Skip to main content

OSCON Thursday

Posted by haroldcarr on July 29, 2011 at 1:29 PM PDT

My OSCON Thursday, 07/28/2011

2 10:40am - Big Data; Small Problems (And a Tale of Two Sequels)

  • Tools and Techniques
  • Tags: mongodb, postgresql, clustrix, sql, newsql, hadoop, pgsql, couchbase, mariadb, riak, mysql, bigcouch, nosql, cloudant, couchdb, drizzle, cassandra, voltdb, database
  • Tim Anglade (Cloudant)

attendance: 65

Tim: and

… talk was too general/high level - no real info …

mongodb, couchdb, riak, cassandra, neo4j, bigcouch, redis?, hadoop?

… by the time he started to get into real info I had tuned out,
installing prerequsites for instead …

3 11:30am - Building a Machine Learning Lab that Scales in Your Garage

  • Citizen Science
  • Tags: mahout, citizen_data_science, hadoop, machine_learning
  • Vin Sharma (Intel)

attendance: 50

… speaker did not show up …

4 11:30am - Wheeler

attendance: 40

… show I went to this - but either I missed to much context (since I
missed the first 15 minutes) or …

I struck out this morning.

5 1:40pm - Utility and Automation: Low Overhead Operations with Amazon and Puppet

attendance: 35

  • man hours; spares; growth; routing/continents; regulation
  • shared infrastructure -> shared resources (multitenant)
  • instances are not servers
  • data persistence is hard


  • postgres + memcache


  • EBS is unstable
    • do not put your DB on it; nor your root volumes; DO use for backups
  • AWS fails all the time; your apps must be FT


  • declarative DSL for config mgmt
  • other options: Chef, Puppet, CFEnginer; AutomateIt, …
  • more work to do upfront, but reusable and easier to figure out than
    debugging config after the fact


  • inventory, sort, filter, execute on multiple instances
  • when you have 100s of nodes, nice to say
    • restart some service on all my nodes that have this other service
    • don't do it all at once and cause a stampede
  • mcollective

6 2:30pm - Hacking Yourself with Open Source Software

attendance: 40

behavior change is difficult

playing games is a way to get people to do things that won't normally
do - or would do, but procrastinate

smoking, over-eating/drinking, drugs, …

games as stealth-health

socially-oriented : pacting, contracts:

book: Carrots and Sticks

mob zombies, figure running, real-life pacman

7 4:10pm - Refactoring SQL

  • Programming
  • Tags: sql, performance, refactoring, performance_tuning, database
  • Jeremiah Peschka (Brent Ozar PLF)

attendance: 50

8 5:00pm - How OAuth 2.0 Influenced Architecture, Design, and Policy

  • Open Data
  • Tags: api, design, oauth
  • Eran Hammer-Lahav (Yahoo!)

attendance: 20

invite code: hueniverse


  • collaborative list-making tool
  • friends and family
  • zero workflow
  • web and mobile
  • live updates
  • platform API

OAuth 2.0

  • spec: last call and ready for impl
  • use 2.0 now - do not use 1.0
  • redesigned for scale
    • use access tokens without client auth
    • refresh tokens
  • decouples authentication for authorization
  • everything extensible
  • crypto optional
  • not backwards compatible
client                       resource

resource                     authorization
owner                        server
                             (i.e., OAuth server - TLS required)

Using Access Tokens

Authorization: MAC id="dfdfdf",
  • 1 resource owner inits flow to client
  • 2 redirect to auth server
  • 3 give password - returns auth code
  • 4 auth code given to client
  • 5 client authenticates with auth server
  • 6 client gets stuff from resource server

sled architecture

  • multiple scopes, client permissions
  • OAuth 2.0 extension grant types
  • everything is client API
  • 3rd party login (facebook, twitter, yahoo!)
  • link multiple accounts
  • no local passwords
  • user -> web server (static files, login, acct mgmt)
  • user -> api server (data calls and streaming updates)
  • web <-> api server (acct validation)

client implications

  • client-side rendering
    • multiple round-trips to display initial view
    • redirection based error handling
  • cross-domain requests
    • CORS overhead
    • limited browser support
  • one-time interaction with web server
    • custom mechanism for refreshing client code


  • config: version; access token route; redirection URI
  • zero deploy client testing: run locally and hosted
  • OSS: clone, config, run


  • visually-based load prioritization
  • roaming load indicator
  • preload data on hover
  • apply local changes; revert with indicator on error


  • default: user data read/write
  • elevated: login, signup, reminder, tos, quit


  • minimum required for each API call

client storage API (to store on server)

  • preferences; last state; scope; name/value

9 7:00pm - The Locker Project & TeleHash Hackathon!

