ScaleConf: Review & Closing thoughts

This week I took two days off from my day job at Symantec as a Product Marketer for Information Protection and went to sit in a room with over 200 developers in Cape Town at ScaleConf. I was the only marketing guy in the room (we checked, trust me, no-one else whooped with me, or were too ashamed) but thankfully my random past which included many years as a developer and keeping in my hand in a few projects since meant there was only a few times I got lost!

Before I get into reviewing the conference and my personal thoughts, my key takeaway for my employers is that somebody needs to be at events like this representing Symantec. Symantec probably has the most relevant “vendor” portfolio of anyone to this topic and not everyone in the room was a bootstrapping start up using open-source. Nobody mentioned Symantec’s private/public scaling and availability technologies once (except for me, over beers, 1:1).

The conference itself was well run, up to the event and the event itself. Any organisational glitches that I experienced (my double-ticketing, bad WiFi performance, speakers getting ill) were handled quickly and neatly (extra ticket cancelled, bandwidth increased, moved to a panel instead). The coffee may not have rocked anybody’s world, but they had Rooibos tea! Logistics were handled by LessFuss so we should probably have expected nothing less, but there’s a more than small difference between keeping Capetonians lives in order and 200+ geeks happy for two days (or is there Jen?).

Content on the other hand was a mixed bag with some superb talks and some downright dreary talks! There were nine speaker sessions and one panel at the end (to cover for the unfortunate illness of Coda Hale) with each session lasting forty-five minutes. This is a perfect length for a talk. Thirty-five minutes of content, ten of interaction. It’s perfect in two ways. Firstly, because if the speaker is great, they can get a lot of information and inspiration done in that time and leads you toward how you can learn more from them or online. Secondly, because if they suck, you can easily entertain yourself on Twitter in the backchannel or answer some email and not feel like the whole thing is a waste of money.

Before I get into the session reviews let me say these three things so I don’t get struck off as a rude ignoramus forever:

  1. Organising conference content is really, really hard. I help do it at least three times a year for internal and external events at Symantec. People suck. At commitment, sticking to deadlines, doing what they agreed, the list goes on and on.
  2. Speaking amongst peers is really, really hard. Especially in the Stack Overflow/Hacker News era when everyone’s done a quick search after hearing something new and is an “expert” suddenly.
  3. Every single speaker at this conference had something good to say. There’s a bible verse that roughly says “Don’t hide your light under a basket but put it on a lamp stand for all to see”. Some people needed help with their basket and lamp stand skills.

Lastly, one more caveat (do I sound nervous of a backlash much?). I said on Twitter during one session of the conference:

95% of the room are trying not to kill themselves, the other 5% think this is the best talk of the day. #iminthe95 #scaleconf

Though I may be exaggerating with my 95% claim, I think from the Twitter traffic during the “not so great” sessions I was in the majority in most of my opinions. In a non-tracked conference, I think you have to try and satisfy majority requirements. With that in mind here’s my own personal, very high-level view of the sessions and any slide links I could find.

Session 1: Jonathan Hitchcock - Clearly I Have Made Some Bad Decisions
Great talk after Jonathan (one of the conference organisers) got over initial nerves, had some anecdotes to get some points very clearly across. Top points for me from [Slides]:
  • Know what is normal (don’t start monitoring when you think you have a problem, too late)
  • Server config is code (make consistent repeatable deployment easy)
  • Continuous deployment can give devs a sense of lots of small achievements

Session 2: Craig Raw - Webscaling Tips
Competent, confident speaker but for me delivered little that couldn’t be found on a blog post about 101 Scaling. Maybe not a bad choice putting Craig from Quirk in for level-setting early on in the conf for newbies. Got some grief on Twitter for not even mentioning Python during his comparison of “safe/unsafe” languages for scale! [Slides]

Session 3: Zach Holman - Scaling GitHub
Easily the slickest presentation of the two days by Zach and not just pretty content but delivered well and with the right amount of detail. Top points for me from [Slides]:

  • Happy employees make Productive employees (and visa-versa)
  • Everybody (not just managers) should worry about co-workers happiness
  • Look for seemingly separate datasources to graph for interesting correlations (e.g. “Twitter mentions” against “production deploys”)
  • When using Git for source control, Master should be ready to go at any time, branch for new features and merge back to Master

Session 4: Miles Ward - Scaling using Amazon Web Services (AWS)
This was the first of the sessions where the quality of content was hidden by the delivery. Subsequent appearances and comments by Miles during the panel showed clarity of thought and humor. AWS is amazing, getting better all the time, but I knew that before. Whizzing through many reference architectures, mumbling into the mic is not a great use of 35 mins. What I would have like to see here is examples of bad scaling decisions when using AWS contrasted against the good. [Slides]

Session 5: Bryn Divey - Nimbula’s bIC: lessons learned and challenges faced
Firstly, I couldn’t shake the feeling that Bryn reminded me of Captain Jack Sparrow! Well delivered talk that touched on how they overcame some of the challenges in building a solution for developing private clouds. I felt like I knew a bunch of this stuff because I work at Symantec (we do it too) but key lesson for me in design was the criticality of tracking and documenting service dependancies is critical for multi-service-multi-server systems. One improvement could have been a bit more structure on being explicit on the lessons learnt.

