Containerizing monolithic apps: An argument for
Transcript
[Music] hey how's the one it's Keith Townsend from the CTO visor rule here at Interop with Eric Eric right disco passkey on Twitter we'll get into that hole where the twitter handle came but from one day but introduce yourself your day job title and why you interrupt aha good question yes my name is Eric rates I'm the Technical Evangelist for turbo nomic by day by night so I'm a blogger at disco Polycom twitter handle was at disco posse and yeah why interrupt I'm actually on the
review board for Interop which is pretty cool I love I love this community it's a very open community not just open source and Ross Conway but like actually open agnostic technologies out of traditional enterprise technology here and it's also a neat merger of like business focus so you've got like the CIO the leadership track and security CTO get further down into the practitioner it's a miracle we can actually pull this off with the handful of folks that are doing it like the UVM team that is
it's great because I've been a part of Interop for four five years now as a presenter as well as running some stuff I'm doing a mentoring track here and I'm also the guy who drags everything out on a 5k way of mine my legs are crying and I'm still hurting from that nine and a half minutes face so Eric I get you on because I got into a really interesting debate on LinkedIn about monolithic applications be eyes at Turin onyx do a lot of workload optimization
more and where the foot workloads we were talking off-camera about the Netflix story about how they took their Java applications put them in two containers and then get great optimization from ec2 that sounds like an infrastructure led offering what's your opinion just looking at Netflix and then looking at the average enterprises saying you know what is that a little JIT approach to take monolithic applications and cram them into containers right yeah it's a great question and you can clearly see those two camps right up the
advocate for the cloud native focus is gonna say it's a terrible idea right not unlike we would look back and say the pita bees are now a terrible idea however it was our path to virtualization so if I look at monolithic apps and containers again what's the actual goal we're trying to achieve do we want to get to cloud native or do you want to get to a new construct on which you can adopt new practices and I can tell you hands down in the hundreds
of enterprises that I talked to that's the right path it's not the well say it's the appropriate path it's maybe not be the right path right the right print family to refine your services yes so while I understand where there's resistance to that as a method to think about what you're achieving by doing it you're moving further up the stack your abstract away the Aya's layer to containerization and who's to say that maybe that's not where we're gonna go down the road we tried it with
you know OpenStack and it was like no one wanted ease openstax they didn't have live migration and those folks enterprise consumers got browbeaten by the open source community saying you should just develop your apps to be resilient but yet the application layer but what if you have a thousand applications and you see that I mean look at the customers and folks that you talk to it's a different scale I mean Netflix is funny of course cuz they have one act yeah obviously there's a lot of
other things and they they write their full stack but if Netflix is doing it there's a reason because maybe they couldn't get to the refactor early but what they did need was to be able to run multi-cloud great I could potentially you know start restart respawned this and then again it's like you've moved the construct into something that now you administer and operate differently and then you develop against the new pattern and that's very what it is so I agree with is the pattern adoption that
was the win the fact that the running a model Hey look not the best approach not the meth approach but what was the alternative the alternative is spend seven months for developers refactoring an application that maybe doesn't actually need to be a cloud native application there's actually a Randy shout if you follow Randy on Twitter he's Randy Sh OUP he's phenomenal you know and he talks he's got real great long-form blogs as well about why micro-services is a terrible idea for a lot of your apps
mmm because it's a it's an anti-pattern to a lot of particular applications why would I take an application that gets used three four times a week that gets updated once a year why why would I refactor that and spend engineering effort just let it live in a container Ron run you VM whatever and I think one of the push backs that I'm hearing from the DevOps community that you know what getting a Java app per se to run inside of the container it's just it's not
worth the trouble but Netflix put out some numbers that were really impressive they reduced their ec2 costs significantly by just taking containers and using containers more using ec2 it is this is more efficiently I don't know whole thing is on it you know I agree you know I think that's that's right I hate to make this a just a bunch of violent agreement the things but but it is true and and why do I agree because again what is the pattern that they've adopted which is
appropriate and we talked we've talked before as well you and I and a lot of folks about you know low density virtualization why would you do low density virtualization because the practice you've adopted means common administrative practice across every virtualization host why run a container on a VM again I've I get screamed at continuously about you know why would you do that because if I've got all these orchestration practices around patch management and all this stuff for that underlayer doesn't let it live live it works
fine I'm not trying to build a race car with containers containers is about portability a lot of enterprises containers is about packaging have cross or multi-cloud portability in efficiency so you know and then of course the ICB and then the ability to get to a point where i've refactored application is that it that's appropriate right and and if you look at you know again can what were the original containers they're typically they were deployed like ECS is on ec2 so you're deploying an ec2 instance it
puts containers on top of it now we're obviously getting into native you know we've got eks you've got Fargate you've got the kubernetes orchestration layers but again it's like because you can Tanner eyes that app you can now have complete freedom as to where you're gonna land it versus it's a wrapper for a VMware or a hyper-v or a rev installation so move to the container and whether you need to or not you've immediately made it portable and at that point now you can do is
it appropriate to do this next thing which is refactor where should it live and I can see I can tell you hands down the performance inside containers versus performance inside VMs its provably good so you heard it from the word of a person whose job if optimize workloads put them on the most cost-effective platform for the workload where can people find out about Trevor Nunn's yeah so Caesar could a turban Amma calm and we actually just had a release very recently I say recently because it
was a few hours ago from when we're talking depending on when they watched us now and and yeah so being able to do stuff in in cloud on-premises that's it's pretty cool so definitely encourage folks you can reach out to me as well it's act disco Posse and Twitter connect on LinkedIn as well and just search for Eric right disco Posse there's only one disco I can guarantee that all right and thanks for joining us from the floor of interrupt ITX talk to you next CTO
advisor CTO DOS you can follow me on the web the CTO of Viacom defying CTO doses at CTO dose calm talk to you next CTO toes
