Working with AWS Cloud: Expert Advice

wire and our topic today is working with Amazon Web Services in the cloud if your business hosts on the AWS cloud what are some best practices for working with the cloud giant to talk about that I'm joined by three folks we deep experience with AWS with us today is Jamil elaziz he's the lead infrastructure what are your title Dame I'll explain this to me today you're the lead of the infrastructure team at hi-5 correct yes I am okay got that out give me like the 11 second elevator pitch for what exactly hi-5 dies they said we'll listen now so we're a video conferencing company in our entire goal is to simplify the variant so we try to kind of approach it from a consumer perspective and try to build an experience that customers love to use and in our goal to get a video conferencing room out there also where this is Don Ferguson the chief technical officer I'd seek a TV hello to you don hi Don are you there done yeah can you hear me okay I sure can loud and clear you tell us I think probably a lot of people know what secret TV is why don't you give us kind of a short rundown on what secret TV is a Sukkah TV is primarily a streaming platform for independent web series producers in addition to allowing them platform for streaming it's a way for them to build an audience so they can interact with the viewers they can chat they can have discussions and we also see some interest for similar technology deploying it you know for marketing or product information in enterprises but it's um it's a mix of how details story with you know ten to fifteen minute videos how do you guide people through the flow right and the the other thing I do from a cloud perspective is own I've been teaching a class on Amazon Web Services at clunker yes like the past three years so it's kind of interesting too well yeah the thing that's interesting about that the first talk I became in service was you know can you teach old dogs new tricks because I was a little dog trying a new trick with them when I started teaching at Columbia this well can you teach new dogs the tricks because yeah I've got all these masters and undergrads who learned how to program but never use the cloud so it's it's an interesting experience great also with us is Bernard golden co-ed naviga and author of Amazon Web Services for dummies as well as other books low to you Bernard hey how are you today good good so Amazon where every service is I would ask the three of you you know what what is you know a key piece of advice you'd give to some business that's getting on board with AWS which was sort of the secret sauce to working with them it's on Web Services Jamil I thought you had an interesting idea yesterday what was your like gala top secret tip for working with AWS I think for me the the first tip I like to give any company I join in is honestly get to know and befriend your account to wrap around can make all the difference in terms of understanding AWS getting you in touch with the right people over ne WS getting in touch with the solution architects and even just think the documentation get you the files so it's like you have to romance but the account rep a little bit you think that Amazon Web Services would be kind of banging over backwards to cater to this the customers were you're saying that it the customer maybe should take a proactive approach to reach out to I mean your account rep usually will will take a very proactive approach and reaching out to you and being involved it's the the trickiest Ashley listen to them respond to them develop that relationship you know I've been at several previous companies and the relationship is different every time and the ones that are the most successful are the ones that actually nurture that relationship what they count about what would you say if you know forced you know II gave like one essential piece of advice that the company's going to have we're with AWS what would you say like oh definitely keep this in mind for you for your AWS deployment well I guess what I would say is that a lot of times where I do with companies is they sort they sort of approach it like well we've had our own datacenter but now we're going to use AWS so we'll use your infrastructure or asset of ours and its infrastructure your service I think I really emphasize is that's a really limited way to think about Amazon Web Services it's well well beyond that it's now a collection of services that you can sort of assemble you know on kind of a tinkertoy basis to build anything you want and if you just look at it as oh this is a convenient way to get computer storage you're missing out on what I would argue are really the core pieces to build next-generation kinds of applications like artificial intelligence you know data analytics what other tools might you say make sure people think about well you know there's a whole set of things that help you assemble these kinds of things so they've got notification service so you can send out alerts or you know stuff like that they've got an IOT service to help you if you're building some kind of device they've got an API gateway so that you can bring events into the system it got lamda which the Don of course is an expert in it reaches server list type stuff which is pretty revolutionary they've got a whole set of code building tools and DevOps stuff so you can kind of say I'm just gonna automate that without having to do all the community so it's a really remarkable collection services that really it's really important to go beyond just think of it as like oh it's it's a data center at the end of a wire right right there's more there I would say that's kind of like you know the very 2011 approach to cloud computing is well we can we can outsource our infrastructure and you will do with Amazon or maybe we'll do with Microsoft but really there is no this amazing collection of tools it seems like Amazon keeps on putting more tools out all the time yes you're right that it's a 2011 viewpoint but it's amazing how many people live in the past and yes really it's a big challenge and I would say even for the three of us who sort of live and breathe in this world to keep up with all of the stuff that Amazon develops and delivers it's it's incredible done what's your view and obviously you built caecus you certainly know the Amazon Web Services system and they're not mentioned you receiver less expertise said oh well what would you say to a business that is getting on board with a Amazon Web Services what's the best practice to keep in mind I'll say three things the first one is Amazon Web Services people people think of it being a cloud it's actually a platform I mean when people hear a cloud they think it the structure as a service software as a service but