Cloud Crunch
Cloud Crunch

Episode · 5 months ago

S2E13: 5 Strategies to Maximize Your Cloud’s Value: Strategy 5 - Reducing Cloud Spend While Accelerating Application Deployment

ABOUT THIS EPISODE

Covering the 5th and final strategy in our series on maximizing your cloud's value, cloud cost optimization expert, Jeff Collins, joins us today to examine our 7 Pillars of Cloud Cost Optimization. Find out how to apply these pillars to reduce your cloud spend without decreasing the benefits of your cloud environment.

...involve solve, evolve, welcome to cloudCrunch, the podcast for any large enterprise planning on moving to or isin the midst of moving to the cloud hosted by the cloud computing expertsfrom Second Watch, Ian will be chief architect Cloud Solutions and Skip Very,executive director of Cloud Enablement. And now here are your hosts of CloudCrunch. Welcome back to cloud crunch. If you'vebeen listening for the last few weeks, you know, we've been talking aboutstrategies you can use to increase the value of being in the cloud. We'vealready examined the first four strategies and have arrived at thefifth and final strategy we're going to discuss in the Siri's reducing cloudspend while accelerating application deployment. Today, I'm joined with myco host, Skip. Very Skip. Good to see you. Good to see you in Awesome.Awesome. And today our guest is Jeff Collins, who is one of our productmanagers here at Second Watch, who has a lot of experience in this area andparticularly out modernization in all kinds of things along those lines. Sowhat we want to do Jeff, is welcome you to the show first. Thank you guys.appreciate. Yeah, Now this is great. We have the opportunity to work with Jeffquite a bit, so we know his experience out in the field is very relevant. And,you know, that's what we try to do on this show is is bring the real worldexamples and our experience back to our audience. You know, with the businessand the technology focus, of course. And Jeff is an ideal candidate to talkabout this topic today. So, Jeff, let's just dive right into this. Let's talk alittle bit about how you can reduce your cloud spend without decreasing thebenefits of your cloud environment. Yeah, and I guess that, you know, Cloudspin is probably one of the one of the most important elements when you domove to the cloud, you know, because that's kind of the the initial thoughtprocess. You know, I can get out of my data center and I can save X amount ofdollars on network and power and storage costs, and I could move intothis fancy new cloud thing and just click some buttons and deploy my stuffthat was in the data center into the cloud. Problem is you can actuallyspend way mawr if you're not careful. So you gotta kind of know what you'rewhat you're doing. And typically, when you do a migration into the cloud, itcould be like a one for one transition. So a lot of customers, if they're notquite to that point where they can start thinking about optimizing or appmodernization? Well, basically, just take an export of what's there runningin the data center, say they're on being where they've got a 200 VMS, forexample. They want to put that in the AWS. So they go and spend up the samenumber of instances with the same amount of horsepower. Set up a kind ofthe same type of networking configuration. And that's how they dotheir migration without any kind of pre thought into what you know, What can Ido to make my environment better than what I had before and take advantage ofsome of these additional technologies? The cloud provides. So wait a minute.You're telling me that you could end up spending more in the cloud? It'spossible. Yeah, yeah, yeah. What is it? That what they used to call it thedouble bubble bump. Whatever the first bump or whatever, when you put workloadin there, is that still a thing? Double bubble, Yeah, especially if you don'tknow what you're doing. There's lots of even click on. You know, when you login to the console lots of buttons to click on, you click on the wrong thing.You you could easily start over spending. Yeah, I think we shouldprobably explain the audience that may not know that term. The double bubbleis basically when you are in the process of migrating to the cloud andyou're both you're paying for that application resource is both in thecloud and on your prem data centers wherever that may be. So the promisethat, you know, I just want to backfill this is that that bubble will go awayin a relatively short period of time. How short? Hopefully shorter the better,right? I mean, that's that's what we're trying to shoot for here and get backto that payback period. Brings in the interesting topic of optimization. Aswell, right, Right, right, right. And...

