Autonomic Computing

Rating:
Summary: Interesting overview
Review: Optimistic in its projections, and sometimes exaggerating what is currently possible, this book gives an interesting overview of the status of the 'autonomic computing' project of IBM. Targeted to the network manager reader, the book views autonomic computing as a new approach to computer and systems management. Reducing costs are its goal, as well as increasing the quality of the service. An autonomic system must, according to the author, have knowledge of itself, have the ability to configure and reconfigure itself, as well as self-optimize itself. It must also have the capability to self-heal, to protect itself, and the ability to discover knowledge of its environment and context, and then adapt itself if needed. Most importantly, the system must be able to deal with any type of environment, and be able to anticipate and adapt to the user needs. The complexity of computing systems though is what is considered to be the real driving force behind research into autonomic computing.

These requirements can be met, says the author, by the use of intelligent hardware and software. Just what constitutes intelligence is not really discussed by the author, and hence readers will have to impute their own notions of intelligence when reading the book. The notion of complexity is also not quantified explicitly, although the author does give a few criteria that signify it, such as the size and cost of a system, and the shortage of skilled labor. This last criterion though is somewhat troubling from the standpoint of the labor markets of today. There is definitely not a shortage of skilled IT labor; in fact, there is an overabundance of highly skilled IT workers, definitely more than the current market can absorb.

This reviewer does not know of a machine or system that has all the requirements that the author lists as necessary in order to be considered as a fully autonomous system. The author claims though that such systems are available, in that IBM customers are using it at the present time. He gives a few examples of firms that are using autonomous systems, but does not give the (quantitative) details of how these firms are using it, nor does he discuss just how much these systems improved the overall functioning and management of the systems deployed by these firms. There is ample evidence from the literature on artificial intelligence that will back up some of the author's conclusions however. Indeed, there are myriads of examples that will substantiate the claims attributed to IBM's Alan Ganek that the efficiency of a business increases as more human error is taken out of the loop. But again, there are no systems being used now that satisfy all eight of the criteria for autonomy.

Part 3 is the most interesting part of the book, for it goes into more of the details behind autonomic computing, including how to build these kinds of systems, and even how to implement it in personal computing systems. The author considers 'control loops' as the basic management elements in autonomic computing architectures. Control loops will hide complexity from end-users are able to communicate and negotiate with each other and with resources within and outside of the autonomic computing architecture. A hint of what the author considers as intelligent behavior arises in the discussion on control loops: they are able to request additional processing cycles when needed, can install software and upgrades, can restart a system after failure (but can they restart the systems that they are part of?), can initiate backups after daily processing, and can shut down systems are an intruder has been detected (but can they shut down the systems that they are part of if an intruder is detected in these systems?). Certainly if control loops are able to do all of these things they are emulating functions and tasks normally done by human engineers. The author considers the ability to do these things as a definite sign of intelligence.

The author considers open standards as essential for autonomic computing, and he devotes an entire chapter to explaining his reasons for this. Open standards give more agility and flexibility to users, he says, and he compares proprietary and open models in terms of reliability, interoperability, risk, speed of updates, quality, and cost. The Web Services interoperability standards play an essential role in the open source movement and the author believes, in autonomic computing.

Very helpful is the author's discussion in chapter 10 on how to implement autonomic computing. It is helpful because it does address the attitudes about autonomic computing among users and IT managers. From personal experience, this reviewer has encountered a lot of anxiety and caution in the use of this technology. The author is aware of this however, and therefore he defines the notion of different levels of autonomic computing, that will allow IT managers to incorporate autonomic computing in phases. There are five levels in all, labeled as 'basic', 'managed', 'predictive', 'adaptive', and 'autonomic'. The tools, processes, and skills become increasingly sophisticated as businesses progress through the levels. At the basic level, IT workers set up, monitor, and manage all systems. At the managed level, management technologies are deployed to allow administrators to collect and synthesize information using fewer consoles. Technologies able to pattern match, perform optimization, and give expert advice are deployed at the predictive level. At the adaptive level, systems can automatically take action based on information available to them. Then at the autonomic level, the systems satisfy all eight of the requirements for autonomy, and the IT infrastructure is governed solely by the policies and objectives of the business.

Autonomic computing is definitely an exciting and profitable development in information technology. It remains to be seen just how fast the business community will integrate it into their operations. With hardware and bandwidth costs dropping, and with software development and engineering becoming more automated with each passing day, autonomic computing seems to be the soundest and most reliable approach to IT systems management.

Rating:
Summary: Interesting overview
Review: Optimistic in its projections, and sometimes exaggerating what is currently possible, this book gives an interesting overview of the status of the `autonomic computing' project of IBM. Targeted to the network manager reader, the book views autonomic computing as a new approach to computer and systems management. Reducing costs are its goal, as well as increasing the quality of the service. An autonomic system must, according to the author, have knowledge of itself, have the ability to configure and reconfigure itself, as well as self-optimize itself. It must also have the capability to self-heal, to protect itself, and the ability to discover knowledge of its environment and context, and then adapt itself if needed. Most importantly, the system must be able to deal with any type of environment, and be able to anticipate and adapt to the user needs. The complexity of computing systems though is what is considered to be the real driving force behind research into autonomic computing.

These requirements can be met, says the author, by the use of intelligent hardware and software. Just what constitutes intelligence is not really discussed by the author, and hence readers will have to impute their own notions of intelligence when reading the book. The notion of complexity is also not quantified explicitly, although the author does give a few criteria that signify it, such as the size and cost of a system, and the shortage of skilled labor. This last criterion though is somewhat troubling from the standpoint of the labor markets of today. There is definitely not a shortage of skilled IT labor; in fact, there is an overabundance of highly skilled IT workers, definitely more than the current market can absorb.

