When I was a boy, one of my favorite movies was The Time Machine. The film- based on a book by H.G. Wells-tells the story of an inventor who creates a device to carry him thousands of years into the future. I often imagined what it would be like to have such a machine and to explore the vast reaches of time.
There's little doubt that AS/400 managers would also like to get their hands on a time machine to find out how IBM will enhance the AS/400. Armed with this knowledge, they could develop their information systems strategies with ease and win the respect and admiration of corporate managers.
In a small sense, this article is a time machine; I've designed it to take us on an excursion into the AS/400's future. Before this journey begins, I'll briefly review the IBM strategies that will affect its premier midrange system. During our trip, we'll stop at four points: the summer of this year, the fall of 1996, the fall of 1997, and early 1999. At each point, we'll see how the AS/400's hardware and software architecture have changed since our previous stop and discuss the potential impact of these changes on the AS/400 community.
Before starting, I must mention that the future we will see will take place only if IBM sticks to its current plans. I've gained access to these plans through extensive interviews with IBM executives and the AS/400's chief architects. However, their future choices (and those of other companies) could alter the course of history. This is one of the unfortunate sticking points about traveling to the future. It's not a big one, however, because IBM has a strong commitment to making this future a reality.
Open Blueprint: A Quick Review
As we discussed in last month's article (see "IBM AS/400 Strategy: The Next Three Years," MC, December 1994), IBM has charted its future course-and that of the AS/400-in a strategy it calls Open Blueprint. At the heart of Open Blueprint is a plan to create a family of high-performance systems that:
o Network and share applications with virtually any computer. o Use common hardware and software components to reduce production costs. o Offer a consistent, object-oriented application development environment. o Make it easy to integrate client/server computing into any enterprise.
IBM calls this family of systems the Workplace Environment-a family that includes the AS/400, the RS/6000, and PCs running OS/2. The company has also decided that the AS/400 will be the server of choice within Workplace. In this role, the AS/400 will be a data server, object repository, and network manager to IBM and non-IBM systems across enterprises.
To become this kind of system, the AS/400 will have to evolve rapidly over the next several years. As we step into our time machine to witness this evolution, we'll be looking at two key areas: the hardware and software architectures. In future articles, we'll talk about changes in other areas such as application development, openness, and client/server. Buckle your seatbelt. Our first stop is summer 1995.
New RISCs, New Rewards
From our new vantage point, we can see an AS/400 that looks like the Advanced Series of 1994 on the outside, but is quite different on the inside. The system uses 64-bit, reduced instruction set computer (RISC) chips that conform to the PowerPC standard. The chips probably run at speeds between 150 and 200 megahertz (MHz) and deliver up to twice the performance of previous AS/400 processors.
This CPU switch will be only the start of the hardware changes. IBM may double or triple the number of buses within each system and increase bus speeds by several times. It could also increase memory capacities two to four times. Because of these extensive changes, customers upgrading from Advanced Series models will probably have to replace system buses as well as processor and memory boards.
When it comes to upgrades, IBM plans to offer plenty of options. All systems- from the D-series to the latest Advanced Series models-should be able to upgrade to RISC without losing their old serial numbers. In addition, IBM may give owners of new AS/400 Advanced 36s an upgrade path-a viable option because the RISC-based systems will run both SSP and OS/400.
Meanwhile, owners of RISC (and possibly non-RISC) AS/400s could get some new I/O processors (IOPs)-one IOP acting as a Lotus Notes server, another providing communications with other systems over Asynchronous Transfer Mode (ATM) lines. These and other IOPs will play an important role in the evolution of the system.
The software changes of the future are just as dramatic. All the new systems will require an OS/400 release that IBM currently calls V3Rx. This release, which IBM plans to offer at no charge to V3R0M5 and V3R1 license holders, will be unique in many ways. Unlike previous releases, V3Rx will support both OS/400 and SSP, the S/36 operating environment. These two operating systems will sit on top of a new, object-oriented machine interface called Systems Licensed Internal Code (SLIC). Through SLIC, applications running under OS/400 and SSP will invoke and communicate with each other.
