scale.bythebay.io: Petr Zapletal Interview
Recording: scale.bythebay.io: Petr Zapletal Interview
okay so my name is batters up at Tahoe sometimes Peter here I work for free cake solution /c bump dioxide Disney so basically I work for like for a large corporation alright I used to work as a consultant for a long time my my role is like a tie cleat or that consultant so I'm mostly doing like like overviewing systems repairing like architectural design sometimes writing some code reviewing code of other people interviewing people what's it like at the point when we are right now I think we know how to build like a large-scale distributed systems we understand it and we were able to build it I think right now we're tackling most like cost efficiency problems for example because like I can I can give you a number but we are paying like a huge amount of money to AWS like every month and it's quite hard to take a lot like because like dry like pros and cons always and yeah this is probably the problem I'm trying to tackle these days um apart from that in like a typical scaling problem or typical programs are distributed systems at least from my perspective is like state management so how do you want to manage state how do you distribute state how to make sure that our state is consistent or available or something else where else in case of programming we are we are using X Co alike like almost exclusively and I think we are tried it so I don't think like Scala or writing programs is the issue right now my finger issues are like from bigger scale [Music] was so many of them like for example just oh my hat I think we had to agree to redo the system because we have like wrong database cure for example or we had very using like wrong sterilization for example else yeah it was like we were incorrectly using gaseous like okay she's our something he's like so critical in distributed system because like it's same sort of time but it's like one of the hardest problems and in distributed environments are like cache invalidation [Music] like the project I'm working on has like projected scale and like millions of concurrent users these days so yes it's hard to write it it's right - it's hard to value I correctly and it's also like hard to ride it like a fishin enough that we can afford it because for example yeah like issue that if we collect all data we own it you would have to pay like tens of thousands of dollars a day just for a depressing class structure like robot to take oh I [Music] think we are quite functional like what are we doing I don't faint at functional programming is a sort of bullet which solves our problems definitely it's not it's certainly house with correctness but yeah it's like no I can always advantages like sometime it's hard to like on board people because like the number of people or number of people I know where we can hire first class for functional programming it's this pool is quite small secondly I'm sometimes I think that there's like it's not that efficient so there are like parts of the code which you need be like really efficient and functional programming is not always the best choice here we in the past with its couple factors from like really nice purely functional go to something less functional but more efficient but I'm sure that a lot of guys would argue attraction oh this is my favorite confront actually I was here like last year I think I had a like another talk I think we had the interview what I like here is that it's not commercial which is like awesome because I've been a couple conferences like in last year's like you know like spark something the reactive some aid and all these things and they are like trying to pushing some kind of product to this is like very uncommon there's not commission at all basically jan what I really like is they're all like free separate like how do we collect like channels or strikes three separate tracks which is so I think like everyone can like find something I'm conference it's like a great idea I had no chance to go Google yet but I definitely plan to do that else I like that talks out just 20 minutes because I think it makes total sense because like from the talk usually want to have like some basic idea and introduction and you are interested into that you can just you know dig into it or do something I think like 40 minutes talk we shall I think like more common are like I think people just lose conference concentration after like 10 15 minutes anyway so I think like 20 minutes course talks are also you