Tuesday's sessions began with the Advanced Topics Workshop; once again, Adam Moskowitz was our host, moderator, and referree. We started with cable management 101 in separating the large bundle of cat-5 cable into strands for us to connect our laptops to the local network switch, as there are enough of us that we overload the wireless access point. We followed that with introductions around the room. Due to a variety of reasons, several of the Usual Suspects weren't at this year's workshop. Despite this, in representation, businesses (including consultants) outnumbered universities by about 4 to 1; over the course of the day, the room included 6 LISA program chairs (4 past, present, and future, up from 3 last year) and 11 past or present members of the USENIX, SAGE, or LOPSA Boards (up from 5 last year).
After the setup (untangling the bundled cat-5 cables, connecting them to attendees' laptops and the local switch, getting the moderation software up and running, setting the correct time zone on the server, and so on), we did the administrative announcements and overview of the moderation software for the new folks. Then, we went around the room and did introductions.
Our first topic was on management versus technology. About half of the room are managers either full- or part-time, and we discussed some of the problems we have interacting with our management. Some of the concerns were knee-jerk, oo-shiny managers, cultural differences when your manager is in another country, and managers who used to be in Sales positions. Some folks discussed their specific situations and asked for advice in solving them. One common suggestion was to communicate differently; remember that managers (especially those on the financial side who approve capital budgets) tend to speak business-speak and not techie. They don't care about the new gee-whiz neato-peachy-keen technology, but rather in how this new thing will solve their problems and provide a decent return on investment.
A side discussion took place on cultural issues that differ from the North American standard most of us are used to, and how that can affect communication styles as well as resumes.
After the morning break, we discussed career concerns. Most of the people in the room have 15 or more years of experience, and many of us have more than 20 years of experience. Assuming that retirement isn't an option (for whatever reason, be it financial or boredom), what's the right thing to do if you wind up looking for work? One person discussed how he neglected to ask questions of the company during the interview process and after accepting the offer and working for some length of time, realized he was a bad fit for the position. One suggestion for avoiding this in the future was to ask better questions before accepting any offer; another suggestion was to consider contract-to-permanent position since it gives both parties an out without the company having to let a senior person go. One topic that fell out of this is whether there's a technical growth path at your company or if senior implies a management position. Another topic was on the technology refresh rate for individuals, and whether staying generalists or becoming specialists was the better course of action (consensus seemed to be the former). For those who are retiring in the sufficiently-near future, you have to make peace with what's good enough and remember that "good enough" isn't necessarily the same as settling. Do what needs to be done and find enjoyment in that. Whatever you're doing, remember to work, to play, and to live, not just to exist. Do things to keep yourself interested and awake at your job; don't just settle into a rut.
We next discussed patterns as an abstraction layer in system administration. There's apparently some controversy in patterns; some think they're a good way to abstract problems and provide a common shorthand; others think they're not worth the electrons and doubt they're applicable to systems adminitration.
After our lunch break, we discussed things we should have done differently. One example was the whole IPv6 rollout. A lot of places don't see any need to deploy it and wondered if ignoring it now will cause problems later. CFOs don't see the benefit from or ROI in another technology refresh. Widespread adoption of something new requires there be some kind of benefit on the business level, and right now businesses don't tend to see a need for IPv6. Right now, there is very little IPv6-only content out there; if services or content were made IPv6 only that could drive folks to convert, assuming that their equipment is IPv6-capable (not all SOHO equipment is, for example).
We next had a brief discussion on the different uses of DNS and search engines. Both are treated as a way to find resources: DNS is a way of finding IP address-to-name mappings and search engines are a way of finding some specific document or site.
We next went around the room to discuss our favorite new-to-us tools from the past year. Common examples were AFS and ZFS, certain KVM cards, load balancers, ntop, Puppet, Ruby and jRuby, spam management tools, svk, tcptrace, the iPhone, Time Machine in MacOS 10.5, virtualization, and zones in Solaris. Several people chose Puppet for their configuration management, mainly because it was faster to get it up and running and it had a sufficiently-low learning curve for installation and configuration.
Our next discussion was on virtualization. At least one participant has done nothing with it and wondered if it was worthwhile; the consensus seemed to be that there are some areas where it's not a benefit, such as a high-performance computing environment . Someone plugged this year's refered paper, "Decision Support for Virtual Machine Re-Provisioning in Production Environments," by Kyrre Begnum, Matthew Disney, Æleen Frisch, and Ingard Mevå, for performance statistics. Some environments are only using virtualization in non-production environments.
We next talked about delegating identity management. The example environment is a department within a university that uses automated identity management to manage authentication and authorization controls for students, faculty, staff, alumni, and guests (for example, investigators on research grants from other universities). The central IT organization can provide some of the information they need but not all of it. The question was posed, how can they cascade information management systems? The short answer is that processes need to be put into place to incorporate the data to allow for both provisioning and revocation and to make sure essential safeguards are in place such that a failure upstream (e.g., HR's database failing miserably) doesn't accidentally cause a disaster downstream (e.g., deleting all staff accounts).
The next major discussion was on distributed or geographically disparate personnel. We talked about what server infrastructure do you put in remote data centers for part-time workers; the answer generally depends on how many people and how much network traffic there is. Items to consider are VOIP or POTS phones, home directory access over the WAN, docking stations, a printer, reserved offices or cubes ("hotelling"), and possibly a conference room, depending on the size of the office and the data center. We also talked about tools for communication and collaboration; many use some form of instant messenger client (such as internal IRC channels or logged IM conversations), email, trouble ticketing systems, and wikis. If you are using any of these technologies, remember to include them in (as a critical part of, where need be) your disaster recovery planning.
Our final discussion was a quick around-the-room on the cool tool for next year. Ruby, Solaris 10, and virtualization were the most-commonly mentioned, with configuration management tools, Perl 6, VOIP, wiki deployment, and ZFS rounding out the list.
This year's Talkies Award goes to David Williamson.