The new release will offer one other surprise: it will run only on the new RISC-based AS/400s. All new releases of OS/400 after V3Rx will run exclusively on RISC-based systems because SLIC will use 64-bit addressing. The current machine interface only addresses 48 bits. The restriction will probably have the following effects:
o Users will be able to run all their current applications on the new systems without modification. If the applications have observability, V3Rx will automatically recompile the programs on the new system. o Customers will be able to run applications generated under V3Rx on non-RISC systems that are at the V3R0M5 or V3R1 level. These systems should detect that the application was compiled under V3Rx, and compensate accordingly. o Customers who want to take advantage of a new feature that only V3Rx supports will have to upgrade to a RISC-based AS/400 or persuade IBM to offer the feature on a non-RISC release.
The incompatibility of V3Rx with non-RISC systems could cause problems for some AS/400 customers, but IBM executives feel these problems will be minimal. As they see it, V3Rx will offer few functional improvements over OS/400 V3R1, so incompatibility should not be an issue during 1995. In the following years, they expect that customers will be happy to upgrade to a RISC-based AS/400 because of the upcoming improvements in price and performance. It remains to be seen whether the AS/400 community will see the situation in the same light.
Incompatibility concerns aside, the 1995 systems will fulfill many of the promises IBM has made. However, Big Blue will not rest on its laurels. As we travel to the fall of 1996, the AS/400 will evolve once more.
A Whole New Workplace
Before our time machine even comes to a halt, we can see the AS/400 has changed once more. The central processors are 30-40 percent more powerful than last year and the memory looks different. Instead of 16-megabit (Mbit) memory chips, the 1996 systems probably use faster 64Mbit designs.
While these changes enhance system performance, others let the AS/400 handle a variety of new tasks. For starters, IBM plans to put PowerPC chips in all its IOPs and build them from standardized components, making them more powerful and less expensive. The added performance will let IBM put more intelligence on the boards. The company will use that intelligence to give the new IOPs a high degree of independence from the CPUs and to let them communicate directly with each other. IBM may even give them a new name: application engines.
Application engines will let AS/400s perform a wide range of functions offline from the CPUs. To handle database queries, IBM could offer an array of application engines to run parallel queries against DB2/400 and other databases. Some engines may function as servers running AIX or the Taligent application development environment. Instead of buying separate server systems to handle such functions, customers will merely add another application engine to the AS/400 and load the appropriate microcode for the desired task.
IBM could expand the hardware in other directions. It could offer systems with six or even eight CPUs. It may also use ATM connections between clustered systems. This would increase the number of systems allowed in a cluster as well as the maximum distance between systems. The ATM connections could also let customers tightly couple multiple systems together to form single-system images. These changes will dramatically increase the performance and availability of linked systems.
To support these enhancements, IBM will introduce OS/400 Version 4 as a chargeable upgrade for all RISC-based AS/400s. While the new version will make the hardware advances possible, it will also offer important new facilities of its own. Among these, the most important one will be a 64-bit microkernel that supports an OS/2 operating system personality. This personality will let the AS/400 function as an OS/2-based server to a wide variety of client systems.
The microkernel may also support a group of services that OS/400, SSP, and OS/2 applications can all access. These common services will handle a variety of tasks consistently across operating system personalities with each personality accessing the services as if they were another operating system. Some services that will probably find their way into this module are AnyNet, AnyMail, Ultimedia, SystemView, and OpenDoc. For definitions of these services and other terms, see 1.
The microkernel may also support a group of services that OS/400, SSP, and OS/2 applications can all access. These common services will handle a variety of tasks consistently across operating system personalities with each personality accessing the services as if they were another operating system. Some services that will probably find their way into this module are AnyNet, AnyMail, Ultimedia, SystemView, and OpenDoc. For definitions of these services and other terms, see Figure 1.
The new software architecture will offer another important feature: interprocess communications between SLIC and the microkernel. Through interprocess communications, applications running on one operating system personality will work cooperatively with those running on another personality. For instance, a client running Windows may access a multimedia application on the OS/2 server to request data from a DB2/400 database. The OS/2 application could pass the request to an OS/400 application, receive the data, access Ultimedia services to revise some figures, then show the results to the client on a Windows graphical interface. This would all be transparent to the Windows user.
