enterprise Connect family of middleware products enables Sybase users to exchange information from platform to platform, database to database, and application to application. All of the Sybase products in this family are built on the Sybase OpenClient/OpenServer architecture. On the OLTP side, Sybase offers Replication Server for heterogeneous replication (a non-Sybase data source can serve as a publisher of replication data, not just a subscriber), and DB Gateways and Open Server for access to various data sources.
In the data warehouse space, OmniConnect was merged with the Micro Decisionware Inc. code base; a toolkit is provided to enable companies to build their own interfaces to other databases and data sources, including mainframe data. In addition, Sybase has relicensed the InfoPump technology from Trinzic (now owned by Platinum Technology). InfoPump is used for batch-type data movement among platforms (and into a data warehouse).
In the mass deployment category, last year Sybase began shipping its Enterprise Messaging Server, which provides asynchronous messaging among laptop and wireless devices. The company is integrating this technology into SQL Remote, which provides replication among mobile platforms. SQL Remote runs on all Microsoft platforms, as well as Macintosh, Novell, and OS/2.
Sybase is also working on an object middleware strategy -- similar to Oracle's much ballyhooed Sedona project -- called ObjectConnect. This product family is middleware that connects object-oriented applications to corporate databases. It's especially useful for OLE, C++, and distributed PowerBuilder programmers who want to store objects persistently over a network. On the client side, ObjectConnect supports C++, PowerBuilder, and Visual Basic clients (through OLE). On the back end, it supports Sybase, Oracle, and Informix. In effect, the ObjectConnect layer sits between the OO clients and the corporate databases, and provides an OO view -- making data in a database look like a C++ class or an OLE Automation Server. For three-tier computing, any object will be able to invoke methods on any local or remote application server.
The ObjectConnect subsystem will be composed of programmable object interfaces, an Interface Repository that contains the object interface models, and an Implementation Repository that stores physical mapping information and interfaces to methods. Methods provide the implementation for most of the subsystem's functionality, including data source mapping, building and managing the Interface Repository, object dispatching, and generating code for C++ language bindings.
ObjectConnect for C++, which should be available by press time, consists of the following three components:
Object Builder: a GUI toolset that automatically generates C++ classes with database access (this is where a user defines his/her object model)
ObjectConnect for OLE, due out in June 1996, provides an Object Builder, which is a GUI toolset for generating OLE interfaces for database access. It also includes a dynamic interface to an object repository. The OLE Object Server is an OLE Automation Server for improving performance and managing object integrity for OLE applications.
ObjectConnect Server with three-tier support (targeted for a late 1996 release) will support OLE and Common Object Request Broker Architecture (CORBA) application interfaces. Other future directions include ODMG Object Model and Object Query Language support, standard methods for retrieving and manipulating relational and OLE data, support for asynchronous event notification support, support for user-defined data types, Smalltalk and Java support, and support for nonrelational data sources.
On the tools side, Sybase made a strategic merger in February 1995 with the hugely successful Powersoft Corp. With Powersoft came PowerBuilder and InfoMaker (Powersoft's end-user front-end tool), as well as the Watcom compilers and database, which Powersoft acquired the year before. In mid-1995, Sybase bolstered this product offering with the S-Designor modeling tool. Once it had this suite of end-to-end client/server application development tools, Sybase essentially dumped its own application development tool strategy, which was based around the Gain tools it had acquired in 1993. According to Sybase Executive Vice President Robert Epstein, the Gain technology has been re-implemented from scratch to be optimized for the Internet as a personal authoring tool, and renamed media.splash. (See the sidebar, "Robert Epstein Speaks Out," on page 50.) Although no delivery schedule has been finalized, Epstein promises that media.splash will provide real multimedia power that is optimized for the Internet and compatible with PowerBuilder 5.0.
The long-awaited version 5.0 of PowerBuilder, which the Powersoft division plans to ship in the second quarter, is also focused on the burgeoning world of the Internet. Specifically, it will let PowerBuilder developers create plug-ins for Web browsers and distributed objects that are accessible from browser clients. Powersoft will also produce DataWindow OCX and browser plug-ins, which will enable browsers to access databases.
