Around two and a half millennia ago, a Greek philosopher named Heraclitus challenged the thinking of his day by declaring that reality constantly changes. To refute those who believed in an unchanging universe, he would point to the nearest moving body of water and ask his opponent to put his foot in the same river twice. He won many arguments that way.
Heraclitus would feel right at home with IBM's plans for the AS/400 over the next three years, for the plans make it unlikely that customers will ever put their hands on the same AS/400 twice. Like a river, however, the AS/400 will remain the same as it constantly changes. Three years from now, customers will still be able to run today's applications without modifications.
Because IBM's AS/400 blueprint is part of a much larger strategy, we'll start by reviewing Big Blue's master plan. We've based this review on extensive interviews with the chief architects of the AS/400. From there, we'll explore the role of the AS/400 in that plan and how the system will change to fit that role. Next month, we'll dig deeper into the details of the AS/400 strategy and present snapshots of the projected system in 1995, 1996, and 1997.
IBM's Brave New World
In the early 1990s, a small circle of top IBM executives began discussing a long-term plan for the company. They were concerned over IBM's faltering revenues and loss of market share and believed that radical change was necessary. They wanted the company to jettison its "not invented here" mentality, form partnerships with other vendors, open its systems, and eliminate much of the duplicated effort going on in the divisions. Among these executives was John Thompson who was at the time, manager of the AS/400 Division.
When Louis Gerstner took over as IBM's chairman in 1993, he put this group in charge of much of the company. Thompson, for instance, now manages all server systems from the ES/9000 mainframe to the RS/6000, as well as the development of OS/2. Gerstner asked Thompson and other executives to coordinate the efforts of IBM's divisions and establish a unified plan of action. This plan, however, is not like the failed monolithic strategies of the past such as SAA or AD/Cycle. It is flexible, relies heavily on popular non-IBM technologies, and draws upon contributions from all divisions and outside vendors. It is, as its name implies, an Open Blueprint for IBM's future.
On paper, the Open Blueprint plan is hundreds of pages in length, but its aims are simple (see 1). Of the several strategies IBM has adopted to accomplish these goals, the most important states:
On paper, the Open Blueprint plan is hundreds of pages in length, but its aims are simple (see Figure 1). Of the several strategies IBM has adopted to accomplish these goals, the most important states:
IBM and its industry partners will create a group of open, industry-standard, hardware and software components. IBM's divisions will share these components and use them to create a variety of systems.
It's difficult to underestimate the importance of this strategy. If IBM succeeds in building its systems from a pool of common, industry-standard components, it will automatically meet many of its Open Blueprint aims. By reducing the number of components it develops, the company will slash production costs and shorten product cycles. The strategy smooths the path for any-to-any networking and simplifies the porting of data and applications. With this strategy, IBM will create a relatively seamless enterprise computing environment in which developers and end users alike can be more productive.
The strategy is crucial for AS/400 users, because plans call for virtually every one of the common technology components to find its way into the system over the next three years. Let's take a closer look at these components.
o PowerPC. PowerPC is a reference standard that dictates how IBM and other companies will build a large family of reduced-instruction-set computer (RISC) chips. Some of these processors such as the 601 and 604 are 32-bit chips. Others such as the upcoming 620 are 64-bit chips. A variety of computers already use 32-bit PowerPC chips including some Apple computers and the RS/6000. In 1995, IBM plans to release an AS/400 that runs on a 64-bit PowerPC chip that is similar to the 620.
o Storage devices and peripherals. In a world of commodity hardware, it makes little sense to custom-build a peripheral for a single system. To compete, IBM already uses the same disk drive technology in its PCs, RS/6000s, and AS/400s. Its Pennant subsidiary manufactures printers that connect to multiple systems. These trends will continue to grow in the coming years.
o Input/output interfaces. IBM plans to standardize its systems on a handful of I/O interfaces. The most important of these is the Peripheral Connect Interface (PCI) standard that is rapidly appearing on PCs. IBM will continue to support other standards such as the Small Computer System Interface (SCSI) and interfaces for 5250 and 3270 devices. However, it will increasingly avoid the SCSI, 5250, and 3270 interfaces when designing peripherals.
o Communications devices and protocols. IBM wants to simplify the confusing world of communications protocols. At the same time, it needs a standard that will support the extremely high speeds and multiple data types of the future. To meet both aims, the company has adopted Asynchronous Transfer Mode (ATM). Currently, ATM networks can transfer data, voice, and video information at speeds of up to 155Mbps. Future ATM devices may send data at speeds greater than one gigabit per second. Until ATM fully matures, IBM will continue to support other communications standards including token-ring, Fiber-distributed Data Interface (FDDI), frame relay and ISDN.