Some readers may have noticed that we have ignored AIX and UNIX. IBM executives have often talked about running AIX as an operating system personality on the AS/400; however, the system's planners are reassessing this strategy because the AS/400 of 1996 will be UNIX-friendly without an AIX personality. In two years, OS/400 will largely comply with SPEC 1170, a specification that makes operating systems compatible with UNIX applications. As a result, many vendors may port their UNIX software to OS/400, making an AIX personality less valuable.
Regardless of what happens to AIX, the AS/400 will be a destination of choice for server software in 1996. Its standards compliance will let it network with any computer, run thousands of new applications, and connect to open systems peripherals. Its applications engines could let customers configure the system as a database query engine, a file and print server, or a network manager without sacrificing its traditional functions. Its performance will lead to AS/400 clusters with over 1,000 times the horsepower of a 9404-B10. In addition, the 1996 models will offer a fully visual, object-oriented development environment that will exceed or equal anything else on the market.
With all these enhancements, the AS/400 of 1996 will be a true enterprise server system. However, the best will get even better. As we return to our time machine and push the clock forward to the fall of 1997, we'll encounter even more improvements.
PowerPC...The Next Generation
We've now traveled almost three years into the future. While the AS/400's details are getting a little hazy, we can still notice a striking change in its internal architecture. There's a new PowerPC processor in this system, and it's different from anything we've seen before.
To explain this difference, we need to backtrack for a moment. During 1995 and 1996, many of the PowerPC processors in the AS/400 will be multiple-chip designs, because manufacturing technology cannot yet put all the circuits on one chip. Unfortunately, multiple-chip designs suffer a performance penalty because the circuits are farther apart, forcing electrons to travel greater distances.
By 1997, however, IBM expects to have the technology to put all the processor's circuits on one chip. The resulting CPU will be a PowerPC processor that can run at 300 to 400MHz-nearly twice as fast as previous designs-leading the AS/400 into a whole new performance range.
By using this next-generation chip in six-processor and eight-processor systems, IBM could offer AS/400s with over 350 times the performance of a 9404- B10. Since these systems will require much greater I/O and memory bandwidth, there is a strong likelihood that upgrading to the 1997 systems will require complete bus and memory replacements.
IBM may also use communications enhancements to improve the AS/400's capabilities within system clusters. By 1997, ATM data transfer rates may exceed a gigabit per second-over six times faster than the current top speed of 155Mbits. IBM will use these faster ATM links in AS/400 clusters containing up to 32 systems, allowing the largest clusters to offer more than 3,500 times the performance of a 9404-B10 and letting customers link systems that are more than a kilometer apart. The same links will tightly couple smaller numbers of AS/400s into single-system images.
IBM may also put its new line of faster processor chips into AS/400 application engines. This would radically improve the speed of the functions the engines perform, such as file serving, print serving and database queries. IBM could also offer customers a variety of microcode personalities to download onto the engines, allowing customers to reconfigure their application engines to perform specific tasks as computing needs change.
To support the 1997 systems, IBM will probably announce a new release of OS/400-V4Rx-to support all the hardware changes and offer some enhancements of its own. For instance, the V4Rx microkernel may support new operating system personalities for the Macintosh operating system, AIX, other varieties of UNIX, or native Windows NT. These personalities would dramatically expand the AS/400's support for thousands of new server applications.
The new release may also support a virtual application engine. Unlike regular application engines, virtual application engines will have no dedicated hardware resources; instead, they will be microcode entities that borrow main memory and CPU resources on an as-needed basis. IBM planners expect that entry- level AS/400s will be the primary users of virtual application engines as they will be more affordable than hardware-based application engines.
We are now ready to take one last leap forward in time. When we arrive in the first quarter of 1999, we'll see the outlines of a whole new system architecture.
Toward the 21st Century
As we step from our time machine, we find ourselves in a world that only the most visionary AS/400 developers inhabit. The system they are working on is far from complete, but a close look at it reveals a great deal about the AS/400 of the 21st century. The system will definitely have faster processors, but they may not get much of their increased speed from simple hardware improvements. Instead, their performance gains could come from a technology known as Very Long Instruction Word (VLIW) computing.