Sybase's overall Internet framework is called "web.works." Its goal is to Internet-enable existing client/server applications through off-the-shelf browser extensions, or by developing executable content and custom application servers. Sybase's middleware and database infrastructure will provide connectivity to corporate information, as well as to database servers on the Internet.
To this end, Sybase has announced two products: web.sql, which will connect Web servers with SQL Server System 10 and System 11, and Optima++, an Internet application development tool. Sybase has not yet mentioned plans for Internet-enabling SQL Anywhere.
web.sql, due out for Sun Solaris by press time and for other Unix platforms and Windows NT later this year, will integrate databases and the Web by facilitating bidirectional communication between Web servers and DBMS servers. Developers will be able to embed queries in HTML documents using SQL or Perl syntax. To improve performance when processing multiple queries, web.sql will maintain multiple connections between a single Web browser client and the database. Although Sybase focuses on its support for Sybase System 10 and System 11, Web applications will be able to access other data sources via Sybase Open Client connectivity software, which is included with web.sql. In addition, web.sql will support access to databases through Java applications.
Optima++, due out at the end of the second quarter of 1996, will enable developers to build Internet and Intranet applications, including Java applets, Netscape plug-ins, Microsoft OLE sweeper controls, and document viewers that are compatible with Microsoft's Internet Explorer Web browser. It will emphasize the use of wizards, drag and drop, and components to simplify 3GL development. (Optima++ incorporates Watcom C/C++ technology but does not replace the Watcom compiler.)
In addition, Optima++ will let developers create and use OLE controls. A key component supporting database access will be a PowerBuilder DataWindow control that can be used in both Optima++ and PowerBuilder 5.0, as well as other OLE container applications. The product will also include a standalone version of Sybase SQL Anywhere and InfoMaker.
In the words of Berl Hartman, Sybase's vice president, Strategic Marketing, Sybase "can't do it all." Therefore, the company has revamped its sales and marketing infrastructure to emphasize partner solutions. Its umbrella strategy, the Open Solutions Program, incorporates more than 1800 software partners and more than 25 hardware partners.
The Warehouse Works partner initiative provides an end-to-end framework for new data warehouse projects. This framework, formed last year, encompasses middleware, replication, database, and management tools from a variety of third-party vendors. Sybase also offers comprehensive consulting packages, and has more than 1000 consultants on staff around the world.
Sybase offers end-to-end solutions --from tools to databases -- that cover OLTP, data warehouse, mass deployment, and electronic commerce computing environments. However, if all you need is a point solution, Sybase promises that each of its components can integrate with any environment. This strategy underlies the entire Sybase product line. Sybase executives believe that this will be a viable product strategy for the next generation of client/server and Internet computing, and will lead them to sunnier days.
Theresa Rigney is DBMS's features editor, and the editor of OTJ and STJ, technical journals for Oracle and Sybase product users, respectively.
Interviewed by Theresa Rigney
What is the history of Sybase?
Sybase really started with the idea that the LAN would become a dominant part of computing. With the LAN, two things changed related to data: One is that there is no one spot that has the data -- there's no central point. The other thing the LAN changed is the place where software is built. So we used this as the basic premise for our database and product strategy. Originally, we called it "client/server," but then backed off that name. We focused instead on affordable online databases -- databases in which response time is critical on the OLTP side. The biggest impact of our first products was that they enabled more applications to be written. It used to be that to do an OLTP application, there was a minimum cost of building it. But LANs increased the scale of systems that could be built, and that really started the client/server movement.
We then recognized the fact that a new technology creates new ways of doing things, but it doesn't eliminate the old stuff. So the idea behind our product strategy has always been to surround the other systems that already existed and extend their life. Our argument has always been: Don't replace something that already works; instead, build something new that solves a new problem, and integrate it with what already works.