it is really a platform you know I think about it as a set of api's so it's like the dotnet platform the j2me platform these are the api's that you use when you build your system their infrastructure api's and there are some domain api's like historical information about pricing so I don't I I try to try to get everyone to think don't think about it as a cloud think about it doesn't set API such a call because it really is a programming platform the second one could to come back to what know what Bernhard said is he step one is you've got to think differently about when you build these applications because the very first version of a seka we built because of my background was kind of a classic j2ee application trying to run on top of the Amazon API is an infrastructure and you sort of aback look at a year later and you go what the heck was I thinking you know you think about I'm gonna build an app I'm gonna put it in a container and it's just it's a completely different way of thinking about the only applications and then the final one is that in addition to the relationship with Amazon I've found that getting plugged into the community of the users is pretty critical knowing the right forums knowing the right conferences knowing the right people to talk to even outside Amazon it's critical how would someone do that how would someone get plugged into the community what might be a form that you might recommend well there's some really good topical focused conferences on burnin and I've gone to the service conference which is quite good for people that are in that space there are local meetups you can you can find sub meetings at the larger Amazon conferences so sort of meetings with the meetings but I think the meetups and some fire I'm going to a conference in Nevada on middleware and there's gonna obviously be a lot of people there say I mean it's just going going to the right conference because it's down to the right beat ups your peer interaction is important and then part of the reason that's critical is one of the things like the the second slot I've put in my presentations when I teach cloud computing is a picture of Princess Leia talking to your 2d2 saying you know help you stackoverflow you're my only hope yeah the problem with Amazon Web Services to some the tum extent isn't lack of information it's too much information and dated information right and so knowing exactly where to go would rely on yeah um is critical you know let me let me follow let me stay with you for a second Donahue had made rather the topic of best practices for change management and I I want you if you could briefly define what you mean by that and and what what do you think it would be a tip for change management on the AWS platform well this goes back to the the tagging comment unlike a more classical application there's there is there are a lot of independently manageable moving parts moving you know parts artifacts I mean when we would deploy a web application you put down a jtv thing be the server what's the jsps when you lay these things down on Amazon there's you know CloudFront s3 lambda functions SNS so Ted the first comment about tagging was pretty insightful because you've got to tag these things just to know which things are in which application which means because you got lots of little pieces and you basically have to do the tagging to make sure that you know what actually makes up the end the application the second one is we try to take a classic change management approach which is okay we're gonna release a new version of the application we're gonna push it out so it's on it's the sort of the standard way of the way we would have done in an enterprise which is numerous and the application push it out provision everything and what we're doing now is we're following the microservices pattern but like everything's a micro service so we focus just on changing individual little things even if it's just the topic definition or a cue definition so we're changing little bits of things and making sure that what we're doing is additive well compatible so we don't think about changing the application we think about our image changes a little party or putting a version out that kind of thing so we were doing a classic in and you know DevOps but a classic stuffed change to push the change out and we think about it differently which is we don't change the app we change the little Legos Nate do you know you take on that here's a change management how do you think of it and and what might be a best practice in terms of AWS yeah so for me a change management change management has various different vectors in it and and you know on the official ID I looked as I try to think of everything as code right as how can I manage it I mean I try to think of control as my my mechanism of change management so how do I put everything in source control we use git you're a high five and so you know we used a terraform we use we use core OS so for us deploying deploying a machine we use cloud config and cloud in it to manage the actual staydry we perform to manage the state of the infrastructure the actual AWS services were running the instances everything like that and then from actual versioning all of that's the you know software versioning code versioning all that's done through gates so for us we try to keep everything in code so that the our entire state of AWS everything from the infrastructure from the instance that's running to application at the minutes of that other question it's running is done via 'get Bernard your thoughts on change management or would you say to folks about that well I'm gonna answer that and I'm gonna ask a question good so the way I'd answer it is I think it's it's very important to think about and particularly as organizations start making real what I'd call commercial or industrial use of Amazon I mean one of the outer was a downside but one of the challenges of Amazon is it's so selectively easy to get going that you can kind of get started and then much later go oh wait we're missing all this stuff or now we're causing now we have problems because we didn't do X Y & Z I mean there was there wasn't a sense of foresight there the people who didn't even know how to have a sense of foresight perhaps well maybe or they just kind of went you know hey it takes me two minutes to spin up resources you know I don't have time for all that other stuff and I just later when you've got you know 300 differences running and a bunch of different kinds of applications and they're all in different stages of development or production or whatever that's where it gets and tagging is a is a very well-established technique to segregate your resources into comprehensible sorts of aggregation books so you say oh that's