|
|
|
| home |
The Enterprise Infrastructures workshop at LISA 2003 is full. I can't accept new registrations, and I'm having to turn away good people -- I hate that. Watch this space -- we may do a related evening BOF later in the week.
This workshop is dedicated to the premise that a good infrastructure, whether departmental, divisional, or enterprise-wide, is a single loosely-coupled virtual machine or "enterprise cluster", with hundreds or thousands of hard drives and CPU's. An enterprise cluster is heterogeneous, containing multiple operating systems and hardware platforms. Enterprise cluster nodes are commodity items; none are dedicated to a permanent role. The lines between "client" and "server" are blurred, if not removed completely.
Administration of this infrastructure is fully automated, and uses the same code paths as disaster recovery. Any node, regardless of size or current role, can be destroyed at any time without impacting users for more than a few minutes. Any node, once destroyed, can be rebuilt unattended without resorting to backups. Users experience free seating, single signon, and a single system image. Sysadmins spend their evenings at home.
In this workshop, we'll discuss concepts, terminology, and tools which meet this criteria. We'll cover divergence, convergence, congruence, and ordering, and the cause and effect of each. We'll give attention to all phases of infrastructure administration, including bootstrapping a new infrastructure, host installation, upgrades, patch management, and disaster recovery of the entire virtual machine. We'll cover financial and organizational benefits, as well as career implications.
The format of the workshop is facilitated discussion of short presentations; participants are encouraged but not required to present their own findings.
Day: Monday, October 27th
Time: 9:00am - 5:00pm
Location: Garden Salon 1
Bring cash -- we're going to order out for lunch, probably pizza.
These two papers are probably the most common denominator among people attending the workshop -- if you familiarize yourself with the vocabulary used in them ahead of time it will help prevent arguments over semantics:
In addition, see the URL's in the schedule below. If you review those ahead of time you'll be able to follow the presentations much more easily -- they're going to be moving extremely fast.
This is a very tight schedule, and I'm taking some risks by even thinking of letting this many people stand up. But it's a big topic, almost a "mini-LISA". You'll need to manage your time down to the minute; this is going to take a certain level of discipline. If anyone doesn't think they can do this, then let me know ASAP.
Take a look at the schedule below. This is only the current version -- it will continue to change without notice, so keep an eye on it. The amount of time you have is shown in the "Allocated" column next to your name. The amount of time I think you asked for is shown in the "Desired" column. I will be tracking actual time usage throughout the day and having gnumeric adjust "Allocated" accordingly. If anyone runs undertime, then that will give everyone else a little bit more. If anyone runs over, then you'll be taking time away from everyone else, not just the next guy; so don't run over.
If your name appears in the schedule, that means that you should plan to be on your feet, helping to guide what will be a rapid discussion and adding helpful content for the duration of your time slot, not talking the entire time yourself. Use your slides, but don't overdo it -- six hours of solid presentations will put us to sleep.
The format should be interactive, rather than trying to get your entire thesis out during that time slot. I'd recommend using your slides as an overview of what you're working on, to introduce yourself to the group, and then follow up in later conversation. Remember, you're going to be in the room all day, and will have many hours to add information to what is essentially a day-long structured discussion. After reading these four paragraphs, if anyone still wants a longer time slot, let me know.
