DevOpsDays Berlin 2016

Aah, Berlin again, DevOpsDays again - love it!

Here are some notes:
  • Day 1:
    • Looking back at 7 years of DevOpsDays, a keynote by Kris Buytaert.
      Kris showed a compact review from when Patrick Debois and he initiated the DevOps movement in 2009.
    • Infrastructure as a Microservice by Konrad Hosemann.
      It was a very good talk about their journey from a monolithic architecture run in a data center to microservices running in a public cloud. They implemented some infrastructure services and used several distinct accounts for different business units like search and data science. The advantages they recognized were benefits from external innovation, autonomous teams with a "you build it, you run it" attitude, less to no side effects, cost transparency incl. cost awareness. On the other hand, they were facing problems like clarifying the ownership of the infrastructure services, distributed updates across several accounts, avoiding fragmentation and gaining a unified view of their landscape. Konrad also showed interesting numbers: The Graphite setup consists of four 24 core machines with 72 GB RAM and 360 GB HDD each and handles 1.5 million metric updates per second; the ELK landscape utilizes ten 20 core machines with 128 GB RAM, 1.5 TB SSDs and 15 TB HDD each handling 150,000 write requests per second.
    • Communities of practice in large scale DevOps transformations by Martin Thalmann.
      A talk about the transformation towards a DevOps culture by establishing communities of practice in an organisation of about 4,300 IT people (give enough room/time to the members ... and provide beer and sausages). They also founded a local meetup group to open up the exchange with external contributors. Based on the number of questions after Martins talk, this seems to be a topic a lot of people deal and struggle with.
    • Talking to People: The forgotten DevOps tool by Peter Varhol.
      Instead of (over-)using electronic communication (e.g. email, IM) we should communicate more face to face, as we build trust and solve problems more quickly. Everyone, but especially management should lead by example, e.g. by putting the phone away during meetings. A regular "beer thirty" could help to establish communication and trust.
    • Ignites:
      • The 20 Shades of Grey by Michael Hüttermann.
        Michael showed 20 ways to make your project and DevOps efforts fail, e.g. DevOps teams as an anti-pattern.
      • Git and Github Tricks by Michal Bryxí.
        Things I didn't know before:
        - Ignore whitespace changes in diff: git diff -w -b
        - Better diffs by leveraging heuristics: git diff --compaction-heuristic
      • Come to Germany, Pepe! by Anthony Stanton.
        Anthony presented his development and learnings from a classical IT department employee to a DevOps.
      • DevOps? That's not my job! by Nathen Harvey.
        An entertaining ignite about the walls between dev and ops motivated by this blog post about serverless computing. Here is the longer version of that ignite as a full-length talk.
      • Felix Frank shortly introduced the motivation behind mgmt.
    • Hands-On Workshops (in parallel to Open Spaces):
    • I skipped the workshops, but joined the following Open Spaces - the resulting schedule is available here, and someone had the great idea to add further information directly to each card:
Yes, we're all certified now! ;-)

  • Day 2:
    • Information Overload and the Real Costs of Interruptions, a keynote by Hannah Foxwell.
      This was an awesome talk about how interruptions of any kind influence our effectiveness, the quality of our output and particular our health. This is a related blog post by her. Example quotes from her talk:
      - "It takes an average of 23 minutes for the average person to regain focus after being interrupted."
      - "Cognitive losses from multi-tasking are greater than the cognitive losses from smoking weed."
      Some ways to fight against interruptions:
      • Manage your notifications.
      • Do a brain dump (i.e. taking notes).
      • Sort tasks and plan to complete similar tasks at the same time (e.g. email).
      • Don't multi-task.
      • Have a routine that works for you.
      • Decide what NOT to do.
    • The Art of Supportability by TP Honey.
      Quote: "I don't work for the management, I work for the leaders in my team." TP talked about why it is important to produce code that is maintainable. Topics covered: clean code with good naming, good documentation incl. FAQs, meaningful error messages, ease of setup and use.
    • The Mathematics of Reliability by Avishai Ish-Shalom.
      This was a very interesting talk about calculating reliability e.g. taking dependencies into account. The repo contains the presentation and a Jupyter notebook.
For the rest of the day I missed the talks, ignites, open spaces etc., as I met a lot of friends and former colleagues exchanging lots of news, gossip and "war stories".

No comments:

Post a Comment