...here we are very excited. Jeff, youknow, for me this topic from being in pro serve and you know what you all doand you know where we are and working on this Very interesting. Uh, talkabout some of the technologies out there, you know, from if you look at itfrom the pillars of cloud optimization from your perspective. Yeah, andthere's there's various tools that you can deploy to get, do a lot ofreporting and do some recommendations. E don't want a dog. You know anyspecific tools, But if you do that, you're kind of silo in yourself. Sowhat we kind of like to look at is your cloud environment as a whole. And to dothat, there's some pillars that we've kind of set up on the optimization sidethat will help guide that conversation and help look for elements that cansave you money, right? So there's seven of them that we use here a second watch.The first one kind of takes on that whole. If you're familiar with Finn opsand kind of those methodologies of how you can centralize your spin and kindof delegated out to your different business units within an organization,we do take advantage of that, um, that mentality. And that could be anythingfrom kind of looking at reserved instances savings plans, purchases togetting into like GDP and be a negotiations. Kind of further of thefinancial staff was kind of looking at all those various elements and what Imean, what's going to make the most sense for your environment and how it'srunning today so that that's 11 element. Second one would be auto parking, andthe theory behind that is, you know, even if you have coverage within withthe reserved instance or a savings plan, it might make more sense just to cutthe thing off. You know, especially if you only need it running for a couplehours a day, or just 9 to 5 business hours. You're not. You're not payingfor those. Compute Resource is to be sitting there idle, you know, whilethey're not needed. Third one rightsizing that zit. That's definitelya big one. You'll see a lot of tools out there that'll give right secondrightsizing recommendations, basically, just based on your CPU and memorymetrics. A lot of them actually don't pull in the real time memory metrics.They always set that to like 100%. So you're not really getting the actualamount of compute usage that you're using, believe it or not, so thatthat's something to look out for when we're providing recommendations. Ourcustomers here second watch, we pull in real time memory metrics and make surethat your your instances or your VMS are properly size based on what'sactually running on them on the next one would be family refresh. And thatone like you, you know another kind of thing about the cloud. You know, it'sIt's not like this vapor of stuffs. That stuff just runs on right. There isa data center sitting there. There are physical servers, lots of them, um,that air running and occasionally you know, they get old and they have to berefreshed. So a lot of the public cloud providers will will provide incentiveswhen some of that hardware is becoming end of life and they want to migrate tonew families. That will probably give better performance for yourapplications anyway and also provide some additional savings for running onthe on the old legacy stuff. Um, the key point, I believe. Just a drill downon that one, Um, you know, in the in the cloud itself, it's really you know,the panache now becomes how how well you are as faras managing all thisright with the with the ever evolving landscape inside someone else's datacenter, a k a. Someone else's cloud, right? So it's a very salient point.Your former job of how you made that hardware last for three years is now.How do you maximize your spend? Really, in every time something new comes out anew instance or family or what have you? So that's I spent a lot of time talkingto, you know, my counterparts in a customer explaining that situation justa mind shift itself that it takes. Yeah,...

...it's a It's a very valid point becausewe often walk into a customer, will ask them what their refresh cycle is beyond.It's usually between three and seven years. I think when they're at sevenyears, they usually publicly don't like to talk about the seven years died, butfive is pretty typical. But this you know that this family refresh just adrilling for those that may not understand the family. The instancetypes obviously like If you look at Amazon, they have a C type. Theystarted with C one a long time ago. Now they're upto See, I don't know, 900 Atthis point, I believe it's C six, but every time the generation obviously thechip sets that architecture gets better. So But, you know, it's reallyfascinating to me. So if you have, like one family and they come up with a newinstance type mean it breaks the whole refresh model, right? I mean, it's notlike you're not gonna wait for five years. You can wait for the maintenancewindow to hit it. Yeah, just to follow on before you get into the last two.Jeff, I want to take a step back because I think it's a very importantpart to mention Ah, lot of this isn't even technology. It's about contractualmanagement aspect. You mentioned E d. P. Volume licensing. Can you Can you drilldown into that a little bit more to just about you know, not necessarilywhat second watch offers, but just the mindset that we try to instill in ourcustomers. Uh, e guess there's different. You can think of it kind ofis a cake right When it comes organization. There's different layersof that cake that you could apply to achieve the most savings you know. Soand take AWS, for example, savings plans could be a layer of that cake.Reserved instances, even though they're they're kind of going away. Savingsplans kind of give you a better coverage. It's a lot more flexible. Youkind of base your amount of coverage on your actual usage and spread it acrossmultiple cloud services a little bit more flexible. Think of that is thelayer of the cake. And then you also got kind of your overall overarchingcontract with with a W s that you can negotiate better rates for dependent,Understand? So when you take all those different elements, um, it's kind oflike it's kinda like baking a cake, right? You've got this multi layer. Um,it's a great analogy. Which part of the E pretty good, you know. But once youwhat, you get all those pieces in there, you've got a pretty good cake. If youwant to use that analogy s just different. Different things to look ata very critical part of the all this for sure? Yeah. 11 thing That's just atouch on the family Refresh. I did notice that you've seen a lot ofannouncements. Especially, totally reinvent, I would say, probably thelast, I don't know, six months or so from AWS. More, more So it reinventlike I just said, but around Amazon kind of getting into that. You know,there are cloud provider, but now they're getting into the hardware space.We saw it with this. Drives like where they would they would go to, Ah,Western Digital or C gate. I would say, you know, for for the glacier, we wantyou to provide us this speck of disk. You know, we don't If it breaksstraight, we'll throw it away or well, don't throw it away. Well, recycle it.But we wanna is cheapest possible. And now you kind of see that mindset goinginto processors with their announcement of the gravity on to and how muchcheaper they are, you know, So it'll be interesting to see kind of how that'sgonna affect the different families that they come out, especially ifyou're moving away from an celery and B to provide additional cost savings efeel. It's like almost the last dagger to push into the chest of getting outof a data center. Because that seems to be where the I want to say thereluctant are still. But But that seems Thio they're taking that last obstacleout of Why would Why would you still have your own infrastructure and whathave you So it will be interesting to see I agree how that plays out.Especially 21. Yeah. Yeah. Now you have so many choices on these differenttypes of, you know you know? You know, components. I mean, who would havethought that Amazon would be making their own ship sets, You know, I mean,it makes sense when you finally realize...

