Venue
The venue was the campus of the Warsaw University at Krakowskie Przedmieście, Warsaw - the same place as last year. Generally fine - large rooms, good equipment, a lot of space outside to walk, relax and talk with colleagues. If air conditioning worked better and the corridors were a bit more spacious I would say it is perfect. Food and drinks - present, enough to survive through the day full of talks. For those with bigger appetites there are many bars and restaurants nearby (hey, it's the centre of Warsaw after all).
Two interesting features were iBeacons that were supposed to make navigation between conference rooms easier and electronic devices located near exits to allow voting presentations up or down. I did not rely on iBeacons for finding rooms and I always forgot to vote when leaving the rooms - but maybe it is just me. Anyway, I would rather see an option to vote up or down in the mobile application prepared for the conference.
Two interesting features were iBeacons that were supposed to make navigation between conference rooms easier and electronic devices located near exits to allow voting presentations up or down. I did not rely on iBeacons for finding rooms and I always forgot to vote when leaving the rooms - but maybe it is just me. Anyway, I would rather see an option to vote up or down in the mobile application prepared for the conference.Before lunch
There were five parallel tracks, so I had to make some difficult choices. I picked "Disposable dev environments" by Marcin Brański and Przemek Hejman for the first slot. They presented tools that can be used for DevOps deployments or simply improving developer's productivity when provisioning of test environments might take ages. It was an interesting topic for me as I am considering use of these technologies on my current project and my knowledge in this area so far has been only theoretical.
They started with showing Packer - a tool to build machine images for multiple platforms. Other tools covered during the presentation were Puppet, Vagrant, Ansible, Chef, Salt and to a slightly different tune - OS level virtualization solution: Docker, and its useful companion: Fig.
The guys even managed to run 3 or 4 scripts (Packer, Vagrant, Docker), but in my opinion it was an area that actually could be improved. Instead of simply showing a display full of logs after script execution it would be more informative to show how the environments built, set up and started using Packer/Puppet/Docker can be put into actual use by running some simple applications inside them.
It would provide a better learning experience even if it meant that due to extra time required some tools would need to be dropped from the agenda. In other words - a bit more depth at the expense of breadth.
The guys were quite enthusiastic talking about disposable environments, so I'm confident their next presentation will be a bit slower paced and less chaotic.
The next presentation I attended was about working with a difficult client ("Tajniki współpracy z (trudnym) klientem") by Kasia Mrowca. This presentation unraveled slowly, but ultimately provided quite a few interesting insights. Some that I noted are:
The guys even managed to run 3 or 4 scripts (Packer, Vagrant, Docker), but in my opinion it was an area that actually could be improved. Instead of simply showing a display full of logs after script execution it would be more informative to show how the environments built, set up and started using Packer/Puppet/Docker can be put into actual use by running some simple applications inside them.
It would provide a better learning experience even if it meant that due to extra time required some tools would need to be dropped from the agenda. In other words - a bit more depth at the expense of breadth.
The guys were quite enthusiastic talking about disposable environments, so I'm confident their next presentation will be a bit slower paced and less chaotic.
The next presentation I attended was about working with a difficult client ("Tajniki współpracy z (trudnym) klientem") by Kasia Mrowca. This presentation unraveled slowly, but ultimately provided quite a few interesting insights. Some that I noted are:
- Keeping systems small is important because it increases the odds of success. If the system is inherently large and complex - try focusing on no more than 3 key processes at a time. Showing statistics how much more often small projects succeed compared to large ones can go a long way when talking to business owners.
- Be careful when evaluating requirements originating from end users - they sometimes tend to copy the existing manual processes into new IT systems. You might end up reimplementing Excel functionality in a SAP system...
- Screen mock-ups are very useful. Balsamiq Mockups was mentioned as a good tool (I tend to agree as I'm happy to use it in my work as well). Prepare more than one mock-up version to avoid getting your mind fixed to just one layout.
- If preparing a mock-up takes more than a few hours - consider building a prototype instead.
- If building a prototype would take months then the system is probably too complex - see point one.
- Options presented to clients could be visualised using three dimensions: Effort estimate, Risk, Business Value.
- Use cases or user stories can be presented using UML-style notation as well as comic strips. Notation does not matter - choose one that will be most effective for the targeted audience.
- Do not present too many options - it will make decision making more difficult.
- And the most important: do not teach business people how they should run their business - they hate it (from my experience - I strongly agree).
Then I moved to another building to see Wojciech Seliga's talk about innovations. He started with pointing out some innovative practices employed by Atlassian (and Spartez) that are not related to products, like a simple sales model:
Later Wojciech claimed that innovation cannot be effectively planned or produced - you simply cannot order someone to be innovative. Innovation cannot be brought in by processes and policies. Traditional means used by managers in other areas, like money incentives, trainings, processes are not effective when it comes to innovation.
What helps is to create an innovation-friendly environment, where creative individuals can develop and work without being afraid to make mistakes. A zone of "safe failure", where the cost of failure is minimal, helps incubating innovative ideas.
Some ideas shared by Wojciech that work well for Atlassian are:
After the first three sessions there was a lunch break. I missed the warm meal provided by the organizers, but it was not that bad - I grabbed a few sandwiches and enjoyed some interesting conversations.
The main point that I remember is being active on StackOverflow and GitHub helps learning new things (as you get exposed to different problems than the ones you face in your own projects) and improving one's reputation among developer community.
There was also quite a bit of ego pumping during the presentation - not a big surprise given how expressive Jacek is as a person.
One more thing that was mentioned is Scalania - a learning project for people starting their Scala adventure: https://github.com/jaceklaskowski/scalania.
