QCon London 2014 - Friday
March 9, 2014
Keynote: The World after Cloud Computing & Big Data - Gunter Dueck
Upgrade your future or be industrialized. T shaped influence. The autistic spectrum. Computer scientists are like cats. Dogs have meetings cats solve problems. Use your other intelligences.
- Books: Who Moved My Cheese: An Amazing Way to Deal with Change in Your Work and in Your Life - Lean Brain Management: More Success and Efficiency by Saving Intelligence
Real-Time Systems at Twitter - @Brian Degenhardt
Interesting presentation on the architecture at Twitter. Another presentation that said "we build a monolithic rails app, but it didn't scale, so we re-engineered." 5000 tweets per second. 300k timeline requests per second.
How you divide into pieces. Incrementally implement SOA. Separate semantics from execution. Use stats to monitor behaviour.
Talked about Future objects - a composable thing that might or might not have fetched/calculated a result yet.
- Tech: redis - twitter-server - finagle - Scala - zipkin (graphical tracing of architecture calls. Always on. Distributed tracing) - thrift - Iago load test
What can Hadoop do for you? - @Eva Andreasson
The speaker took questions after they'd described the various technologies involved in the hadoop ecosystem, but before covering case studies of what Hadoop can do for me. Disappointing for me, probably worth getting hold of the slides.
Can use sqoop to ETL into Hadoop then process, put the results back in relational database.
Suggested building an Enterprise Data Hub.
- Tech: Hadoop - YARN (scheduling) - Flume (ingest data) - Sqoop (ingest data) - hive (SQL to map reduce) - Pig (same has hive, but for programmer rather than DBA) - HBase - Hue (ui) - Oozie (workflow scheduler) - Impala (real time query) - Solr (search) - Spark (Map reduce on speed)
Lean Under Pressure - @Glen Ford
Lego themed slides - excellent! Chief Architect of ZeeBox. Talked about leadership vs management.
You build it you run it. No handoff to ops. Garden architecture. Chaordic system - can't always be predictable.
Suggested lunchtime tech talks, and guilds for grouping together people interested in a specific technology.
-
Tech: Scala
-
Links: zeebox - Lean Software Development
A Call for Sanity in NoSQL - @Nathan Marz
Started by describing the "Doofus programmer" then going on to discuss Immutable Databases. Always insert, no update. Use Map Reduce to produce consolidated views of data. Can always add in recent records onto results from MapReduce to get up to date details.
Covered equivs - finding out who a user is after they've logged in yet still tagging their non-logged in requests with the same user.
-
Links: Lamda Architecture - HyperLogLog
How Shutl delivers even faster using the Neo4J - @Volker Pacher & @Sam Phillips
Neo4J is a graph database that can be run standalone or embedded in a JVM. Shutl designed to be PayPal for awesome delivery, now bought by eBay.
Graph has nodes and vertices. Bridges of Konigsburg is the famous example. You can also have a directed graph.
-
Tech: neo4j - Cypher Query Language - Geoff
-
Links: shutl
-
Books: Graph Databases
How Elasticsearch Powers the Guardian's Newsroom - @Graham Tackley & @Shay Banon
The Guardian uses an in house analytics tool called ophan, which is based on elasticsearch. 5-6 million unique page views per day. It takes five seconds to get a page view onto a graph. First version used zeromq and ran on a desktop machine after a hack day.
-
Tech: elasticsearch - Amazon Web Services - zeromq - status-app - d3.js - Rickshaw
-
Links: The Guardian