To a great extent, these common technologies are already standards in today's computing world. However, many of the software components of the Open Blueprint are still on the drawing board. Three basic software strategies are vital to Open Blueprint: the Workplace Environment, the common services initiative, and an object-oriented development environment. Because there is more confusion about these elements, we'll take more time to explain each of them in detail.
The Workplace Environment
The Workplace Environment is one of the boldest and most important innovations in Open Blueprint. It proposes an innovative systems software architecture for three of the company's platforms: the AS/400, the RS/6000, and all OS/2-based PCs and servers. These systems comprise the Workplace family of operating environments.
Workplace borrows a page from the design of the AS/400. Like the AS/400, all Workplace platforms will isolate their operating systems from the underlying hardware. This approach lets hardware evolve independently of the operating systems while preserving customer investments in applications. Unlike the AS/400, Workplace does not put a layer of microcode between the operating systems and the hardware. Instead, it uses a microkernel architecture.
A microkernel is a small amount of extremely efficient code that performs the most basic operating-system functions. It is the only code that interacts with the hardware, so it is the only code that changes when the hardware changes. It also provides a standard interface to the operating systems above it. This design gives computer designers an incredible amount of flexibility (see 2). They can rapidly build several microkernels, each of which runs on a different hardware platform, and run multiple operating systems on top of each microkernel. As long as an operating system uses the microkernel's interface, it can run on any hardware platform.
A microkernel is a small amount of extremely efficient code that performs the most basic operating-system functions. It is the only code that interacts with the hardware, so it is the only code that changes when the hardware changes. It also provides a standard interface to the operating systems above it. This design gives computer designers an incredible amount of flexibility (see Figure 2). They can rapidly build several microkernels, each of which runs on a different hardware platform, and run multiple operating systems on top of each microkernel. As long as an operating system uses the microkernel's interface, it can run on any hardware platform.
The IBM microkernel design provides another important attribute-interprocess communications. In essence, this lets multiple operating personalities run simultaneously on the microkernel and communicate with each other. For instance, an AS/400 user could ask OS/400 to run an OS/2 or an AIX application. The system would then use the microkernel to pass the request to an OS/2 or AIX personality. The personality would run the application and return the results to OS/400 through the microkernel. These capabilities will probably appear on the AS/400 in 1996.
The Common Services Initiative
While the simple functions performed by the microkernel are universal to almost all operating systems, most operating systems also provide other more sophisticated services. These services include system management, network communications, and security. In the near future, operating systems will offer other services such as E-mail access, multimedia, software distribution, and workflow and document management.
Because all Workplace operating systems will provide these services, IBM developers plan to package them as a group of common software components that access the microkernel like an operating system. IBM will refer to these components as the common services. The operating system personalities could invoke these services as they would invoke any other operating system function. This eliminates the need to redesign each service for every operating system.
As part of the Workplace family, the AS/400 will probably possess all of the common services by late 1996. Many of these IBM services already have names.
o SystemView provides a consistent set of systems management products, just as NetView provides one for networks.
o Ultimedia is a family of multimedia services that is already appearing on the AS/400, RS/6000, and OS/2.
o AnyNet provides a multiprotocol transport that lets systems network over SNA, TCP/IP, IPX, or NetBIOS links.
o AnyMail offers an integrated E-mail environment, including APIs that simplify communications between systems.
o OpenDoc will provide an environment to create, manage, and share compound documents. OpenDoc is a key part of an IBM initiative to add workflow management into the common services used by all Workplace systems.
Just as common technology components should improve the quality and productivity of IBM's system development efforts, so object-oriented programming should improve the quality and productivity of application development efforts. Realizing this truth, Open Blueprint calls for the creation of a consistent, object-oriented development environment that will run on MVS-based mainframes, the AS/400, the RS/6000, and PCs running OS/2.