...it, but they're they're large consumerof them, but it's fascinate. And I'm gonna pull a skip here, I'm gonna say,But before we go on to the next to I wanna go back Thio phrase because it'sjust kind of lingering out there. Finn ups. Yeah, Let's go back and dive intothat a little bit further. And if you could I mean, you kind of touched onwhat it was, but what's your definition of Yeah, it's more like a kind of amethodology. It's not like a you know, from what I've read, at least it's notlike a you know, there's a set tool to give me a fin office Reports were morekind of like a mentality that you get in. Like, for example, we've got alarge customer that that has kind of like their Central I T department thatdeals with all of the financial aspects of the business, right, So they're theones that will do the E, d, p and the savings plans out. External of that,they've got multiple, different business units that operate fromwhatever guidelines that the central group says right so they don'tnecessarily take can take advantage of all of the discounts that the overallorganization is getting that got set internal rates, and it's really justkind of a way. Not that not that they're charging themselves more moneyor anything. But it's really a way to kind of get your arms around bestpractices financially and make sure that all of your groups are kind ofbeing held responsible for what they're spending their aware of their spend,and it allows that central group to kind of control the mindset of thosedifferent organizations internally. Yeah, yeah, that's the conversationthat I was mentioning earlier, that I have often of, like, trying to get thetraditional infrastructure director or VP to get their mindset in that in thatworld in that realm, right, new. It's a new thought for them. Really? Yeah, wayhave seen that play out. I don't think it was called pinups when we initiallysaw this. But when somebody from finances attached to I t. Particularlytheir cloud center of excellence and then they're they're helping understandand do that, you know, making sure the charge backs everything are all good.It's the first time I ever got the ability of how much I tease costingeach business owner so exactly exactly it's not. It's not a one and done thingeither. It's ongoing, you know. So once you kind of get these like I said, Iwould I would picture it. Maura. Just kind of a way of thinking or amethodology. Once you kind of start adopting those those differentstrategies, you know you can apply and going forward and, you know, seeadditional cost savings and efficiencies. All right, now thatyou've satisfied my a d d me jump back around, let's see if we continue on tothe next couple. Yeah, yeah, way Actually have three more. We have a newone that we had recently, but no one is the next one's waist. So any anythingthat's out there like, you know, unattached storage volume instances orVMS that have been sitting there idle, not not getting in a use for for, youknow, 30 60 days old snapshots. Now a lot of a lot of companies will have, Ah,good backup policy, but not necessarily a good snapshot deletion policy. Somaking sure that, you know you're not keeping things any longer than you haveto because the cloud provider is going to charge you, you know, whether youuse it or not, as long as the storage is sitting there. So just kind ofgetting some good visibility into what those different waste elements are. Thenext one we storage storage has gotten. I wouldn't say complicated, but I wouldsay there's tons of options out there, even though even like with S three andGlacier, there's multiple flavors of each one and kind of having a goodunderstanding of what your customers workloads look like. And setting a dataretention policy that's gonna work best for them, you know, based on frequencyof retrieval. Or you know how, how quickly, how how important a piece ofdata is. You know, can you go through it in Glacier? And if you can pull itout within, you know, a day or two is...

