• Add a Comment
  • Edit
  • More Actions v
  • Quarantine this Entry

Comments (2)

1 localhost commented Permalink

Dave, I know this doesn't have anything to do with definitions but in looking over all the autonomic computing stuff that IBM is doing I am curious about the direction. I am also interested in what you think of my view on it. I am building a rather complex product that plans to incorporate autonomic computing as a feature of it. Having built complex systems for year - too many to be certain - I have started to appreciate how difficult it is to handle the errors that pop up. In the past when an error was encountered in the execution, most times we just printed out a message (somewhere) and stopped execution (I remember the days of OS/360 MFT and the operator console). Today, while I am designing and writing the new system I have taken the approach that went an error happens in my code I will "write out some state" and invoke another system to hopefully fix things up. While this other system is busy working, the current system is suspended. The other system can cause the application to terminate, retry what it was doing when it failed or even just log a message and continue onward. For the "other system" I decided to use work from the "commonsense reasoning" area. My goal is to describe a set of rules, in simple terms, and let logic drive the outcome. An example of commonsense reasoning is how we figure out why a file can't be opened that should be present (these are things a good operator or admin would do to fix the situation). I was wondering what you thought of this approach. Comments?Chuck Wegrzyn

2 localhost commented Permalink

Hi Chuck, Thanks for your comments. I too remember the days of OS/360 and the operator console! We have come a long way but not far enough. With all of the tehcnological advances, one would think we would futher beyond systems coming to a screeching halt with the only hope being for some human to bring it back from dead. ( I am reminded of my laptop and the dreaded bluescreen of death that visited me last night! )The approach to "write out some state" and invoke another system to correct the situation via "commonsense reasoning" is indeed a sound approach. One of our AC partners is not only using this approach for software but also for all elements of the hardware as well including power back up. The idea is that there will always be an autonomic manager available to correct the problem regardless of the problem and state of the machine. The best advice I can offer is to align your work with the autonomic architecture and standards as much as possible to insure the kind of sustainability that stansdards can support and to leverage alot of the architecture and technological constructs that are already in place. This should provide a huge jumpstart. A good reference is the Autonomic computing zone on developerworks.http://www.ibm.com/developerworks Best of luck!

Add a Comment Add a Comment