IBM aims to make e-business infrastructure self-managing

Self-managing Servers

IBM eServer: Project eLiza

April 27, 2001

Some malign microbe bent on havoc probably assaulted you today and, without your even knowing it, your immune system recognized the alien, marshaled a variety of systems, and attacked and destroyed it.

Irving Wladawsky-Berger, vice-president, Technology and Strategy, introduced Project eLiza at IBM's International Technology Symposium

In short, your body managed the situation without calling your attention to it. We do that repeatedly through the course of the day, and not just with disease.

Run to catch a taxi and your autonomic nervous system raises your respiration rate, adjusts your heart rhythm, and increases your energy levels. Our body manages these functions autonomically. We don't give it a second thought; in fact, our minds are probably on work or home, on the more significant things.

In a sense, that's what Project eLiza is all about. Just as the human body manages itself without conscious intervention, Project eLiza's objective, as announced at a Server Group meeting with analysts, is to make the e-business infrastructure, and in particular IBM eServer, storage and software, a self-managing system.

Why are we declaring this Grand Challenge of a self-managing e-business infrastructure now? It all begins with the relentless proliferation of faster, more powerful technology. Exponential improvements in price/performance have become almost commonplace, but IBM Research studies indicate price/performance, in a number of cases, is actually super-exponential, that is, the rate of improvement is actually going up.

Indeed IBM has been a major contributor to the advance of technology with new materials like copper that have raised conductivity, Silicon-on-Insulator (SOI) with its very high frequency operations; our microdrive, which holds colossal volumes of information on a drive the size of a quarter, and flat panel displays with resolution that exceeds that of the human eye.

Technology has reached price points that make it feasible to embed full-function computers in any device, so specialized appliance servers are being deployed in the infrastructure -- single-function servers, like caching, firewalls and content management, that are easy to use and deploy, and relatively low in complexity.

The same technology being used in appliances is also being aggregated and scaled into ever larger, more powerful systems, from the very small to the enormous -- gigantic clusters with colossal processing power. Meanwhile customers more and more need support for all their mission-critical applications, which requires increasingly powerful mainframe-class systems.

Emerging from this profusion of technology is a complex, multifaceted infrastructure, one both distributed and centralized; used by small and large businesses alike; focused on content and transactions; supporting old and new applications; partially owned and partially outsourced.

Clearly, such an environment poses major challenges to our industry: among them, how to manage such a far-flung, complex, and profoundly indispensable infrastructure, specifically, how to take advantage of the latest management technology, while keeping costs under control.

The world faced a similar challenge early in the 20th Century when the telephone began to proliferate. Initially, telephone calls were switched manually by human operators, a perfectly good solution when Alexander Graham Bell and his man Thomas Watson were the only ones with phones.

But had the same management model been followed as the telephone's popularity grew, Ma Bell before long would have had to employ every man, woman and child on earth to manage the volume of calls. The solution was automation; the infrastructure had to manage itself.

The IT industry is on a similar track because of, among other things, an already critical shortage of IT skills. According to IDC, this year will see a shortfall of 455,000 IT personnel, growing to 2 million worldwide in 2004, half of them in the US.

Larger and larger systems, networks of networks, billions of users, a trillion devices, colossal volumes of transactions: all mean new levels of complexity, more and more pieces that must be configured and tuned, more and more points at which systems can fail or intruders penetrate.

In addition, complexity is taking on a new dimension because increasingly we are dealing with whole new kinds of workloads. Where once it was only data; now it is data plus telephony, entertainment, advertising, rich media, massive digital libraries, and a host of other workloads that will no doubt grow over time.

Managing complexity is not a new issue to IBM. We began years ago to address it as our mainframes were called on to do more and more things -- to go from batch processing to online transaction processing, for example.

So, if there is one thing IBM developers have learned how to do over the years it is to manage complex systems. And a major goal of Project eLiza is to spread mainframe-like management qualities, along with new advances from the research community, throughout our IBM eServer, storage and software families.

BM's new self-management focus will have a number of objectives. For example, Server Group developers and researchers working together will seek to move dynamic workload management, built on the Intelligent Resource Director, to another level -- Heterogeneous Workload Management.

Our Advanced e-business Council -- composed of high-level IBM developers, researchers, and customers -- is at the point of defining a Heterogeneous Workload Manager (HWM) that will handle as many as 64 classes of work, spanning the entire network rather than just a single server. It will track each job throughout the network, and automatically tune the network and operating system to reduce delays and assist various classes of work in achieving performance goals.