The new environment will contain several key technologies. IBM has already shipped some of them; one example is the Smalltalk-based Visual Age. Others, such as a C++ compiler, will probably appear on the AS/400 late in 1995. By late 1996, the full development environment should be in place. Here are its major elements.
o Development Tools. These include a C++ compiler and VisualAge. In a certain sense, the development tools also include traditional languages such as RPG and COBOL. IBM will provide facilities that let RPG, COBOL, and objects created by the new development tools invoke each other on the AS/400. The company will also offer system class libraries that contain prewritten objects. By using the development tools, programmers will be able to access these objects to invoke and customize selected AS/400 system functions.
o System Object Models. One of the key advantages of object-oriented programming is that developers can share and reuse objects to perform tasks in many different systems. To promote sharing and reuse, IBM is building system facilities that let developers exchange and link objects written in different languages. IBM calls its facilities the System Object Model (SOM) and Distributed System Object Model (DSOM). While SOM facilitates object sharing and linking at the local system level, DSOM facilitates sharing and linking between systems. Both SOM and DSOM comply with an emerging industry standard, the Common Object Request Broker Architecture (CORBA), that lets objects communicate with each other.
o Taligent. This joint project between IBM, Apple, and Hewlett-Packard encompasses a wide variety of object-oriented technologies. It includes a visual application development environment and a screen interface to that environment called People, Places and Things. Taligent also includes a large number of entities called application frameworks. Frameworks are customizable groups of objects that provide the foundation for an entire application. A framework may track inventory, run a payroll, or handle a system-level function such as saving and restoring data.
Each of the Taligent partners will use the new technology as it sees fit. IBM plans on incorporating much of the Taligent development environment into VisualAge. It could include the People, Places and Things interface in an OS/2 personality that will run on the AS/400 in 1996. Many of the Taligent application frameworks will also find their way onto the AS/400, where developers will access them through SOM and DSOM.
Taken together, these components will revolutionize the way AS/400 developers build systems. However, the revolution will take time to reach everyone. Becoming productive in the brave new world of objects requires training, vendor support, and a critical mass of pretested objects.
Open Blueprint's Impact on the AS/400
We've already said a lot in passing about Open Blueprint's effect on the AS/400. Now, let's take a closer look at precisely what the effect will be.
As we have already mentioned, the Workplace Environment is a crucial element of Open Blueprint. By the same token, the AS/400 is a crucial element of Workplace. In fact, IBM has decided that the AS/400 will be the enterprise server within the Workplace family of operating environments. As such, it will act as a repository for data from all systems, function as a network manager, and provide a focal point for application development in a distributed object environment. The future AS/400 should be to the distributed computing environments of the late 1990s what the mainframe was to the far simpler environments of the 1960s and 1970s.
To become IBM's premier enterprise-wide server, the AS/400 will have to evolve rapidly. At the same time, this evolution must not threaten customers' current investments in hardware or software. IBM has specific plans designed to meet both of these objectives which are detailed in the next two sections.
The Hardware Impact
The first and most striking change to the AS/400 will be a rapid increase in system performance. Over the next three years, a fivefold performance increase is projected for the largest AS/400s. As a result, we can expect high-end systems in 1997 that have more than 350 times the horsepower of the 9404-B10 that IBM announced in 1988.
To achieve this kind of throughput, Rochester plans numerous changes to the hardware. It plans to replace all its current CPUs and most of its input/output processors (IOPs) with RISC-based PowerPC processors. The number of processors in top-end systems will be increased. AS/400 designers have plans for six-way and eight-way systems in the 1996 timeframe.
These powerful processors will require faster internal communications. To accommodate this requirement, the typical AS/400 will have at least three times as many buses and 30 times the aggregate bandwidth of current systems. As a result of the rapid improvements, future system upgrades will typically involve extensive bus replacements as well as new processors and memory boards.
IBM will also improve the AS/400's ability to work in tandem with other AS/400s. It plans to do so through two forms of intersystem connections: clusters and single-system images.
Currently, customers can cluster up to seven AS/400s using a feature called OptiConnect/400. OptiConnect lets clustered systems access each other's databases over fiber-optic cables at distances of up to 100 meters and run applications against the databases. Under clustering, system managers must decide where to locate various applications and how to partition databases between the systems to maximize performance.
Over the next three years, IBM plans to increase the maximum number of systems in a cluster to as many as 32. The maximum distance between systems may increase to a kilometer or more. This will make it easy for AS/400 sites to assure high system availability in case a disaster strikes. In urban areas, customers could put multiple AS/400s on different power grids and replicate their databases over a cluster in real time. If a disaster struck one site, another site with its database could immediately take over.
