scale.bythebay.io: Devin Loftis Interview

I'm Devin losses a vice president of I'm Devin losses a vice president of engineering at vellum L so we deal with engineering at vellum L so we deal with emails on a daily basis and our goal is emails on a daily basis and our goal is to authenticate emails and authenticate to authenticate emails and authenticate let me start back oh okay so things we let me start back oh okay so things we worry about at bella mel are really worry about at bella mel are really authenticating the world's emails we authenticating the world's emails we want to make sure that an email that you want to make sure that an email that you receive is actually a genuine email from receive is actually a genuine email from the person who sent it it's not a fish the person who sent it it's not a fish it's not a malicious email and so that's it's not a malicious email and so that's really our mission and so we have to really our mission and so we have to think about things about like how do we think about things about like how do we do that while making sure that we're not do that while making sure that we're not compromising a person's privacy or compromising a person's privacy or compromising information about a compromising information about a business or something like that and so business or something like that and so those are things that we think about a those are things that we think about a lot and that was kind of the impetus for lot and that was kind of the impetus for my talk here at the conference [Music] [Music] so yeah we we authenticated about three so yeah we we authenticated about three billion messages to date over the last billion messages to date over the last three years we've been around for about three years we've been around for about three years and you know as far as scale three years and you know as far as scale goes we have to make sure that when a goes we have to make sure that when a client has received an email message client has received an email message that that client could call out to us to that that client could call out to us to authenticate an email message to authenticate an email message to determine whether it's legitimate or not determine whether it's legitimate or not and so we have to be extremely extremely and so we have to be extremely extremely responsive we have to be extremely responsive we have to be extremely responsive to the needs of those clients responsive to the needs of those clients and so our problems of scale are ones of and so our problems of scale are ones of you know just returning information and you know just returning information and sub-second transactions and also sub-second transactions and also managing load across a back up on that managing load across a back up on that our so let me start all over again our so let me start all over again so our problems with scale are really so our problems with scale are really being able to respond quickly to being able to respond quickly to inquiries from clients about is this inquiries from clients about is this message illegitimate in our life and so message illegitimate in our life and so I mentioned earlier that we do about 3 I mentioned earlier that we do about 3 billion emails over the lifetime of the billion emails over the lifetime of the product and you can imagine that we need product and you can imagine that we need sub-second response times because if sub-second response times because if we're not responding there's a potential we're not responding there's a potential that a fish or something else is getting that a fish or something else is getting through to a customer so our problems of through to a customer so our problems of scale are really ones of scaling out of scale are really ones of scaling out of systems in a distributed manner so that systems in a distributed manner so that we can respond quickly to and query from we can respond quickly to and query from an email client [Music] I think the issues that we see in the I think the issues that we see in the future are you know a proliferation of future are you know a proliferation of message types that are being sent you message types that are being sent you know may email that's one thing that we know may email that's one thing that we deal with right now but we have SMS deal with right now but we have SMS messages the Internet of Things is out messages the Internet of Things is out there and there's an authentication there and there's an authentication mechanism around that you know I think mechanism around that you know I think you know we see the world getting more you know we see the world getting more more connected and we see the more connected and we see the opportunity for people to act opportunity for people to act maliciously online or impersonate people maliciously online or impersonate people on the line becoming easier and easier on the line becoming easier and easier and so when we're looking at scale it's and so when we're looking at scale it's really looking beyond you know email as really looking beyond you know email as a messaging platform to all messaging a messaging platform to all messaging platforms and and seeing how we can make platforms and and seeing how we can make that better for individuals [Music] [Music] so I'll be very honest with you at Val so I'll be very honest with you at Val ml we don't do a lot in functional ml we don't do a lot in functional programming right now the application is programming right now the application is a ruby stack and then we also use go for a ruby stack and then we also use go for kind of the rapid response email kind of the rapid response email authentication portion of the authentication portion of the application in the past though and we application in the past though and we have looked at functional programming to have looked at functional programming to support some of the things that we do support some of the things that we do with regards to different message types with regards to different message types at the end of the day we're dealing with at the end of the day we're dealing with DNS transactions here and so there's an DNS transactions here and so there's an opportunity for us to use pattern opportunity for us to use pattern matching with different types of DNS matching with different types of DNS requests to respond more efficiently or requests to respond more efficiently or more quickly there to make our code more more quickly there to make our code more succinct for a cleaner you know we succinct for a cleaner you know we haven't taken those those steps yet but haven't taken those those steps yet but there is always an opportunity for us to there is always an opportunity for us to do that down the road but again to be do that down the road but again to be honest we are not a functional shop honest we are not a functional shop quite yet but we do follow the tenets of quite yet but we do follow the tenets of good distributed systems good systems good distributed systems good systems design and we find those to be the most design and we find those to be the most important things to focus on I really enjoy the diversity of talks I really enjoy the diversity of talks that we have here I like the separate that we have here I like the separate tracks I like you know going back and tracks I like you know going back and forth between a reactive programming and forth between a reactive programming and a functional programming and a data a functional programming and a data track they they all overlap to some track they they all overlap to some degree but you get a diversity of people degree but you get a diversity of people here you don't just have if I wanted to here you don't just have if I wanted to ask all our functional programming or ask all our functional programming or Haskell I can go one place but if I want Haskell I can go one place but if I want to get a broader view of how data is to get a broader view of how data is being used or how it's being consumed being used or how it's being consumed there's another opportunity for me to there's another opportunity for me to learn those things here in fact it's learn those things here in fact it's it's great but one of the downsides is I it's great but one of the downsides is I can't see all the talk so I thank can't see all the talk so I thank goodness that are on the line at the end goodness that are on the line at the end of the conference so that I can see them of the conference so that I can see them then then [Music]