that application production that's that application development or with my parent um Don I guess my question is what do you do what do you use in terms of tooling and process around that because it sounds like you've got a finely grained application you try and be able to update individual piece parts within that application so it's not a giant release it's you know we're releasing you know some part a kind of thing what what do you how do you the tooling in the process for that primary Tony's firmly tilting Bernard by the way thank you primary tools we have our music committing Jenkins so when we commit something it's a corresponding Jenkins that determines what needs to get updated and then that may cascade by many than event so it was code commit is the Amazon sirs yeah it's their version again uh-huh yeah it's it's their version again it's their get environment and then that spits out to a Jenkins service or do you have it actually also based learning to be yes we have a Jenkins server that reacts to the event and runs the Jenkins scripts okay Bernard did you get your full answer there well I got enough that probably I'll you know at some point offline you know separately thing actually so that's a really key issue once you start saying yeah this is a core part of the way I do computing for my enterprise for my organization right sit and all that stuff is you know it's kind of the difference between I'm gonna throw a meal together at my house for four people and I'm gonna run er you know run a restaurant with a hundred 50 tables and you know it's cetera you have to have a system for our restaurant 150 tables you can't just sort of go I'm a really creative cook so what about the issue and this is maybe the most important part of this it's the issue of cost Jamil any any tips for people in terms of like you know either either shave a few pennies better manage the whole system in terms of cost what we do tell people yeah so you know we discussed this briefly before this started um the first thing as everybody mentioned and as we brought up multiple times us you know make sure you have a robust tagging scheme out there that's gonna help you identify your resources is going to help you understand where costs are coming from you can see that in your building manager and kind of drill down to see where your costs are coming from the second thing I would look into is Amazon provides you know they don't make any effort to hide cost-saving measures and what I mean by that is used spot instances where you can use reserved instances where you can use s3 storage analysis and I five objects you're not accessing frequently used different storage tiers for that you know there's plenty of ways to go in and reduce your costs but the key is to discover where your costs are coming from and the only way you can do that is by being very aware of what is running where it's running and why it's running and for me you know that's where you know you start with tagging and once you have that in place to start you can start with your area so weight reduce your costs so do you think that Amazon is helpful we've got going over those kind of issues absolutely I mean that's what brought up the tip of brink bringing your AWS rep is uh I think from experience your AWS rep is entire purpose is to make you happy and so actually quite often I find that they give me advice counter into if in the sense that Amazon and will make a lot less money with the advice they're giving me but they give it anyway the reps which is a quick point but I've heard about the reps that the larger your account is the more they pay attention you which is only natural I mean is it you think a small fry might get sort of ignored in the process and the Amazon Web Services Platform not as been growing not a few making it you know they don't you don't that you're betting on Amazon and that you are going to grow with them you know I when you're kind of breaking up there Jamil what's not so that one more time well even a smaller company that I've been at we even high five which is not necessarily the local account we get a lot of attention just by being involved and working with Amazon John what's your taken tiers of saving a few pennies every month any any test for users on that I would echo what Chanel said even if you don't it's a ease the word counter into it even as extremely counterintuitive but they have a lot of good best practices tools guidance on how to reduce your spend it it's it's not something you would expect you would think that they would want you to kind of increase this then but there's a huge amount of material about out there about ticket how to take a look at ringerman actually the decent amount that you're spending it is pretty remarkable that I I mean I think IBM might have done that in the mainframe days because the mainframes are so expensive but I haven't seen anything else like that they seem dials sort of actively put an effort into helping the customers spend less with them and it is remarkable County to it make in the the other one about the tag even I was tagging you've got to have some discipline over what you actually create cuz it's pretty easy for these things to swallow it it's so easy to make a new thing created a new thing it's sort of the you know when we were student some of the cloud stuff early on it was a rogue IT thing right yet and and this could be rogue IT con wild see you really have to put it either control and the console or even just internal discipline in place about who's gonna who's gonna create something that I actually take it down because it's really easy to just leave little fingerprints all over the place shadow IKEA if I remember it seems like they don't hear as much about shadow ID as we used to would you take saving a few pennies in AWS how would you bet would you do that well I mean everything has been said is extremely germane and I would just extend that to say that you know Amazon has a tool called trusted advisors but you pointed your head up and it'll come back and tell you things like because one of the places that people spend a lot of money or waste a lot of money is underutilized resources so they'll you know they'll buy an instance sighs you know it's like wearing a pair of pants that's got a 40 inch waist just in case you ever lose gained weight and you know well they kind of have this thing assist Joey you probably fit into a 32 Ross us and so you know I'm the most effective I've found the most effective thing however is to make sure that your girlfriend gets the Amazon that the credit card with the bill comes because she starts asking lots of questions um so that that's real was really helpful for me I have I'm sure that'll be a new AWS service pretty soon that nagging is a service