Within two years, IBM is also likely to announce the coupling of multiple systems into single-system images. This feature will, in effect, make multiple AS/400s run as one system. Unlike clustering, there will be no need to decide where to locate applications or parts of databases. The systems will dynamically manage storage allocation as if all the systems were a single AS/400.
By combining rapid increases in single-system performance with advances in multiple-system coupling, IBM will greatly enhance AS/400 performance. As 3 shows, an AS/400 cluster could easily yield 3,000 to 4,000 times the performance of a 9404-B10 by 1997.
By combining rapid increases in single-system performance with advances in multiple-system coupling, IBM will greatly enhance AS/400 performance. As Figure 3 shows, an AS/400 cluster could easily yield 3,000 to 4,000 times the performance of a 9404-B10 by 1997.
Meanwhile, the AS/400's IOP architecture will also experience a metamorphosis. At present, the IOPs communicate with each other through the central processing units. For instance, the transmission of data from a disk storage IOP to a workstation IOP still requires several CPU cycles. By late 1996, the IOPs will probably communicate directly, without CPU support.
This change will be necessary to support the transmission of new data types such as voice and full-motion video. Normally, a CPU handles several jobs simultaneously and interrupts each job at numerous points. As a result, data travels around the system in small bursts. This creates problems for voice and full-motion video data, which need to travel at constant, uninterrupted rates to make any sense. By taking the CPU out of the transmission loop, IBM will effectively segregate the AS/400's bursty alphanumeric processing from its multimedia traffic.
At the same time, IBM plans to introduce new IOPs that will offload more tasks from the CPUs. The company has already announced an IOP that manages file serving to LANs-the File Server IOP. Over the next three years, customers can expect other IOPs to handle tasks such as print serving, mail serving, and other client services. Rochester is also drawing up plans for an IOP array that will run parallel queries against DB2/400 databases. The array could alleviate the performance bottlenecks that AS/400s often experience when users run ad hoc queries.
The Software Impact
While all these changes are significant, none of them will make future AS/400s incompatible with current applications. The credit for this is largely due to the system's software architecture. Even so, this architecture will undergo massive changes of its own.
Two new structures will take the place of the existing microcode layer-Systems Licensed Internal Code (SLIC) and a 64-bit microkernel. To a great extent, SLIC will perform the same functions as the microkernel; but it will coexist with the microkernel for two reasons. First, SLIC will appear in 1995, at least one year earlier than the microkernel. Second, plans call for SLIC to support different operating system personalities than the microkernel. While SLIC will support OS/400 and SSP personalities, the microkernel will support an OS/2 personality as well as the common services discussed earlier.
As a result of these changes, the AS/400 should fully support three operating system personalities by late 1996: OS/400, SSP and OS/2. IBM also plans to run Microsoft Windows NT on the AS/400 under the OS/2 personality. Other operating system personalities, such as an AIX or a Macintosh personality, could appear later. Communications between SLIC and the microkernel will let applications on one personality interact with applications running on another personality.
Once the AS/400 supports multiple operating system personalities, it will be difficult for anyone to say it is a closed, proprietary system. To make sure the critics have no ammunition, IBM is also planning to make the OS/400 personality comply fully with Spec 1170, a standard that specifies approximately 1,170 APIs that every version of UNIX should contain. When it meets the Spec 1170 standard, the AS/400 will be more compatible with UNIX applications than many versions of UNIX itself.
In addition to the operating system personalities, plans call for the AS/400 to support a full set of common services. Certain operating system personalities (especially SSP) will not access all of the services because of inherent functional limitations. When a personality can access a service, it will do so through a set of standard APIs. Because these services will be common to all personalities, they will have a similar look and feel whether a user accesses them on a 5250 display terminal, a PC running Windows or OS/2, or an AIX workstation.
Meanwhile, IBM will revolutionize the AS/400's application development environment. That revolution will begin in 1995 when IBM plans to add a C++ compiler to the system. At the same time, IBM hopes to roll out pieces of SOM and DSOM on the AS/400. These system facilities, especially SOM, will play a key role in the integration of third-generation and object-oriented programming languages. Through SOM, developers will be able to access, modify, and link applications written in C++, Smalltalk, RPG, COBOL, and C. In many ways, SOM is a superset of the Integrated Language Environment (ILE) that IBM has already announced for RPG, COBOL, and C.