Got it working - but not easy


node lockerd.js

allow popups to localhost

google APIs -> API Access

http://localhost:8042/map Locker/Me/facebook-2/me.json -> providers add: ="photo/facebook"=

look at hello photos as template for transforming collection data to RDF

10 Other sessions I did not attend but find interesting

10.1 The Right Tool for the Job: Using Hadoop with a Purpose-Built Analytics Platform

  • Products & Services
  • Mingsheng Hong (Vertica, An HP Company)

10.2 Plaid: Programming with Typestate and Permissions

  • Emerging Languages
  • Tags: programming_language, types, plaid, permissions, typestate
  • Karl Naden (Carnegie Mellon University), Jonathan Aldrich (Carnegie Mellon University)

10.3 Not Only SQL: Exploring Alternative Data Stores

  • Programming
  • Tags: mongodb, databases, riak, neo4j, nosql, couchdb, cassandra
  • Roger Bodamer (10gen), Peter Neubauer (Neo Technology), Matt Pfeil
    (DataStax), Tim Anglade (Cloudant), Antony Falco (Basho

10.4 Bridging Unstructured and Structured Data with Apache Hadoop and Vertica

  • Products & Services
  • Steve Watt (Hewlett-Packard), Glenn Gebhart (Hewlett-Packard)

10.5 Open Source Identity With No Lock In

  • Business
  • Tags: saml, startup, business_model, identity, ldap
  • Simon Phipps (ForgeRock), Allan Foster (ForgeRock AS), Lasse
    Andresen (ForgeRock)

10.6 Higher Order Javascript

  • Programming
  • Tags: programming, refactoring, javascript, higher_order_programming
  • Piers Cawley (BBC)

10.7 OpenStack + Ceph

  • Products & Services
  • Jonathan Bryce (The Rackspace Cloud), Ben Cherian (DreamHost)

10.8 CUBRID HA - Guaranteed Way to Never-Die Web Services

  • Products & Services
  • Esen Sagynov (NHN Business Platform)

10.9 Leveraging Commercial Graph DB Technologies in Open Source and Polyglot Application Environments

  • Products & Services
  • Brian Clark (Objectivity)

10.10 Magpie

  • Emerging Languages
  • Robert Nystrom (Google)

10.11 Just What Defines A Web App?

  • Products & Services
  • Oren Teich (Heroku)

10.12 Ensemble And The Art Of Service Orchestration

  • Operations & System Administration
  • Tags: upstart_linux_server
  • Clint Byrum (Canonical)

10.13 The Kotlin Programming Language

  • Emerging Languages
  • Andrey Breslav (JetBrains)

10.14 Collaboration - An Emerging Trend in the Healthcare Open Source Model

  • Healthcare
  • Tags: healthcare, hospital_ehr, open_source_for_healthcare, vendor_collaboration
  • Vickie Hoffman (Roberts-Hoffman Software), Thomas Jones (Tolven), Brady Mathis (Roberts-Hoffman Software, Inc.)

10.15 Visualizing Economic Data Using Perl and HTML5's Canvas

  • Perl
  • Tags: perl, economics, data, government, html5, javascript, web_development
  • A.Sinan Unur (Unur)

10.16 Gosu And The Open Type System

  • Emerging Languages, Java & JVM
  • Carson Gross (Guidewire Software)

10.17 Monitoring at Scale

  • Moderated by: Jeremy Brinkley

10.18 ActiveMQ Clustering and High Availability

Moderated by: Dhiraj Bokde

Related Topics >>