well but the tool basically will tell you you know you've got this underutilized resources sort of gives you a place to go drill down and from the from my experience in the research I've done underutilized resources quickly underutilized instances is a huge money pit and by managing those more judiciously you could save a lot of money and then another technique that sort of aligned with that is and this is sort of getting back changing the way you think about things rather than sort of buying up you sort of spread horizontally and that way you can add and subtract resources in a more granular fashion and match your cost that are to your demand curve can you give an example of a spreading horizontally what you mean brother or not well in the old days when it was really hard to get more servers installed what you do it is you'd say oh I need more capacity I'll just buy a bigger server and then Temptation lot of times was well where's this thing gonna be in three years or two years well I'm not really sure so I better buy a size 40 pair of pants because it goes we're not just in case and so you end up with these things with tremendous underutilization a different technique when infrastructure is no longer hard to obtain like an Amazon where you can just say oh I need another instance it took me to Mississippi torn up is you typically will first all you'll have redundancy so you might have two web servers instead of just one because it's easy to have two and and then what you do is rather than using big one big instances or larger instances you use smaller ones and if demand hit we'd say all spin up another one or two more or five more and use load balancing to spread traffic across those and that way the amount of computing resource you've got matches the actual demand and then if the demand drops off you can easily slough off those extra instances and go back that's a that sort of horizontal scaling of her - or horizontal growth is a good technique to avoid this kind of I got a big honking machine that's running at 4 percent right you know it's a lot of good stuff I think you guys said it I anyway like one last thing you'd want to say like if you're gonna be an Amazon Web Services definitely think about this it's you know my my real secret John what would you say like one last thought about like working on Amazon Web Services there's something that's particularly important for businesses to keep in mind yeah I they've got a very sophisticated security policy model and I don't think people put enough thought into that and how to use it it's extremely powerful it's extremely complex but I think it's a part of the solution that people don't really think through very well it's a question of knowing how to use it or knowing it's there you think people know about it and they don't just simply go no I can tell you know huh and it's to avoid getting hacked is that a part of its a to avoid getting hacked but more for a part of it is just being able to build the applications that are intrinsically secure so lemme has the caller running as the you know the principal it's really a quite a sophisticated mechanism uh-huh so I think people will understand that Jameel you're sort of last piece of advice you'd give to this this person on a Tovah AWS I mean I watch something Donald esenin also a you know sleeper not mentioned is the entire level is AWS is all about the api's and in order to really understand interview us and and to do well in anybody s honestly read the documentation and read as much into the documentation as you can it's a very easy to get started on AWS and because it's very easy to get started maybe yes it's very easy to make mistakes that are hard to recover from and it's and once you have an application running in production it's almost impossible to recover from some mistakes so it's really really important to understand what the limitations are what the caveats are before you get star otherwise you end up in a situation where you know you can't recover and a lot of that is just planning and we documentation what's an example of so much terrible mistake someone might make right up front and they can't recover from easily a one that's actually one of them hardest to recover from is choosing an appropriate subnets for a V PC so Amazon pushes a you know it's been pushing VP Caesars for the last few years and a very common mistake I see at some companies I've worked at is you you just choose an inappropriate subnet and what I mean by that is you either choose a you know subnet prefix that is too small so you end up or you end up with too few IP is available to you or you just end up with conflicting ranges so you can't appear to be PCs there's a bunch of different situations but you know even at the networking level if you make a mistake it's very hard to recover in Amazon so unless you just completely create a new environment and then and that's why it's one of those things where it's just read the documentation understand the limitations and understand what future is you might want to use and plan according to that regard what's your what's your final tip for the the AWS user what was a really important thing to keep in mind well I I think I would use continue that last thought about the sort of the VPC of the VPC peering I mean this is a very different approach to things pretty different than you know then what most organizations have had in their own data centers they sort of didn't have this kind of approach and and that sort of sort of ties back to where to my first piece of advice which is don't think of this as just your data center at the end of a wire you're old same as your old data center just the end of wire it's a it's a very really different kind of environment it's got a rich panoply of services and incredible portfolio that's growing all the time and and to sort of recognize you're dealing with a different animal and to and to embrace that rather than say well it's the data center at the end of the wire and get frustrated cuz well how come it doesn't act just like I used to like used to you know and I think that's that's very important to sort of come to it with fresh eyes rather than the same eyes that you used to use when you were looking at your you know your rack right I'm always thinking you know the the word cloud may be slightly obsolete metaphor and that the idea of cloud presumed it was this distant thing you might tag into but actually it's it's really almost very intellectual business at any rate thank you the three of you for our expertise also did the the link when it's all edited we can we can tweet it greatly briefly appreciate your thoughts thank you Tennessee thank you

This transcript was generated automatically from the video's captions and may contain errors.