The 1995 shipments of object-oriented technology should give AS/400 developers enough tools to start experimenting with object-oriented development. In late 1996, those developers will probably get the rest of the tool kit. The most important additions will be the Taligent development environment and a set of application frameworks. While the development environment will run on client workstations under OS/2, the application frameworks will reside on the AS/400. With these new tools, most AS/400 developers should have the critical mass they need to begin building object-oriented production applications.
Putting All the Pieces Together
We have now reviewed all of the components in IBM's AS/400 strategy. Each one plays a vital role in transforming the system into an industry-leading enterprise server. By late 1996, IBM plans to announce most of the components. By 1997, we expect that virtually all of them will exist on the AS/400 in a relatively stable form. 4 pieces the components together to form a picture of the new system.
We have now reviewed all of the components in IBM's AS/400 strategy. Each one plays a vital role in transforming the system into an industry-leading enterprise server. By late 1996, IBM plans to announce most of the components. By 1997, we expect that virtually all of them will exist on the AS/400 in a relatively stable form. Figure 4 pieces the components together to form a picture of the new system.
This picture has a lot of parts to it, but a top-to-bottom review helps make sense of it. On the top, server applications and development tools will access any of the operating systems that have interfaces to support them. They will also access common services and program models. Most of these applications and tools will communicate directly with OS/400. OS/400 and the other operating system personalities, services, and program models will freely interact with each other through consistent interfaces.
Underneath the operating systems, the microkernel and SLIC will provide the interface to the hardware. These components will use interprocess communications to interact with each other and integrate the operating systems and services directly above them. They will also communicate with the IOPs that attach to the hardware bus.
The IOP level will handle a variety of specialized tasks for the AS/400 without burdening the central processing units. The IOPs will have the intelligence to communicate with each other over the bus. They will also access SLIC and the microkernel and, through them, communicate with all the higher levels in the architecture.
The Challenge for the AS/400 Community
To get the AS/400 from its current position to the system that 4 represents-and to do so while preserving customer investments-the AS/400 Division will have to work hard over the next three years. At the same time, AS/400 managers and programmers will have to log some overtime to deal with all the changes. They will have to decide which of the AS/400's advanced capabilities are important for their companies, when to upgrade to the most current OS/400 release or hardware platform, and how to integrate new functions with legacy applications.
To get the AS/400 from its current position to the system that Figure 4 represents-and to do so while preserving customer investments-the AS/400 Division will have to work hard over the next three years. At the same time, AS/400 managers and programmers will have to log some overtime to deal with all the changes. They will have to decide which of the AS/400's advanced capabilities are important for their companies, when to upgrade to the most current OS/400 release or hardware platform, and how to integrate new functions with legacy applications.
Because these decisions are crucial ones, we'll provide more details on the AS/400's evolution in next month's issue. We'll analyze the major system revisions slated for late 1995 and late 1996 and discuss their implications for customers. We'll discuss the upgrade strategies customers may adopt and suggest some tactics that could save you time, effort, and money. Finally, we'll get out our crystal ball and forecast what the AS/400 will look like in the 21st century.
Lee Kroon is an industry analyst for Midrange Computing.
IBM's AS/400 Strategy: The Next Three Years
Figure 1 IBM's Open Blueprint Goals
1. To create any-to-any networking between all IBM and non-IBM systems. 2. To give customers the ability to easily move data and applications between all IBM and non-IBM systems. 3. To give developers the kind of tools that will rapidly increase their productivity as well as the quality of the applications they create. 4. To give end users the interfaces they need to easily send, access, and manipulate data, whether it is inside or outside the enterprise. 5. To speed up the hardware and software development process within IBM while reducing the cost of the development effort.
IBM's AS/400 Strategy: The Next Three Years
Figure 2 The Microkernel AdvantageUNABLE TO REPRODUCE GRAPHIC
IBM's AS/400 Strategy: The Next Three Years
Figure 3 Projected AS/400 Performance GrowthUNABLE TO REPRODUCE GRAPHIC
IBM's AS/400 Strategy: The Next Three Years
Figure 4 The 1996/1997 AS/400 ArchitectureUNABLE TO REPRODUCE GRAPHIC
MC Press Online