This reviewer does not know of a machine or system that has all the requirements that the author lists as necessary in order to be considered as a fully autonomous system. The author claims though that such systems are available, in that IBM customers are using it at the present time. He gives a few examples of firms that are using autonomous systems, but does not give the (quantitative) details of how these firms are using it, nor does he discuss just how much these systems improved the overall functioning and management of the systems deployed by these firms. There is ample evidence from the literature on artificial intelligence that will back up some of the author's conclusions however. Indeed, there are myriads of examples that will substantiate the claims attributed to IBM's Alan Ganek that the efficiency of a business increases as more human error is taken out of the loop. But again, there are no systems being used now that satisfy all eight of the criteria for autonomy.

Part 3 is the most interesting part of the book, for it goes into more of the details behind autonomic computing, including how to build these kinds of systems, and even how to implement it in personal computing systems. The author considers `control loops' as the basic management elements in autonomic computing architectures. Control loops will hide complexity from end-users are able to communicate and negotiate with each other and with resources within and outside of the autonomic computing architecture. A hint of what the author considers as intelligent behavior arises in the discussion on control loops: they are able to request additional processing cycles when needed, can install software and upgrades, can restart a system after failure (but can they restart the systems that they are part of?), can initiate backups after daily processing, and can shut down systems are an intruder has been detected (but can they shut down the systems that they are part of if an intruder is detected in these systems?). Certainly if control loops are able to do all of these things they are emulating functions and tasks normally done by human engineers. The author considers the ability to do these things as a definite sign of intelligence.

The author considers open standards as essential for autonomic computing, and he devotes an entire chapter to explaining his reasons for this. Open standards give more agility and flexibility to users, he says, and he compares proprietary and open models in terms of reliability, interoperability, risk, speed of updates, quality, and cost. The Web Services interoperability standards play an essential role in the open source movement and the author believes, in autonomic computing.

Very helpful is the author's discussion in chapter 10 on how to implement autonomic computing. It is helpful because it does address the attitudes about autonomic computing among users and IT managers. From personal experience, this reviewer has encountered a lot of anxiety and caution in the use of this technology. The author is aware of this however, and therefore he defines the notion of different levels of autonomic computing, that will allow IT managers to incorporate autonomic computing in phases. There are five levels in all, labeled as `basic', `managed', `predictive', `adaptive', and `autonomic'. The tools, processes, and skills become increasingly sophisticated as businesses progress through the levels. At the basic level, IT workers set up, monitor, and manage all systems. At the managed level, management technologies are deployed to allow administrators to collect and synthesize information using fewer consoles. Technologies able to pattern match, perform optimization, and give expert advice are deployed at the predictive level. At the adaptive level, systems can automatically take action based on information available to them. Then at the autonomic level, the systems satisfy all eight of the requirements for autonomy, and the IT infrastructure is governed solely by the policies and objectives of the business.

Autonomic computing is definitely an exciting and profitable development in information technology. It remains to be seen just how fast the business community will integrate it into their operations. With hardware and bandwidth costs dropping, and with software development and engineering becoming more automated with each passing day, autonomic computing seems to be the soundest and most reliable approach to IT systems management.

Rating:
Summary: Let's try again
Review: [For some reason my first review disappeared without trace.]

Yes, I know it's an IBM Press publication, so dial up your "self-serving effluent" filters - but not too high. Overall this is a really useful book. While it's targeted at CIOs and their staffs (folks who have read, and bought into, the Autonomic Computing Manifesto), it's not afraid to dive the details and point at source code to back up the architectural diagrams. It discusses what's going on in the research community and what competitors are up to. And I like the way the author models "customer maturity"; the readiness and ability of customers to take up some of the things described in the book. I disagree with some of his numbers, but without this kind of model the temptation to believe one's own propaganda is irresistible.

There are a few goofs (mobile agents? please, no), as well as some yawning gaps (systems modelling and policy languages). And while it's reasonable to skip the IBM-heavy business stuff at the front on a first reading, don't put the book away without going back to it. In particular, don't skip chapter 2, on the costs of complexity.

And please ignore the petty criticisms of those who can't see past minor production issues. If they actually understood this book, they wouldn't be distracted by such irrelevant details.

I've bought two copies of this book so far and given them away to people that need to be thinking about this stuff. And no, they weren't business expenses.

Rating:
Summary: Let's try again
Review: [For some reason my first review disappeared without trace.]

Yes, I know it's an IBM Press publication, so dial up your "self-serving effluent" filters - but not too high. Overall this is a really useful book. While it's targeted at CIOs and their staffs (folks who have read, and bought into, the Autonomic Computing Manifesto), it's not afraid to dive the details and point at source code to back up the architectural diagrams. It discusses what's going on in the research community and what competitors are up to. And I like the way the author models "customer maturity"; the readiness and ability of customers to take up some of the things described in the book. I disagree with some of his numbers, but without this kind of model the temptation to believe one's own propaganda is irresistible.

There are a few goofs (mobile agents? please, no), as well as some yawning gaps (systems modelling and policy languages). And while it's reasonable to skip the IBM-heavy business stuff at the front on a first reading, don't put the book away without going back to it. In particular, don't skip chapter 2, on the costs of complexity.

And please ignore the petty criticisms of those who can't see past minor production issues. If they actually understood this book, they wouldn't be distracted by such irrelevant details.

I've bought two copies of this book so far and given them away to people that need to be thinking about this stuff. And no, they weren't business expenses.