...that fine? You know some of the data?Yeah, some of it may be not. So you've got a kind of balance, those differentelements to make sure that that you're not over paying for your storage costsbecause they don't typically go away. You know, either store Cem Cem data oryou don't. Right. So they're going to charge you if it's sitting there. Evenif there is ah, you know, compute tied to it. You might not not necessarily becharged for the compute portion of it, but storage is gonna take up space, andthey're gonna they're gonna ding you for those charges, regardless on thenThe next one is kind of a new one. Like I mentioned, it's kind of more of astep into continuous optimization, and it's it's specifically around spotresource is so if you have customers that can take advantage of that, usedto be kind of a pain, right? You had to go to the market place. You had Thio.Basically go buy some unused resource is Jeff. If you could just explain toour audience what you mean by spot? Yeah. So spot is a is a methodology ofbuying. Compute resource is that people no longer need They might havepurchased it and they're looking to they don't need any more. So they sellit on a marketplace, which it used to be a pain to manage a lot of the cloudnative tools over the past year to have gotten a lot better at it. We partnerwith a company called Spot I O um, basically takes that whole headache outof the way. So basically, you you target a workload. It could be a autoscaling group. It could be a just an instance of VM. Uh, they can also becontainers. It could be ks kubernetes E C s. And you basically loaded into theinto the tool and it finds those spot resource is based on availability, andit's going to run them, you know, constantly. So you could You couldactually run production where close. And there is well these days. And iffor some reason that spot resource goes away, it'll just convert over to ondemand. So basically takes away all that headache of ensuring that you'vegot those resource is available and automates it so that that's that'sanother one for workloads that can take advantage of it. Now, that's great.Yeah, we're seeing more and more of that. It's getting more sophisticatedas well. Yeah, there's a lot of people that I think are terrified of usingspot because particularly have state collapse, that it may shut down on them.And then they're they're losing some data. So obviously we want Thio thatgets a nap. Modernization. How do you re architect that app to make sure thatyou could take advantage of all these things as well. Do you want to tell onestory, though you have to be careful. Obviously, you follow these rules, youget the optimization, you get your backups in place and then you find outthe most expensive part of your platform for and I'll pick on what Iwant is the name of it. But there's Ah, very traditional big data platform thatpeople often migrate directly to the cloud. And if you start backing that upfrequently like you should your cross your spiral out of control. So you haveto kind of look, you know, that's when you need to start at modernizing usingcloud native Technologies as well. We had one client that we didn't set thispart up for him, but they were spending almost $200,000 a month on back up justfor one application. So we had to go back and re engineer that with them andsay There's a better way of doing it and it really dropped it down. So justbe aware that you gotta gotta watch those costs on a daily basis when youdouble bubble with people would be mighty upset with. It's really it'sreally looking at the whole thing holistically, eso Like I said earlierthere, there's no easy button to do this, You know, you gotta you gottaknow what you're looking at, you know? And you've got to be a pull in theright data and apply different algorithms to get the right. Um, I'llsay the accurate recommendations, you know, based on on actual usage, youknow, And then once you make the recommendation, then you kind of gotthe the whole issue of getting the customers to go deploy them and makethose changes. You know that that's a...

...whole another issue because there's gotsometimes the recommendations. There's reasons why things are set certain ways,and, you know, those could be excluded kind of on the as you as you startdoing optimization reviews, you know, I'm on an ongoing basis, but yeah, theremediation part can be could be just a challenge. And what's their identified?Hey, when we when we met with Willie last year, we talked about this alittle bit to quote the episode number. But you're your perspective, Jeff.What's the single first place to start? You know, obviously we have the pillarshere or what have you, But if you're if you're just new to this obviouslyoverwhelming right, it could be overwhelming potentially. But where'swhere some small winds media gains that you would pick on? I guess it zits justbeing familiar with with what you're gonna be building in the cloud, right?So if you can get so you've got a VM our environment and a data center, moreaccurate information you can pull out of that in terms of actual usage. Thatwould probably be a a good start. And a lot of this like I mean, from whatwe've seen, that a lot of customers won't even get into optimization untilthey're already migrated. But the sooner you can start thinking aboutthese different things and do a little research, see what's gonna make themost sense the better. And this is how it kind of ties into at modernization.So maybe maybe you're, you know, 100 VMS could run better and again,containers are on service. So it's kind of looking at that application andidentifying potential candidates. Thio, you know, take advantage of some ofthose higher and cloud services s. So the sooner you could do that, thebetter. You know? Then I would say from from a lot of customer standpoint, youknow, they just want to get out of their data center. They want to get ourthey have to get out of Davis and a lot of them are hybrid. You know, maybethey wanna build another environment, but they don't want to go, You know, byanother 10 rat cage and all this additional power network Thio power it.So they want to go build it in the cloud. Right? So it was kind of havingthat mindset, knowing what their applications actually require, howflexible they are. That could go a long way. And I think also often what we'reseeing is there maybe five environments for a particular application, only onebeing production and they're not working the whole 168 hours a week. Ithink that's the number of hours in a week so that auto parking is anotherone. I find that yeah, we put out there so that you're only running at about24% of the time. I mean, you talk about that's a 75% savings. If you only havethat on this production environments. Yeah, from experience. That's thebiggest mind shift really to get around. People said, I know what you'rereferencing before, and you and I have intimate knowledge of that situationand to get there even from So you had you referenced the backup. But evenfrom a Q A dev test environment of where you don't need that fullonslaught of you know, that the calvary of what you had on Prem It's a greatexample of where you could save a lot of a tremendous amount of money. Yeah,Yeah, that. Make sure those back up in data retention policies to on those nonproduction environments. You need some, of course, for compliance and otherthings, but you don't necessarily need the same level. So making sure that youcreate those, uh, different categories of data retention I think are verystrategic as well. Let me let me ask interesting perspective here becausethis is something that last week I was challenged with as well. If a companyis looking customer client, what have you is out there looking to start thisinitiative? Where do you guys think it fits in their environment? Is this likean initiative of a c o E. When they start? That is that the ops team is ITMor from a you know, the Dev side of the house. How do you think it getsingrained into the culture or to start there? Yeah. I don't know if it hitsanyone specifically, you know, because I think each each team is gonna havedifferent requirements and drive additional e wants additional, butdifferent types of costs potentially.

