Author: simon

Making great conference video on a budget – Ryan Verner and Ben Hutchens

• Guys from team doing lca2011 conference video
• BOF on this later in conference
• 2004 used tapes, hard to scale
• Volunteer AV very different from using pros, skillset limited
• common chellenges from conferences, not LCA specific
• each component relatively simple, very easy to underestimate, many components, many rooms
• incomplete AV workflow, focus primary on tech
• Volunteers – limited manpower, expeience, training
• unforseen tech issues – Small varience in hardware
• workflow – automate as much as possible
• workflow – manage schedule , feedback loop (recording sheets) , rapid post editing, distribute tasks, automate transcoding/.uploading
• mistakes – LCA 400 hours of video, too much to do ANY manual editing, combining vga,main,audio feeds
• vga capture – twinpact 100 – $600
• Basic mixer – mike, usb sound – $300
• Firewire camera – $300
• Laptops running linux, DVDSwitch software mixes into single DV file
• Audio quality – often neglected
• Get a VGA capture device
• 100% test befoe conference, impossible to solve once things start
• Train volenteers
• Managment important, clear roles, delegation of specific tasks
• Clear Documentation
• Examples http://pythin.mirocommunity.org
• http://videokollektiv.org – example of best practices
• DVswitch software – video mixing, recording, streaming
• Designed for free software confs, limited budget
• Ofter used, sometimes without streaming

• Like “test driven development”
• Initially unit test focussed
• business doesn’t care about underlying implementation, just the functional results
• “executable specification” – written in spoken language
• “infrastructure as code”
• code without tests is “bad”
• Taking BDD and adapting to infrastructure development
• Tools – cucumber – write specs, execute/test specs
• cucumber-nagios
• DEMO of cucumber and cucumber-nagios
• example – continuous server builds as you update you config manager
• migration to config management. – Use BDI to test existing, test CM env with same tests to make sure it replicates
• Monitoring System : notify -> test -> repeat
• Current monitoring systems miss many things
• cucumber provides a common specification format that dev and ops can share
• removes duplication of tests
• Libraries being built of commons test (some already in cucumber nagios)

Geoff Huston

• New technologies from the 70s – Unix and Packet switching (TCP/IP)
• Open technologies – anyone can implement
• TCP/IP not better than competing technologies but it’s openness greatly helped it win
• One thing to “be open” , another to “stay open”
• Useful technologies are rarely static
• Technology evolves, uses change (eg growth of wireless), exploitation models change
• Challenges – net neutrality, next generation networks, mobility and mobil service evolution, Triple/Quad play schemes
• The really important thing is “We are running out of addresses”
• 190 m addresses given out in 2009, 248m in 2010
• 300million new things on the network
• 9 million new addresses just in Australia
• 7 /8s left, rate of 1/month
• plan that IPv6 transition would happen before ipv4 ran out
• Only 0.3% hitting google IPv6
• IPv4 will run out during 2011/2012
• Need to transition to ipv6 in 200 days
• Won’t happen, have to muddle thing with ipv4
• NATs are an externalized problem
• ISP NATs, multi-level, within ISP network
• aperture through through which the Internet can be seen and used. Reduced port space
• transition to ipv6. Dual stack requires everybody to have a ipv4 address
• If you run 6to4 15% of connections don’t connect.
• Transition could take 5-40 years
• Making ipv4 addresses last longer, they will cost
• TCP/IP is the network monoculture
• Will openness be lost in the transition?
• Telcos being asking to make big investment in ipv6. No really in their interests to have an open network.
• Similar for large Internet companies like google and amazon
• Delays help the incumbents, open network infrastructure is at risk
• Need to figure out how to motivate big companies to goto ipv6 and open infrastructure

Re-imagining the Internet – Vint Cerf

• 768 million machines and 1966 million users in mid-2010
• Already more chinese on the Internet than americans
• No particular applications in Internet design
• non-national IP address structure
• Open standards, No IP attached to TCP/IP
• Anybody can build a piece of the Internet and connect to it
• Recent developments: ipv6, int domains names, dnssec, rpki, sensor nets, smart grid, mobile devices
• Two factor authentication really needed by everybody
• Security problems on OS, Browser, Interpreter boundaries
• We privacy laws, lax user behavior
• Invasive devices
• Cloud to cloud missing, data between them has to go via user
• List of unsolved, research problems
• Buffer bloat problem might have to mean reduction in buffers

Finding you feet without losing your head – Alison Young

• Tech writer at redhat
• starting a new job
• make sure people know your name, spelling and what you prefer to be called
• drop nickname you don’t like when switch jobs
• find out dress policy
• hopefully have a buddy to get you going
• preferred communication ways at company ( talk vs email vs IRC vs skype ), need to allign with this
• management style. hands off vs micromanagement
• work from home. at Redhat be qualified
• transitions at workplace (dinners, cake days)
• breaks, present-ism, how intently are you expected to work.