We are working in other areas to make self management a reality. Scalable cluster management, for example, will benefit from a major research project. It's called Oceano, a highly integrated, parallel system that monitors everything from compute and network resource use to application workload and database performance, and automatically allocates compute resources to various workloads.

Self-management is the goal in other areas as well, distributed server management, for example, where we want to establish wireless management of anything from anywhere. In addition, we will build on the Research Division's work on Blue Gene to develop truly self-healing systems.

Project eLiza formalizes these and many other self-management efforts underway in Server Group, Software Group, Storage Systems, Research, and across all of IBM, and focuses them on a coordinated, common goal -- building on the automated management function already available in our individual servers to create self-managing systems. This is a goal of the most ambitious sort, one truly worthy of IBM's development teams around the world, and a fitting challenge for IBM's talented researchers.

The immediate object of Project eLiza is to drive the e-business infrastructures we are helping build to become self-managing. Just as the autonomic systems one finds in nature regulate and protect us, Project eLiza is working toward systems that can configure, optimize, heal, and protect themselves, while the user focuses on the more significant things.

Why is it named "Project eLiza?"

ELIZA was also the name of a program, one of the first to try to bring "intelligence" to a computer, artificial intelligence as it became known. ELIZA permitted natural-language conversation between a human and a computer.

The program was designed to reflect your words back to you. Were you to say "I need my Mommy", ELIZA would reply, like a good Rogerian therapist, "Tell me why you need your Mommy?" In fact, ELIZA was the inspiration for a program called "PC Therapist," written by Joseph Weintraub, a mainframe programmer with a degree in Psychology.

PC Therapist then became the first AI program to pass the Turing Test for artificial intelligence. Conceived by mathematician Alan Turing, the test proposed that a judge talk to a computer and a man over phone lines and try, by conversation alone, to identify the human. If he was convinced by the computer program, then it was indeed artificially intelligent.

Why the lizard?

A few years ago IBM researchers, using assumptions in Ray Kurzweil's "The Age of Spiritual Machines," estimated the processing power of the IBM supercomputer Deep Blue as roughly equivalent to a lizard's brain. People began to ponder why such a creature would need all that brain power.

The answer is apparent in the fact that the lizard has a far more complex task than a computer devoted to a single objective. Getting a machine to do one thing well -- like Deep Blue's beating a chess Grand Master or Blue Gene's unlocking the mysteries of protein folding, is truly a Grand Challenge, requiring tremendous power. But the environment is completely controlled and comparatively predictable. It is a single-minded pursuit under tightly circumscribed conditions.

The lizard, on the other hand, like all living organisms, has had the grandest of challenges -- survival, beginning millions of years ago, in a completely unpredictable environment requiring constant, instantaneous, instinctive -- autonomic -- responses to new demands and dangers. "It's a jungle out there!" That's why the lizard needs all that intelligence.

IBM eServer has its own jungle to contend with, complete with totally unpredictable demands from unknown numbers of users, sudden threats from predators, and the thousand natural shocks the system is heir to. For eServer, like the lizard, the Grand Challenge is survival through intelligent self-management.

Project eLiza Questions and Answers

Q: Why do we need self-managing servers?
A: Project eLiza addresses a looming crisis in the business world -- in five years, given the rapid proliferation of technology and the growing shortage of trained systems administrators, the majority of corporations will not be able to manage their systems using the current computing technologies. Project eLiza's goal is to give businesses the ability to manage systems and technology infrastructures that are hundreds of times more complex than those in existence today.

Q: What does the 'eLiza' name and the lizard represent?
A: The original eLiza was a project in the mid-1960s to develop seamless communication between people and machines. The name eLiza also harkens back to the "Deep Blue" project, when we estimated that our chess-playing supercomputer had the transaction capacity of a lizard brain. Self-managing servers look beyond transactions towards prediction and decision-making capabilities.

Q: How will Project eLiza be managed?
A: The leaders of Project eLiza are Ross Mauri, vice president of eServer Development, and Irving Wladlawsky-Berger, vice president of eServer Technology and Strategy. In addition to a wide range of Server Group professionals, eLiza will draw on the talents of thousands of IBMers from Research, Technology Group, Software Group and Global Services.

Q: Are we delivering any self-managing capabilities today?
A: Yes, even as we launch this initiative, we are already delivering self-managing technologies to our customers. Examples include:

• Intelligent Resource Director, which automatically allocates resources for multiple jobs according to demand
• eServer clustering, which updates software and manages workloads across massive clusters with a single operation
• Chipkill technology, which is used in our mainframes to virtually eliminate memory failures, and is now available on IBM eServer p620 and p660.

All statements regarding IBM's future direction and intent are subject to change or withdrawal without notice, and represent goals and objectives only.