Um, you know, like your your APP teamcould take mawr advantage of Dev ops types, tools, server, less things that,you know, since they're constantly producing new code for things that willenable them to roll it out faster, you know, but yeah. So there wasn't, youknow, E. But I would say Skippy touched on C c. O. And we've talked about thatbefore a class center of excellence, which is a group of stakeholders acrossthe organization. And it should include somebody from finance. Right? And Ithink if you've got if you have somebody from the financial aspect fromday one in your cloud strategy sessions, they're going to keep asking, whatabout the cost, And then it's gonna be everybody's job because they know thatquestions coming right. They gotta answer it. And plus, now with thatvisibility that you've never had before Like, I know how much my applicationcosts now, you never knew you kind of knew, but you didn't really know. Now,you know, probably too much, right? Right on. Well, this is great. So, Jeff,before we get going, is there anything else that you wanted to share with us?Yeah, there was one. And I guess it's more kind of Okay, well, what's nextfrom optimization perspective, right. That's that's kind of, you know, whatdo you What do you look Thio Add to the service or where is it going? Probablythe next, you know, year so and it's going to get into the application space.So I guess the idea would be to kind of take these principles that are in usetoday around optimization and apply them higher up stack because, you know,infrastructure, you know, rightsizing, auto parking, that that's not goingaway anytime soon. That's always gonna be a piece of it. Let's say you'redoing a recommendation session with a customer. You make theserecommendations, they go make changes. The next time you get on there, they'regonna say, OK, what's next? You know how you save me X amount of money onthis last one? What? You know, we haven't really changed anything oradded anything, our environments fairly static. What you know, How can you Howcan you save us money on this? Go around. So it's kind of taking thoseprinciples and and looking mawr into the application layer. There's a coupleof products, like any kind of a PM type of products that will give youvisibility into the application flow and identify elements in theirbottlenecks that if you could either re factor the app, potentially move it tocontainers or service, you're going to see X amount of additional performanceon that particular application. It might not save you on the public cloudside, you know. But if you can speed up your transaction process, say, forexample, and you could now sell 10 more widgets per second, then you couldbefore you know your your profits are going to go up. Not not on the cloudside necessarily. Think your spin your profits would not because you're notmaking money on the cloud. Yeah, you would see it on the other side, right?So you could it affect your bottom line? So that's, um, that's kind of wherewe're looking to go with, kind of how optimization will morph into additionalpillars specific to app modernization. Very salient point. Sounds like a nextpodcast. Thio. Yeah, absolutely. Yeah, that's where today we talked about howto save money in the bottom line, but yes, but once you take these principleson to the next phase, you're adding, hopefully money to the top line. Sothat's what that's what we all strive for. Well, thanks, Jeff. I reallyappreciate your time. Skip as usual. Thanks for joining us, Jeff. Enjoyingabsolutely thanks everybody for listening, please. Email is that cloudcrunch at second watch dot com. We welcome your ideas, suggestions andcomments. Thank you again.

You've been listening to Cloud Crunchwith Ian Willoughby and Skip Very. For more information, check out the blogged.Second watch dot com slash company slash vlog or reach out to second watchon Twitter.

In-Stream Audio Search

NEW

Search across all episodes within this podcast

Episodes (30)