The Autonomic Udell If there was ever a set of metaphors subject to potential marketing its the biological:
|
Being and nothingness. Jon Udell: "'[A]utonomic' for IBM has become what '.Net' is for Microsoft: an umbrella marketing term that encompasses everything and nothing in particular." Good analysis by Jon as usual. I think both MS and IBM are onto the Next Big Thing, but they can't figure out how to explain it. The marketing efforts make the confusion worse. One can only hope that the companies listen to the technology and the markets as they go forward, rather than letting the slogans define what they build. [Werblog] |
IBM has provided an 8-point definition of autonomic computing and if it sticks to it and journalists like Jon Udell remain vigilant, the value of these functional concepts are less like to dilute.
As grid computing takes hold, autonomic computing plays a vital role in enabling scalable service. Without complex adaptive systems, patterned after biological theory, managing the sheer number and interplay of elements, applications and policies becomes untenable. As processing, storage and applications move towards the domain utility service they would experience the same complexity constraints as utility bandwidth service.
Telecom carriers have been in the business of transforming utility elements into services for some time now. And in absence of adaptive systems they have experienced systemic failure. At the network element level, the vast majority of inventory records are inaccurate, despite advances in auto-discovery, leading over 85% of customer orders failing to achieve straight through processing to the point of provision. As a result, network and service management is largely a manual process. People don't scale well. And failure to automate results in high SG&A costs and an unsustainable business model. Autonomic computing uses decentralized agents to adaptively manage a system, and in contrast to centralized management, scales.
Back to marketing abuse. Last week I recieved a press release from Cloudmark claiming they cracked the Genetic Code of Spam. Essentially, Cloudmark uses collaborative identification of spam, the spam message is inspected for common semantics (a "spamGene"), which are aggregated as "spamDNA" and fed to a Bayesian classifier which determines when a message is blacklisted.
I was about to slam them, but called my friend and life science expert Zack Lynch for some fact checking. It is a stretch to call Cloudmark a complex adaptive system, let alone "evolutionary." And its a little rediculous to say they have cracked the genetic code. However, if you assume that the people enlisted for spam identification, those on SpamNet, are acting as autonomous agents and you buy into Bayesian statistical qualification as evolving, the metaphor holds to some degree.
From a business model persective, decentralized user feedback also enable the model to scale. These individual users experience vendetta-like satisfaction for pressing the "Spam" button, gain spam protection and enhance the enterprise solution. As the spam wars escalate, competitors with centralized management like Brightmail may experience diseconomies of scale.
11:18:02 AM 
|
|