| As of | 10/27/03 15:45 | |||||
| crunch factor | 0.630769230769 | |||||
| day end | 17:00 | |||||
| Desired | Allocated | Start Time | Major Section | Who | Topic | Notes |
| 9:00 | ||||||
| 0:30 | 0:18 | 9:00 | Overview, definition of terms and concepts | We'll go over terminology and framework, to get everyone on the same page in terms of what we're trying to do. See the URL's in the recommended reading above. | ||
| 0:13 | 0:08 | 9:18 | David Snyder | CNN.com | ||
| 0:05 | 0:03 | 9:27 | Version control | CVS, SVN, Bitkeeper | ||
| 0:25 | 0:15 | 9:30 | Andrew Cowie | SubVersion as a backend | http://subversion.tigris.org/ | |
| 0:15 | 0:09 | 9:46 | Directory servers | DNS, LDAP etc. | ||
| 0:05 | 0:03 | 10:15 | Piotr Poznanski | Quattor CDB | http://cern.ch/quattor/ | |
| 0:05 | 0:03 | 10:18 | Host install (imaging) tools | |||
| 0:15 | 0:09 | 10:21 | Thomas Lange | FAI | http://www.informatik.uni-koeln.de/fai/ | |
| 0:15 | 0:09 | 10:30 | Break | |||
| 0:15 | 0:09 | 10:40 | Steve Traugott | SystemImager, etherboot | http://www.systemimager.org/ | |
| 0:05 | 0:03 | 11:10 | Piotr Poznanski | Quattor A.I.I. | http://cern.ch/quattor/ | |
| 0:05 | 0:03 | 11:13 | Authentication | |||
| 0:20 | 0:12 | 11:16 | Derwin Skipp | Arizona State University | ||
| 0:15 | 0:09 | 11:40 | Network file storage | General discussion, AFS, DRBD, OpenGFS, etc. | ||
| 0:05 | 0:03 | 11:49 | File replication | General discussion, SUP, rsync, cfengine | ||
| 0:15 | 0:09 | 11:52 | Thomas Lange | fcopy | http://www.informatik.uni-koeln.de/fai/ | |
| 0:10 | 0:06 | 12:05 | Luke Kanies | rsync triggers, cfengine | ||
| 0:10 | 0:06 | 12:11 | Patrick McNeal | Radmind | http://radmind.org | |
| 0:05 | 0:03 | 12:17 | Patch and local binary management | |||
| 0:10 | 0:06 | 12:20 | Patrick McNeal | Radmind | http://radmind.org | |
| 0:05 | 0:03 | 12:27 | Piotr Poznanski | Quattor SPMA | http://cern.ch/quattor/ | |
| 1:00 | 0:37 | 12:30 | Lunch | |||
| 0:10 | 0:06 | 13:30 | Graydon Dodson | RPM-based management of binaries | pitfalls | |
| 0:10 | 0:06 | 13:36 | Steve Traugott | Isconf 2 | http://www.infrastructures.org/bootstrap/isconf.shtml | |
| 0:20 | 0:12 | 13:42 | Luke Kanies | Isconf 3 Genesis | http://sourceforge.net/projects/isconf/ | |
| 0:20 | 0:12 | 13:55 | Stephen Schaefer | An Isconf 3 Rollout | ||
| 0:20 | 0:12 | 14:07 | Steve Traugott | Isconf 4 | ||
| 0:05 | 0:03 | 14:20 | Configuration management (non-binary) | General discussion of the issues, how ordering doesn't fit this, some alternatives for Isconf 4, etc. | ||
| 0:05 | 0:03 | 14:23 | Graydon Dodson | RPM-based management of configuration data | pitfalls | |
| 0:05 | 0:03 | 14:26 | Piotr Poznanski | Quattor NCM | http://cern.ch/quattor/ | |
| 0:15 | 0:09 | 14:29 | Break | |||
| 0:13 | 0:08 | 15:24 | Mark Roth | PSGconf | http://www.uiuc.edu/ph/www/roth/LISA03-psgconf/ | |
| 0:15 | 0:09 | 15:38 | Paul Dlug | Handling External and Transient Data | ||
| 0:15 | 0:09 | 15:47 | Monitoring | General discussion, Nagios -- http://www.nagios.org/ | ||
| 0:10 | 0:06 | 15:56 | Backups | General discussion, tapeless, etc. | ||
| 0:10 | 0:06 | 16:03 | Disaster Recovery | What it's all for... | ||
| 0:05 | 0:03 | 16:09 | Putting it all together | |||
| 0:15 | 0:09 | 16:12 | Bryan Gartner | StiBET | http://www.infrastructures.org/workshop/STiBETExecSummary.pdf | |
| 1:00 | 0:37 | 16:22 | Rapid Response Time Slot | If we can stay on or close to schedule, then we'll be able to use this time slot to cover as a group some of the things that usually splinter off into hallway conversation. By the end of the day, we'll have a load of things that we'll want to discuss -- I'll be putting together the schedule for this last slot as we go. Expect topics like the need for standards and standard API's, professional issues, and "presenters" who I elect on the spot because of something they said earlier in the day. ;-) | ||
| 0:00 | 17:00 | Close |
|
© Copyright 1994-2003 Steve
Traugott,
Joel Huddleston,
Joyce Cao Traugott
In partnership with TerraLuna, LLC and CD International |