To describe it as simply as possible, VLIW takes certain functions that modern RISC chips perform and gives those functions to the compilers. Currently, PowerPC processors can execute several instructions simultaneously, but the chips must predict which sequence of instructions they can safely execute in parallel. This prediction process consumes valuable time and slows down the processors. With VLIW, highly sophisticated compilers will assume the task of predicting the proper instruction sequence. During compilations, they will use their knowledge of the CPU and its instruction engines to line up long sequences of instructions-or very long instruction words-for simultaneous execution. By relieving the CPUs of their prediction duties, VLIW systems will dramatically improve performance.
To support VLIW and its performance increases, IBM will probably introduce OS/400 Version 5. What could make Version 5 unique is an extensive use of workflow technology within the operating system. When an AS/400 customer loads Version 5, the operating system will ask the customer about operational policies and business processes. This information will give the system enough knowledge to intelligently interact with operators and users and make decisions on its own.
For instance, an AS/400 that encounters an error during an overnight payroll run could automatically detect where and why the problem occurred, then search for and notify the appropriate persons. It could also determine whether or not to restart the application on its own, where to restart it, and what steps to take to backup or restore data. It could also analyze the situation and design new business rules and procedures to handle similar problems in the future. Users could customize the rules to guide the system in its decisions.
The incredible speed of these systems could let customers implement many innovative technologies including speech recognition, multiple-person video conferences, and virtual reality interfaces. By the 21st century, many AS/400 users may talk to their systems more often than they use a keyboard. The AS/400 will talk back, offering guidance based on the company's business processes and each user's work habits.
Back From the Future
We've gone as far into the future as we can, so it's time to go back. As we return, let's consider how the changes we've witnessed could affect the AS/400 community. We can, with reasonable confidence, predict the following things:
o The AS/400's price/performance will improve dramatically. By late 1997, customers may be able to purchase three to four times more performance per dollar than they can today. At upgrade time, most customers will find they can get a performance boost for much less money than in the past. o Most upgrades will involve more than swapping a few boards. Any move from a non-RISC AS/400 to a RISC-based system will require extensive modifications. Customers who stay with the most recent generation of hardware-or even a generation behind-will have to upgrade hardware every year and pay for a new OS/400 version every other year. o The AS/400's software functionality will expand at a breathtaking rate. As a result, customers will have to invest their hardware savings on operating system upgrades and staff training. They will also spend more time modifying their applications to take advantage of the new functions. o Many customers will decide they do not want to change as quickly as the AS/400 architecture is changing, and will stick with their non-RISC models. Since they will not be able to run OS/400 V3Rx or above, the AS/400 community will split into two distinct groups, leaving software vendors to decide if they should enhance applications for both non-RISC and RISC platforms. Inevitably, non-RISC customers will face a gradual erosion of software and hardware support.
Though a split in the AS/400 community could be difficult for vendors and customers alike, it may be necessary. For the AS/400 to continue to have a future, it must evolve rapidly to compete as a server in an increasingly client-based world. If it does not succeed in this new role, it will hurt all customers, including those who are happy with the AS/400 as it is.
We are now at the end of our journey. As you leave our time machine, remember that we will continue to explore the AS/400's future in upcoming issues. I hope you'll join us for the next trip.
Lee Kroon is an industry analyst for Midrange Computing.
