Month: October 2016

Another Reactive Summit 2016 talk. Speaker was Steve Moore – IBM story strategist. My comments at the end. (When writing about heros I used “she” as pronoun, if you are sensitive to this use sed s/she/she/he/g).

Thinking differently helps you gaining new perspectives on your work. That’s the work for story strategist and this is the idea behind this talk. Finding analogies between reactive systems and movies.

Stories behave like reactive systems you can see three analogies:

1. hero as scalability,
2. plot as resilience,
3. genre as elasticity.

Scalability : heroes are under constant pressure to scale. Scaling is usually achieved via some resource. The best resource a hero has at her disposal is a new understanding of the world.

There are three form of escaping in the scaling up of the hero.

Escape the skin. The hero is trying to escape the fake reality she’s living to reach the real and more complex reality behind. Paranoia and questioning foundation of the system are the mean to achieve this escape. This can be summed by the question: “Are you being paranoid enough”?

Escape stoicisms. The hero lives in a simple world that avoids emotional entanglement. The hero takes more responsibility that she think she can bear. Scalability and generosity go hand in hand. Question: “Are you being generous enough?”

Escape the story. The hero lives in a world she wants to escape. This is different from the deception of the escape the skin (first form). The question is “are you being patient enough?” Sometime you need to take the time to observe and understand.

plotasresilience

The plot directs the hero where is needed by the story. Stories are all about failures – the plot is just a sequence of failures that get the hero further and further away from his goal.

Heroes never ever give up. Likeability is not required by heros.

3 ways the hero fails:

Lure of the invisible. This is a kind of accumulation of something that is somewhat hiding. Can be a log file that is increasing without anyone noticing, or some peril that grows unnoticed over time. “What is the system accumulating?” This is not an issue until it is. This way of failure includes also non technical accumulation such as management attention, customer complaints…

Lure of inclination. The hero would like something, but she has to put away that to pursue his goal. Following inclinations may get the hero in hot water, but in the end brings benefits.  “Can you make the system fragile?” [NdM sorry, it is likely I missed something in the translation]

Lure of independence. The story has a lot of moving parts apparently in an unrelated way. Using micro services the system may be better managed and written by smaller teams. “Could coordination be fun?”

Coordination among the teams is needed.

Genre is elasticity.

Elasticity means that response does not change when workload changes or accidents happen. What defines the genre is a set of rules. Genres are categories of antagonists. An antagonist puts strain on the hero.

Three different kinds of genres:

Human v. Human. The analogy on system is threats posed by human hostile actions. “Can you change the game?”

Human v. Nature. The natural world is challenging the hero. Natural antagonists are simple, unbeatable and unchangeable. These stories are about survival. If you live, you deserve to live. “do you deserve to live?” If the system were in danger do these people would do something to protect it?

Human vs self. Hero is also the antagonist. The hero doesn’t want to push away something that is abundant in his life. This kind of abundance forces the hero to questioning the reality “can you endure abundance?” We need to transform ourselves and the system at the same time.

This is the only way the system has to survive.

This talk was very interesting even if not very practical or even theorical. I like the idea that adding word to vocabulary and concepts to the programmer resources may help to devise better system or to enhance the understanding of the system. It makes food for lateral thinking (lateral food?).

One of the first person we met at the luminaries barbecue the day we arrived, had the following job title “Chief Storyteller”. Maybe this is a sign of times.

Here I am, warped through jet leg in the sunny Texas. Austin. The reason is to attend the Reactive summit 2016. Two days crammed with interesting speeches by luminaries of the reactive community. I’ll try my best to provide a brief summary of each talk I will attend, reportin faithfully what the speaker said and adding my thoughts. If you know about reactive systems and you read something unreasonable it is likely I missed something in the process.

This talk is by Edward Hwu – mesosphere.

(I’ll leave out the quote about Justin Bibier being the reason for having reactive systems… in part because I missed something).

If you compare the number of Internet users today with the same number a few years ago you will notice that it has increased manyfolds. Almost everyone owns a smartphone and almost everyone use it to interact with Internet. Regardless companies that make business over Internet need to be sure to be able to comply with these large number of interactions in order not to lose customers and business opportunities.

Because of the increase in the number of Internet users an architecture shift was needed: distributed systems are the only way for industry to cope with this growth.

Modern applications that runs websites are made from open source software, VMs have been replaced by containers that have shorter startup time and are lighter to run.

Additionally containers provide the means for elasticity – they can be used just for the time needed for run the service – and allows for fast and frequent releases.

Data services provide connection and persistence.

Mesos is the core component of DC/OS whose purpose is to manage and abstract data center in the same way a traditional OS abstracts from hardware. Where the traditional OS operates on a single computer, DC/OS operates on the whole data center.

Operations such as installing Kafka or Spark on a cluster, with the proper configuration and replica count, is handled with a single click. You can find services ready to install and use on the open universe app store. As of today you can find 40+ apps

Here’s my thought. Well DC/OS seems to be the real shift in distributed systems. My impression is that the era of custom made software to handle the specific aspects of distributed system (such as load balancing, or provisioning) is going the way of the dodo since structured and standard solutions are being developed.

Having the facility to handle the whole datacenter with simple commands is possibly going to revolutionize datacenter in the same way docker revolutionized application deployment. The latter avoiding developer to keep track manually of all the dependencies and providing a sensible installer, the former avoiding manual install on multiple node with configuration twiddling.

Maybe this technology is a bit ahead – I’m saying this because I found no trace of the universe app store on the internet (maybe I didn’t search enough), but for sure is very promising and being open source, free for all, is something that for sure will provide widespread adoption.