OSCON 2009 – Day 1 (liveblog)
July 22, 2009 Leave a comment
I’m trying a new format for this conference. Instead of live blogging one session per post, I’m going to aggregate the entire day into a single live post. Feedback on the format is welcome.
* mobile phones are now really a collection of sensors that cooperate with cloud data services.
* Google will be able to do speech recognition better than anyone else, because they know what people are searching for, and therefore more likely to say,
* As we rely more and more on data stores that live in the cloud, we need to be concerned with whether that information should be centralized or federated.
* We’ve moved to a vending machine government
Dirk Hohndel on netbooks
* Linux on netbooks need to be:
fast – specifically when booting
graphical – move part of graphics subsystem into the kernel, clutter, Intel gem, non-root X
connected – all connectivity between different devices needs to be cohesive
* pronounced: Butter-eff-ess
* Designed to provide big new features not available in other Linux file systems; snapshots, fault tolerance and simple administration were high on the list
* Performance was not, and is not, the number one goal
* Some features
– Multiple devices per filesystem; shared pooled storage
– checksums; trust nothing
– snapshots; quick, cheap and writable
– directories are indexed for speed; indexed for two different cases
– small files are packed, as resierfs did
Virtualize vs Containerize
* Virtualization is great for system consolidation and protection, but you pay a heavy performance penalty
* Most popular containerization solution is OpenVZ.
* Are limited to running hosts/guests with the same kernel with containerization
* The two seem to be converging as they both evolve
Cobbler and Puppet
* Automating System Builds and Maintenance
* Always remember, “temporary solutions become permanent”
* Cobbler is basically a front end to kickstart (and therefore somewhat RHT-specific) with some additional functionality – https://fedorahosted.org/cobbler/
* Puppet is a declarative language for expressing system configuration, a client and server for distributing it and a library for realizing the configuration. It’s written in ruby. http://reductivelabs.com/products/puppet/
Ten Usability Epiphanies for Your Open Source Web-app
* Hot mouseover (affordance), using a button (for primary action) vs. link (for secondary action), use loading spinners, ensure drag & drop functionality is explicit, avoid iconitis, avoid too many steps (especially for common tasks), choose your choices (follow HIG when possible), language (be descriptive, succinct, necessary and edifying), implement undo functionality and implement keyboard-based functionality (including keybord shortcuts).
Linux Filesystem Performance for Databases
* ext2, ext3, jfs, xfs, reiserfs and attempted ext4
* single disk, raid 0, raid 1, raid 5 and raid 10
* did NO filesystem tuning at all, no mount options and all used all OS system defaults
* adding a second disk to raid 0 gets you almost no performance gain, but > 2 disks does
* raid 5 performs better than most people think
* software raid often performed better from an i/o throughput perspective, but at the cost of added CPU usage
* LVM overhead is not quite as bad as many people think
* turning off atime wasn’t as big of a win as expected
Hacking the Open Government
* There is currently a renaissance in available data and content from the federal government as well as some state governments. Everyday it seems that some new web site or service makes publicly available over the internet what previously was only accessible on paper or other non-digital format. What does this mean for the hacker, the agitator, the political scientist, and the average internet user? How do we ensure that
1. the data is in a usable form and
2. that it is put to good use and
3. that citizens are continually empowered to raise the bar a little higher?
The first full day of OSCON was once again highly interesting and engaging. Off to the Expo Hall Reception and LinuxFund party. I’ll be live blogging tomorrow as well; stay tuned.