The AS/400's Future: A Brief Time Trip
Figure 1 Definitions of Terms
AnyMail - An evolving group of middleware services that will simplify communications between different E-mail systems on IBM computers. AnyNet - A set of multiprotocol transport services that will let IBM systems network with each other and non-IBM systems over SNA, TCP/IP, IPX, or NetBIOS links. Through AnyNet, the AS/400 will support all four of these network protocols. Application Engines - A new hardware device that will replace the AS/400's I/O processors in 1996. Unlike I/O processors, application engines will use standardized hardware components and communicate directly with each other without CPU support. IBM and its customers will be able to reconfigure them for various tasks. Asynchronous Transfer Mode (ATM) - A communications standard that defines network services for the transmission of data at speeds of up to 155Mbps. ATM offers better support for multiple data types (such as voice and video) than other standards. As a result, IBM intends to standardize on ATM in its future networks. Interprocess Communications - An ability of the future AS/400's microkernel architecture to support communications between applications running on different operating system personalities. IBM plans implementation on the AS/400 by the end of 1995. Microkernel - A small amount of extremely efficient code that performs the most basic and universal 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. Open Blueprint - An IBM plan that outlines the company's goals and strategies for all of its systems. The AS/400 plays a key role in the Open Blueprint strategy. OpenDoc - A group of services that will let AS/400 users create, manage, and share compound documents with each other and with non-AS/400 systems. OpenDoc will be a key part of an IBM initiative to add workflow management to all of its Workplace Environment systems. Systems Licensed Internal Code (SLIC) - A layer of object-oriented code that will act as the new machine interface for RISC-based AS/400s. As a machine interface, it will make hardware resources available to two operating systems: OS/400 and SSP. The RISC-based AS/400s will be able to run both operating systems simultaneously. SystemView - A group of system management products that manage a wide variety of IBM systems using a consistent interface. IBM is expanding the SystemView product line to include facilities that manage remote and non-IBM systems. Ultimedia -A family of IBM multimedia services that lets users create, edit, access, and transmit multimedia documents. Ultimedia services are already appearing on the AS/400, RS/6000, and OS/2-based PCs. Very Long Instruction Word (VLIW) Computing - A computing method that uses highly intelligent compilers to help processors run multiple instructions simultaneously. The compilers determine which instructions the processor can safely execute in parallel, then assemble these multiple instructions into single, long instruction words. Virtual Application Engine - A microcode entity that borrows main memory and CPU resources on an as-needed basis to perform a specialized task. IBM plans to announce virtual application engines in 1997. Workflow Computing - A form of computing in which information systems use explicit business rules and procedures to regulate the flow of information between workers. The AS/400 will increasingly support workflow computing over the next several years. Workplace Environment - A software architecture that IBM will implement on the AS/400, the RS/6000, and PCs running OS/2. At the heart of this architecture is a microkernel that will let these systems run multiple operating environments, otherwise known as "personalities." The AS/400 will act as IBM's enterprise server within the Workplace family of systems.
The AS/400's Future: A Brief Time Trip
Figure 2 An AS/400 Architectural TimetableHARDWARE ARCHITECTURE 1995 PowerPC-150-200MHz Memory capacity increases by 2-4 times Improved bus bandwidth New I/O processors (IOPs): o Lotus Notes IOP o ATM IOP o Others 1996 PowerPC-200-250MHz 64Mbit memory chips 6-way (8-way) systems IOPs become application engines: o PowerPC used in all o Peer-to-peer communications o Standardized hardware New application engines: o Parallel database query o AIX/UNIX (under assessment) o Taligent Improved AS/400 clusters: o CPUs connected via ATM o More than seven systems o Greater distances between systems Tight system coupling creates single-system images 1997 PowerPC-300-400MHz 64Mbit memory chips 8-way systems Improved bus bandwidth Improved clustering/coupling: o More systems o Greater distances Improved application engines: o Faster CPUs o User programmability 1998/1999 PowerPC uses Very Long Instruction Word computing for performance gains 256Mbit memory chips Improved bus bandwidth Faster application engines Note: Red text indicates there is a lower than average degree of certainty that this product or technology will ship during the year. SOFTWARE ARCHITECTURE 1995 OS/400 V3Rx 64-bit addressing Systems Licensed Internal Code supports: o OS/400 o SSP Interprocess communications: o OS/400 to SSP o SSP to OS/400 1996 OS/400 V4 64-bit microkernel supports OS/2 and common services Interprocess communications established between OS/400, SSP, OS/2 and common services Architectural support for: o Single-system images o New application engines o Intersystem ATM 1997 OS/400 V4Rx Microkernel supports: o AIX (Under assessment) o Other operating systems Architectural support for: o Larger clusters o Larger single-system images o New application engines Virtual application engines 1998/1999 OS/400 Version 5 Workflow computing implemented in OS/400
MC Press Online