Invited Talk Proposal: "Why Automate?" Hi Invited Talk coordinators, I'd like to present an invited talk, or, if you'd like, chair a panel to address a common controversy in systems administration shops: automation vs. "just get it done". Some of you may have seen the 1998 LISA paper Joel Huddleston and I presented, "Bootstrapping an Infrastructure". In that paper, we discussed a sequence of steps which can be used to "bootstrap" an infrastructure into existence, using lots of automated tools and techniques. If you want to know more about me and my background, you can find my resume etc. at http://www.stevegt.com. An assumption we made, and tried to justify in that paper, was that automated systems administration is a good thing. Years later, we still see shops who don't see the need. Those shops do have a point: automated systems administration takes time; time to write the code, time to test it, deploy it, rewrite it when you realize your original assumptions were incorrect, and, far too often, time spent in meetings and hallway conversations justifying the political need. Deployment of even the most simple solution can easily take months. I'd like to outline three of the reasons why I generally prefer automation: 1. Cost Reduction A significant portion of the IT spending in an organization is labor cost. Automating enterprise systems administration decreases recurring labor costs. [boos and hisses from the back of the room as some of the more union-minded members of the audience decide "decrease labor costs" means "get paid less"...] ;-} I can provide anecdotal evidence from my own experience to support this, as interim IS director of Netscape, a VP at Chase Bank, etc. My wife, Joyce Cao Traugott , was until recently MIS financial analyst for a major chipmaker, and is currently digging up old contacts to get us the numbers. 2. Accuracy and Repeatability Humans tend to make more mistakes than machines do. Humans are not at their best when performing complex repetitive tasks, and tend to inject randomness into the process. I have yet to see a manually-maintained infrastructure in which all member machines are consistently administered. This randomness and unpredictability can be expensive for the organization. If we let machines control their own administration, accuracy and consistancy are more likely. 3. Knowledge Base Writing software is a process of recording knowledge which would otherwise be locked up in the systems administrators' heads. By recording and releasing this knowledge, we build and publish a knowledge base which can benefit both the organization and the computing community as a whole. Fewer sysadmins doing manual system administration will result in more time to write administrative software, creating a positive-feedback effect. I have repeatedly found that once an organization automates sufficiently to get "over the hump" of the trouble-ticket curve, managing new creative projects becomes easier. Depending on what format and time slot we set up for this, I may also take a few minutes to delve into an analogy I've come up with, comparing today's generally accepted manual SysAdmin practices with the early days of automobile manufacturing, in which individual craftsmen painstakingly built each unique vehicle from scratch. Mass production changed all of that, and brought us the economies of scale that allowed ubiquitous ownership of personal cars. We have a comparable situation in today's system administration community; we are still at the stage of the craftsman system administrator building and baby-sitting individual machines. This makes computing infrastructures unnecessarily expensive to own and troublesome to maintain. My cell phone is 800-852-5654, and home address is 472 Kings Court, Campbell, CA, 95008. You can show "Infrastructures.Org and TerraLuna, LLC" as my organizational affiliation. Steve