We are here, have always been here – Donna Benjamin

• 5 minute history of feminism
• $7000 to digitise “The Dawn” , fund-raising effort
• In past women more common in computing, cheaper to hire skilled workers
• Less common today since women get less computers when young
• examples of Women in Australian computing she uncovered
• “we are not fucking unicorns”
• Challenge to find women for Ada Lovelace day

Roleplaying Session – Val Aurora

• Roleplaying sessions of people doing sexist activity and options for replying

Arjin Lentz – creating the business you want

• – ex mysql
• – left when 500 people
• – growth in company revenue doesn’t always mean good elsewhere in business
• – remote services for mysql
• – no emergencies – time with daughter, sanity
• – started pre-GFC, prices reasonable, published, stuck to.
• – no emergencies = no worries after hours, do oncall infrastructure
• = Pool of people who won’t work weekends
• – biz processes = some cases no *real* reason why it’s done that way, but hard to change if other
• things depend on it.
• – hard to disrupt yourself
• – no borrowing, external funding etc.- big affect on how run. See rules
• – big growth , floating, being bought doesn’t always benifit customers
• – lives below the big companies, keep pricing below them
• – total value of biz-space is too small for “china” to enter
• – value curve, invest in different balance of value for your product than your competitors, feature set
• – nintendo wii, amazon
• – list of cool books
• – bigger clients require different sort of company to service

13 years of LWN – Johnathon Corbet

• – most audience lwn subscribers
• – establish 1998, 3 emplyees, x000 subscrbers, >100 company subs
• – Programmed Cray 1 – #3
• – drifted up to mid-level management in 96/97
• – little correlation between work and reward
• – Starting off “Linux Consulting” company ( eklektix.com ), start website to show how smart we are – not many $
• – we’ll do linux support – became linux support partner – program went away
• – linux training company – crowded market – didn’t work out
• – Maybe online news company
• – lesson – business skills matter
• – lesson where money coming from – pay attention to what customers want
• – be ready and will to change plans
• – acquired by tucows. went for mainly cash
• – seemed like good people, money over pure stock.
• – after dotcom crash, tucows handed back.
• – advertising revenue big drop from pre-crash
• – business very cyclical
• – real customers are the advertisers. Other sites did articles for advertising spend
• – Blocked microsoft.com, hard to block all the other variations for the name
• – other ads for dodgy products, soft core porn, political ads
• – ruins customer experience, javascript, flash, popups
• – hard audience to advertise to
• – donations didn’t work
• – July 2002 put up message that calling it quits
• – $35,000 in tip jar over 1 week. “why don’t you try subscriptions”
• – Nobody pays, Linux users less likely to pay.
• – listen to customers, especially when they are offering money
• – credit card company, reverted donation surge
• – credit cards,; extra feels for: discount rate, transaction fees, “international charges”, affinity charges, some arbitrarily
• – banks nervous about extending long to credit to merchants
• – chargebacks. customer always wins. 5 chargebacks over 10 years
• – credit card security, big dangers, huge potential downsides, pci compliance
• – credit card lessons, keep money from somewhere else
• – alternatives to Cc – 5% of stream, works okay, cost 4%
• – Checks – pain to deal with internationally
• – Corp PO cycle – big pain to deal with, Be patient
• – other services, amazon. Haven’t investigated heavily
• – lesson – have 6 months in the bank
• – where are we now
• – subscribers get access to feature content
• – free after 1-2 weeks
• – ability to disable advertising
• – other features
• – Basic cost $7/month , higher and lower cost alternatives, group subscriptions
• – aligns interests with our readers
• – people want to support us
• – subscriptions are a business expense for most people
• – non-cyclical
• – 2008
• – subs steady
• – adverting dies
• – many competitors die, freelancers writers more avaibale
• – amazon affiliate , not good results and then amazon pulled plug on all of Colorado
• – lots of revenue sources good. biggest business is 5% of revenue
• – why doesn’t it work. audience is too small
• – people don’t want to pay
• – we are terrible at selling
• – pricing is really hard, raised prices by 40%, minimal loss of subscribers
• – “design the business as a functioning system” – hard to do with periodical

Lindsay Holmwood – Devops

Consistency

• config managemnt – puppet
• puppet workflow – write->apply->debug
• testing via snapshots, apply,test, keep/revert
• puppet roles

Repeatability

• capistrano – cap-sub for config management
• railess-deploy (capistrano, externsion)

Visability

• Collectd
• git hub newsfeed
• mk-query-digest (slow sql queries)

Lots of ideas in his talk, far to quickly said for me to type them in a all in.

Deltaclound – David Jorm & Stephen Gordon

• Redhat project
• deploy workloads to multiple clouds (some, internal, some external, multiple vendors )
• free and open standards
• REST API that abstracts other cloud APIs. Drivers for different clouds.
• Support EC2 and Cloundfiles, others being worked on
• Build images, store, push to cloud provider and then deploy instances

Cloud Computing in Govt – Pia Waugh

• Cloud vendors vs govt people – Vendor gtee’d no personal data in their cloud
• SOA buzzwords drop straight in Cloud Buzzwords
• http://soafacts.com/
• Govt + cloud = $$$$ for vendors
• Lots of vendor hype, have to ask the hard questions
• Jurisdiction – govt data can’t go overseas
• standards – avoid vendors, technology lock-in
• data – reliability, what if vendor goes out of business
• AGIMO – official govt cloud strategy doc
• private vs public clouds
• Govt needs good advice
• To influence Govt – “be helpful”
• More consultation events planned in near future
• AGIMO actually using cloud for some things

Just a reminder that the Call for Papers for the Sysadmin Miniconf at Linux.conf.au 2011 will close on Friday the 22nd of October (just 2 days away).

We have accepted 7 proposals so far and have a few more pending but we still have space.