Session 6: Deon Erasmus - Highly available infrastructure on the cheap
OK … last session of the day, so where do I start with this one from Deon. Erm… I didn’t make any notes on this one. But I spent a lot of time on Twitter.

Session 7: Ashley Peter - Scaling a mobile social network
This was an excellent start to day two from Ashley, very practical and very inspiring given 2go’s massive growth (16m registered users) and low numbers of staff (3!). One thing I picked up about scaling which I think I’ve paraphrased was “there’s nothing to do, until there’s something to do, then you need to do it really well, really quickly”! The structure of this one broken down into a focus on how they themselves worked through a combination of Vertical scaling, Parallelism and Horizontal scaling was great. [Slides]

Session 8: Simon de Haan - Learning to Fail
Simon is from Praekelt which is a fascinating dual commercial/foundation organisation. Their focus on solutions for Africa and building Open Source tools to enable that is both well engineered and worthwhile. I took lots of notes on this one, probably worth a post on its own and would like to meet up with the team behind Praekelt again to dive in deeper on some of their organisations set-up. My favourite tip though was “Avoid One Hammer To Rule Them All” … in other words, don’t use your favourite tool (framework, language, database, etc.) inappropriately! [Slides]

Session 9: Mark Phillips - Building Healthy Distributed Systems
Mark is from Basho, makers of the NoSQL database Riak and split his talk into talking about the characteristics of a distributed company, community and system. Great talk delivered with a sense of humor that I like and the right amount of clear advice which was something that some of the talks missed out on highlighting. [Slides]

Session 10: Wesley Lynch - Software Architecture and building E-Commerce Websites
Wesley certainly had a tough job ending the speakers content. He presented a case study of the job Realm Digital did (and continues to do from the looks of things) in bringing Exclusive Books online operations to life. Unfortunately it seems they had to bring order to a ball of “kak” and its taken its toll on Wesley’s positivity and it felt more like a rant than set of learnings with guidance! Full marks for attempting to stoke more audience participation and to the crowd for trying to draw some positives out (I tried… and got a nice t-shirt for my efforts!).

Concluding Thoughts
The conference closed with a speaker panel which had the right amount of wrapping up questions, audience participation and “in jokes” referencing the rest of the two days.

Overall, this was a great use of two days “off work” for me personally. It was an impressive setup given it was the first time the team had attempted a conference of this size and subject. Well done to Jonathan Hitchcock (@vhata), Duncan Phillips (@nuknad) and the LessFuss crew (Jen, Bradley and Marije).

Lastly, I tried to build a Twitter list of attendees by monitoring for mentions which you’ll find here.

Looking forward to the next event!

 

7 thoughts on “ScaleConf: Review & Closing thoughts

  1. Hey. Great writeup. I think your descriptions were spot on. Even the boring talks were ok, I think the twitter negativity was a bit of mob mentality sometimes (which I admit to getting caught up in). Hopefully there are no hard feelings.

    We need more things like this to get developers in CT together: conferences, talks, drinks, meetups. Maybe you should pull out a “lessons learned @ symantec” talk and have some people over for beers? :)

    • Thanks … I agree on the devs together. There’s a spotty amount of activity, mostly in language-based silo’s. We could do with a BarCamp style gathering…

      I’m not sure Symantec can host in CT because of the security status of our building. Worth asking though so I’ll look into it.

      • GeekDinner was a fairly successful silo-free networking event. Jonathan was heavily involved in keeping that going (Joe Botha started it, Jonathan Endersby and others helped keep it going as well), but it petered out when he took a break to let others run it.

        Jonathan and I organised a Barcamp-style conference a few years ago as well. A two day conference at AIMS for general tech stuff.

        I think Jonathan has proven that there’s a market for full-on conferences as well (and Rubyfuza has done the same for language-based conferences).

        The hard part of building these sorts of things is overcoming inertia, and maintaining momentum. Financial gain is one way to motivate for momentum – I suspect that if LessFuss had existed earlier, it could have been used to keep GeekDinner running. In many ways, it really is as hard as finding a venue and setting a date. For a Barcamp-style conference, it is the same, with finding some food added on. Getting a sponsor to pay for LessFuss’s services to do these admin things that geeks generally dislike would be pretty easy.

  2. Pingback: 2012 Resolutions: 1 Month In | Smarter Messaging [Mike Jones, SA]

  3. Pingback: ScaleConf – Report Back » GSDH Advertising Agency | Design Illustration Animation Corporate Design Flash Programmer Logo Design

Leave a Reply

Your email address will not be published. Required fields are marked *

*

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>