My Report on Linuxfest Northwest

The eighth annual Fest was April 28th and 29th. There were almost 900 attendees this year. The biggest yet. This year they added a second day. So, for an additional small percentage of overall trip cost I was able to take in twice as many sessions.

• Load Balancing with Linux Virtual Server
  by Jed Reynolds, Bitrachet
• Hacking you Cellphone
  by Nimret Sandu, Nimsoft/Motorola
• OpenID and the State of Distributed Identity
  by Aaron Klemm, Unripped.com
• Network Management Best Practives
  by Thomas Stocking, GroundWork Open Source, Inc.
• Security Monitoring with Snort and Sguil
  by James Affeld, South Seattle Community College
• Scaling: LiveJournal's Story and Secrets
  by Brad Fitzpatrick, Silicon Mechanics
• Practical Honeypots
  by Michael DeMan, Open Access Network Services
• What Went Wrong with my Disaster Recovery Plan
  by Brian Martin, Martin Consulting Services

I saw on the other Montana LUG maillists that a crew was putting together a car pool to head over for the Fest. I first ran into Scott Dowdle in the first session. I first saw Warren Sanders that afternoon. I never did see Donnie.

Load Balancing with Linux Virtual Server
by Jed Reynolds, Bitrachet

The first session of the weekend and everyone was still working on coffee. At least, that is my excuse. If I had done my homework a bit better I would have realized that "linux virtual server" was not virtualization using Xen, VMware, KVM but Linux Virtual Server Project. No matter, as I would have attended this session anyway.

I found a blog entry Jed made after the Fest.

Up to this point my knowledge of load balancing amounted to DNS round-robin for routing traffic to multiple servers and Heartbeat for server failover. Jed introduced some new technologies with application in this area.
keepalived
ldirectord
Ultra Monkey

What looked like the most interesting design was Direct Routing. The drawback of this one is that the real servers in the cluster must be on the same "physical" network segment for exchanging arp. Using IP tunneling adds a bit of overhead which then allows the real servers to be geographically dispersed. I don't see this as buying anything since it is the load balancer front-end that needs to be geographically dispersed.

Hacking you Cellphone
by Nimret Sandu, Nimsoft/Motorola

A high-caliber presentation for writing software for cellphones. I took lots of notes. If I were to take the time to google the many terms I am sure I would learn lots of stuff about cellphones as an application platform. The big thing I took away from this is that not all cellphone services providers are equal and Verison is the worst of them in terms of restricting access to the cellphone you own.

I am going to just throw some of my notes here with no explanation. If you want to dive off in this direction googling on this stuff will get you going.

XHTML MP
phone capability database

• wurfl
• j2me polish
• phonescoop

ringtone creator - java library and application

sms ringtone via gmail forward to cellphone as multi-mime message

moto4lin

logica - open source smpp tools (sms protocol)

SMS has no client state so server ends up being a state machine arch

optional APIs

• location - JSR 179
• SIP - JSR 180
• bluetooth - JSR 82
• payments - JSR 229

jetty - web server

EclipeseME

Have fun. :-)

OpenID and the State of Distributed Identity
by Aaron Klemm, Unripped.com

I have been hearing about identity from Doc Searls for sometime now. This blog entry "Let's go bust some silos" is a good example and provides some background and context.

OpenID is a pretty simple way of managing login as participating websites. Yet it is only one possible solution to a small part of the overall problem of identity management.

Network Management Best Practives
by Thomas Stocking, GroundWork Open Source, Inc.

Thomas Stocking introduced a concept he called GCAP. I tried googling on it briefly but did not find anything relevant.

Gather
Collection
Analysis
Presentation

• Gather is comprised of various sensor technologies. Using such tools as fping, rmrtg, nmap, nagio, net-snmp.
• Collection
  captures the data output from the gather tools. Examples include syslog, syslog-ng, and RRDTool.
• Analysis
  might be provided by a tool such as SEC (simple event correlator).
• Presentation
  There are a number of tools for presenting data such as jasper, birt, network weathermap, Nagios, cacti, and RRDTool.

Besides promoting his own companies product he also mentioned Ganglia which is aimed at clusters and grids. Nedi is a network discovery tool. Or as described on its website a bunch of php scripts.

Security Monitoring with Snort and Sguil
by James Affeld, South Seattle Community College

I have used snort in the past. Getting a handle on its output is like trying to get a sip of water from a firehose. This is where sguil comes into play. James gave a brief demo that was just... wow! The caveat here is that one still has to do all the setup and fine tuning of snort rulesets, the storage engine (barnyard), and such that had to be done before but now instead of a wild firehose of output, sguil provides a gui console to filter and drill down on events of interest. Nice.

Scaling: LiveJournal's Story and Secrets
by Brad Fitzpatrick, Silicon Mechanics

Brad started LiveJournal while in college back in 1999. Now over 10 million users he has a bit of experince in getting a website to scale. LiveJournal does blogging, forums, social-networking, and aggregator, so it has to keep track of a huge amount of data.
PDF of his slides

Practical Honeypots
by Michael DeMan, Open Access Network Services

Warren Sanders and I sat through this one (though I got up and left when the QandA started. Later Warren made a comment that "honeypot" in the title of the session was bait to attract us in. I agree.

To be fair, Michael did address the point "practical honeypots" by noting that as busy as we (sysadmins in general) are setting up a fake host is not something we have time for. He then went on to layout a specification for a well thoughtout distributed network management system.

What Went Wrong with my Disaster Recovery Plan
by Brian Martin, Martin Consulting Services

Brian defined a recovery plan as having three levels.

1. technical
   (host, lan, etc)
2. I/T plan
   (resources, transportation, facilities)
3. organization
   (office space, personnel)

Define what the threshold of events are the comprise a disaster. With the intent being to shrink the gray area between what clearly is and what is clearly not a disaster. Then the real biggie is to remove any penalty for the responsible person making a wrong decision regarding declaring a disaster or not.

Last three points that all plans should include

1. communications plans
2. travel plans
3. expense plans

I had a great time and want to thank Key Computer Consultants for sponsoring my travel, hotel, and food expenses. This year was bigger and better than last. I am looking forward to next year.

Thank you.