Scale By The Bay 2019: Jason Swartz Interview
FunctionalTV interview with Jason Swartz.
Listen
Download MP3 ↓Follow the words
Read the transcript
[Music] they call me the captain but my name is they call me the captain but my name is Jason Schwartz and I'm the edge Jason Schwartz and I'm the edge engineering manager at twitch oh I organized the first few Scala by oh I organized the first few Scala by the bay conferences with Alexi we had the bay conferences with Alexi we had previously worked on the first Silicon previously worked on the first Silicon Valley Scala symposium back in 25th 2014 Valley Scala symposium back in 25th 2014 he did the organizing I helped out there he did the organizing I helped out there but we Co organized the first few Scala but we Co organized the first few Scala by the bay by the bay we had a great one at Fort Mason with we had a great one at Fort Mason with food trucks and everything and then the food trucks and everything and then the following year I think 2016 in Oakland following year I think 2016 in Oakland at the Kaiser Center so it's great to be at the Kaiser Center so it's great to be back here this is a beautiful venue for back here this is a beautiful venue for for scale by the bay and I always have a for scale by the bay and I always have a good time it's great it's so great to good time it's great it's so great to always meet new people and see old always meet new people and see old friends it's always a very friendly friends it's always a very friendly outgoing group at this conference most important part was I repeated most important part was I repeated several times best way to do several times best way to do high-performance server server lists high-performance server server lists functions with Scala is use vanilla functions with Scala is use vanilla Scala and then compile it to native Scala and then compile it to native Linux executables with Gras VM and then Linux executables with Gras VM and then run that on AWS lambda and if you do run that on AWS lambda and if you do that if you actually run your Scala code that if you actually run your Scala code in as a Linux executable cross compile in as a Linux executable cross compile to Linux with Gras VM it runs really to Linux with Gras VM it runs really fast and there's almost no cold boot fast and there's almost no cold boot times you there cold boot times go down times you there cold boot times go down to maybe 200 milliseconds regular to maybe 200 milliseconds regular execution time maybe 50 milliseconds it execution time maybe 50 milliseconds it is really fast and there's no lag at all is really fast and there's no lag at all and it's just a perfect way to run these I manage and lead the edge team which I manage and lead the edge team which handles the core API traffic coming into handles the core API traffic coming into twitch we have millions of concurrent twitch we have millions of concurrent users watching streams and participating users watching streams and participating and chatting and being part of these and chatting and being part of these wonderfully immersive multi-user wonderfully immersive multi-user experiences and we have searches we have experiences and we have searches we have we have new things happening we might we have new things happening we might have fortnight events you might have have fortnight events you might have League of Legends games there's a lot of League of Legends games there's a lot of wonderful eSports entertainment eSports wonderful eSports entertainment eSports and entertainment and events that happen and entertainment and events that happen where we'll get searches of traffic and where we'll get searches of traffic and we have to make sure to support that we we have to make sure to support that we might go to half a million to several might go to half a million to several million requests per second coming into million requests per second coming into our edge services to provide a wonderful our edge services to provide a wonderful dynamic user experience for users and we dynamic user experience for users and we want to make sure no one's left behind want to make sure no one's left behind we wanna make sure creators can use all we wanna make sure creators can use all the creator tools they need to properly the creator tools they need to properly broadcast and push video up and broadcast and push video up and communicate collaborate with people we communicate collaborate with people we want to make sure the twitch extensions want to make sure the twitch extensions run on the stream and are fully run on the stream and are fully interactive we want to make sure interactive we want to make sure everyone can join the channel join the everyone can join the channel join the chat contribute bits emotes be a chat contribute bits emotes be a subscribe be immersive part of this and subscribe be immersive part of this and we want to make sure the scales for we want to make sure the scales for everyone everyone even when there's really high traffic even when there's really high traffic events that happen so we do a lot of events that happen so we do a lot of work to make sure that we scale and our work to make sure that we scale and our edge services can talk to the hundreds edge services can talk to the hundreds of downstream services that make up the of downstream services that make up the twitch user experience at the company yeah I was looking forward to the zio yeah I was looking forward to the zio talk and the the way that this is an talk and the the way that this is an improvement on futures improvement on futures I still use Scala futures it's really I still use Scala futures it's really one of the only concurrent means I know one of the only concurrent means I know of I think it works of I think it works I think the monadic future works really I think the monadic future works really well but learning about fibers and the well but learning about fibers and the fact that there is a more performant and fact that there is a more performant and interruptive way to do concurrency interruptive way to do concurrency actually sounds in some ways a little actually sounds in some ways a little bit similar to using go routines and go bit similar to using go routines and go channels which I really like in the go channels which I really like in the go language I want to learn more about zio language I want to learn more about zio and fiber-based concurrent programming and fiber-based concurrent programming after that looking forward to growing the team and looking forward to growing the team and hoping to find people at the conference hoping to find people at the conference who be interested in working at twitch who be interested in working at twitch otherwise we're looking to create more otherwise we're looking to create more of a dynamic user experience and make of a dynamic user experience and make sure that we can continue to scale and sure that we can continue to scale and meet a lot of the challenges of scaling meet a lot of the challenges of scaling up to have millions and millions of up to have millions and millions of people participating in the same event people participating in the same event on the Internet [Music] [Music] the special speaker washroom what didn't the special speaker washroom what didn't have quite all the amenities I really have quite all the amenities I really like I am kind of a Vitalis no no I like I am kind of a Vitalis no no I don't like not about that it's about the don't like not about that it's about the people I love coming here and meeting people I love coming here and meeting people and it's great uh I'm not sure people and it's great uh I'm not sure what to change I love the venue it's what to change I love the venue it's really fun to be here the talks are really fun to be here the talks are great and it's so much fun to meet new great and it's so much fun to meet new people and make new friends it's a people and make new friends it's a that's the most enjoyable part of the that's the most enjoyable part of the conference for me conference for me [Music]
Recovered English captions. Automatic transcription may contain errors.
Keep exploring
Follow the guest, their work, and the ideas behind this conversation in the Devreal knowledge graph.
Jason Swartz on Devreal ↗