Then I moved to a session about NoSQL hosted by Marcin Karkocha. Most of the information was a fairly standard stuff about types of NoSQL databases, what are their relative advantages and disadvantages. The most valuable pieces for me were:
By the way - I wonder why MarkLogic is so rarely mentioned during presentations about NoSQL. It did not make it into Marcin's presentation, either.
I skipped the next session - I preferred to have spirited discussions with colleagues when sitting in a deck-chair with face towards the sun (did I mention the weather was just wonderful on Saturday?)
The last presentation for me was about organic architecture by Jarosław Pałka. I found it to be too focused on development practices and too little on actual architectural stuff. I cannot understand e.g. how removing unused code rather than commenting it out is related to IT architecture.
An interesting point was about defining architecture as a process of moving the system from one state to another. It highlights the fact that architecture definition takes time.
Another good point was about right tools - the right tool is whatever will help you get to the target state, regardless of what the current technology hype is (microservices, DDD, NoSQL or other BPMN).
Another interesting product of Azul is ReadyNow. It precompiles classes that are normally used during program execution to avoid delays due to JIT compilation at runtime. Since precompilation takes place at application startup, you get slightly longer startup time but faster and more stable performance later. A popular solution to this JVM warm-up problem is to run several dummy requests through the system to give JIT compiler a chance to kick-in before live traffic is allowed in. It seems that ReadyNow can avoid this nuisance.
Finally sentence of the conference: "It's better to ask forgiveness than permission" - heard twice on Saturday, once from Jacek Laskowski and once from Jarosław Pałka. In both cases it was in the context of improving codebase.
- publicly available, low prices
- no discounts
- self-service
Later Wojciech claimed that innovation cannot be effectively planned or produced - you simply cannot order someone to be innovative. Innovation cannot be brought in by processes and policies. Traditional means used by managers in other areas, like money incentives, trainings, processes are not effective when it comes to innovation.
What helps is to create an innovation-friendly environment, where creative individuals can develop and work without being afraid to make mistakes. A zone of "safe failure", where the cost of failure is minimal, helps incubating innovative ideas.
Some ideas shared by Wojciech that work well for Atlassian are:
- brown bags - knowledge sharing sessions at lunchtime.
- FedEx Days (renamed to Ship It Days) - a 24 hours long hackaton.
- 20% of working time to be spent on side projects.
- Ensuring that newjoiners are heard - they challenge status quo and bring a fresh view.
- HackHouse - a week long camp for graduate newjoiners to have fun and code.
- Company-wide transparency - all company information is available for everyone, except private and law-protected documents.
- Easy access to information - regardless in which system it resides.
- Collecting feedback, especially from users.
After the first three sessions there was a lunch break. I missed the warm meal provided by the organizers, but it was not that bad - I grabbed a few sandwiches and enjoyed some interesting conversations.
After lunch
The first session I chose after lunch was by Jacek Laskowski. The topic was how StackOverflow, GitHub and Twitter can be used by programmers for professional development.The main point that I remember is being active on StackOverflow and GitHub helps learning new things (as you get exposed to different problems than the ones you face in your own projects) and improving one's reputation among developer community.
There was also quite a bit of ego pumping during the presentation - not a big surprise given how expressive Jacek is as a person.
One more thing that was mentioned is Scalania - a learning project for people starting their Scala adventure: https://github.com/jaceklaskowski/scalania.
Then I moved to a session about NoSQL hosted by Marcin Karkocha. Most of the information was a fairly standard stuff about types of NoSQL databases, what are their relative advantages and disadvantages. The most valuable pieces for me were:
- Difficulties in defining SQL and NoSQL databases. Some SQL databases do not support ACID transactions (e.g. MyISAM engine for MySQL). Some NoSQL databases support SQL language for queries although they do not use relational data model. Not Only SQL seems to be the best expansion of NoSQL.
- Basing on Marcin practical experience, the following architecture seems to work fine in practice: MongoDB + PostgreSQL (user data, transactional data like payments) + Redis (cache).
By the way - I wonder why MarkLogic is so rarely mentioned during presentations about NoSQL. It did not make it into Marcin's presentation, either.
I skipped the next session - I preferred to have spirited discussions with colleagues when sitting in a deck-chair with face towards the sun (did I mention the weather was just wonderful on Saturday?)
The last presentation for me was about organic architecture by Jarosław Pałka. I found it to be too focused on development practices and too little on actual architectural stuff. I cannot understand e.g. how removing unused code rather than commenting it out is related to IT architecture.
An interesting point was about defining architecture as a process of moving the system from one state to another. It highlights the fact that architecture definition takes time.
Another good point was about right tools - the right tool is whatever will help you get to the target state, regardless of what the current technology hype is (microservices, DDD, NoSQL or other BPMN).
Other interesting events
I had an interesting discussion with chaps from Azul Systems at their stand. They sell Zing - a JVM implementation that makes GC pauses go away. It can be used wherever SLAs are strict and stop-the-world activities might lead do SLA breach. An example could be a trading platform, a messaging hub or even a human facing system if the underlying heap is massive and GC pause might take a few seconds, leading to terrible user experience. Azul offers a free jHiccup tool to monitor application responsiveness and facilitate diagnosing problems that could be resolved with Zing.Another interesting product of Azul is ReadyNow. It precompiles classes that are normally used during program execution to avoid delays due to JIT compilation at runtime. Since precompilation takes place at application startup, you get slightly longer startup time but faster and more stable performance later. A popular solution to this JVM warm-up problem is to run several dummy requests through the system to give JIT compiler a chance to kick-in before live traffic is allowed in. It seems that ReadyNow can avoid this nuisance.
Finally sentence of the conference: "It's better to ask forgiveness than permission" - heard twice on Saturday, once from Jacek Laskowski and once from Jarosław Pałka. In both cases it was in the context of improving codebase.
Brak komentarzy:
Prześlij komentarz