If you look at where we are today, we're recognizing the fact that one implementation doesn't meet all needs. If you tell a design team that it's got to be the world's smallest and fastest database, and it has to do analytics and OLTP using one code base, you have what are called "universal servers." I believe these universal servers will be universally slow. We believe the implementation of OLTP has a different set of requirements than the implementation of a data warehouse, or a small home database. The idea is that they should look the same to the application developer and system manager, but the implementation should be specialized.
How does Sybase competitively position itself?
If we line up our strategy with the strategies of our main competitors, the interesting point is that there is no company that is a pure competitor. Oracle is a marketplace for our tools and middleware. We compete for some database accounts, but we are the data warehouse for Oracle accounts, and we can be the mass deployment database for Oracle OLTP. So while the marketplace would like to line everything up into competitive situations, Oracle is a market for about 80 percent of our product lines.
Informix is a very specific player in certain areas, primarily in the large data warehouse market, but we're concentrating on what we call the "interactive warehouse." This approach says that there are specific subject areas, and instead of having one percent of the people studying the warehouse 100 percent of the time, 100 percent of the people should access the warehouse one percent of the time. Response times must be quick, and you can't guess what the questions are going to be ahead of the time. Our strategy with Microsoft is very simple: Every API it invents, we support.
What excites you about the future?
The biggest change in the next few years will be the shift to using computers as a way to let people communicate. One of the things the Internet will enable is the opportunity for companies to decentralize further and create electronic alliances with other companies.
In the next couple of years, there will also be the opportunity to redefine what a computer really is. Most people can foresee that communications costs could change, and the Internet protocols will be the way that things work. Eventually, display devices will change to accommodate people's needs and requirements for less bulky, smaller, portable devices. This will further accelerate the idea that computers are there to help you communicate.
What do you think of the network computer idea?
It's silly to make a PC dumber. Dumb at half the price won't sell. What would make sense is smarter, cheaper, and more mobile. The PC industry right now is dominated by computers that are more mobile -- and people pay a premium for them. So imagine if computers become more mobile, cheaper, and more powerful. That will sell.
Is there an answer to Oracle's Sedona in the works?
It's already here. Our strategy is based on our PowerBuilder and Optima products, from our Powersoft division. This strategy uses the Microsoft component architecture as the communications device -- which we think is a lot more mainstream. A way to think about this is that a component encapsulates things so that it looks like a well-defined function. Applications simply attach themselves to components. And components are language-independent, so they could be built in Java, C++, or PowerScript. How you map and store these components depends on your performance optimization plan.
How is PowerBuilder 5.0 going to get that old energy back?
There are a couple of ways. One is that PowerBuilder lets you do either PC-based client/server, or Internet with the browser within the same application. So productivity on the Web for business applications increases dramatically. The second thing is that PowerBuilder generates compiled code, and now you have the ability to select whether you want to use C++, Java, or PowerScript. All the tools work with the same component architecture, and the architecture is 100 percent compatible with the present, and 100 percent compliant with what Microsoft is saying the future of computing will be.
What happened with the old Gain technology?
Gain was a technology that was very rich with multimedia, but it had no delivery infrastructure. Along comes the Internet, so we've taken the technology, re-implemented it from scratch to optimize it for the Internet, and renamed it media.splash.
What is Sybase doing in the OLAP market space?
OLAP is an industry in transition. An emerging marketplace is one where OLAP is middleware. In other words, it represents the way things should look to the viewer, but not the way it is stored. So that allows us to take things, transform them, and do orders of magnitude more work in less storage space.
I also believe that the custom user interface for the data warehouse will become extinct. It will be replaced by the Web browser. This is very important when you look at our visual component strategy. Warehousing will turn into a set of visualization components that plug into browsers for universal access.
What about object-relational?
We have one main belief in this area: Specialization of implementation yields better performance. For specialized data types -- and there's really just a handful of them -- you can build optimized implementations for them, and then use our component architecture. The specialization will be transparent to the user, but it will run a lot faster.