 Everyone thank you very much for coming to this session. So so my name's Sebastian. I'm the founder and CEO of scaler and scalers this Abstraction layer, it's called a cloud management platform And it allows you to manage public and private clouds and when when we've been building this platform We've seen a lot of people that have been Building hybrid clouds and so we've I've been exposed to a lot of Myths of or false expectations of what hybrid cloud means so so we put together a really nice panel for you today with some industry leaders that have been have Real experience building hybrid clouds and with that I'd like to introduce you to Anthony Skinner Who's the CTO at Moz? And has been has migrated workloads from Amazon to open stack and will be able to contribute his real-world experience there Do you want to add a little thing anything to that note? That was perfect. That's perfect. All right after that I'd like to introduce you to Mark Williams. Mark Williams is the currently leading the application modernization practice at Redapt prior to that he was the very first operations engineer at Zynga and grew their Their practice their cloud usage on Amazon and then their might subsequent migration to cloud stack And very last I'd like to introduce you oh anything to add good. All right We'll get to the details later And then I'd like to introduce you to Randy bias Randy's bias is the founder and CTO and CEO all three of cloud scaling that was recently acquired by EMC And has a lot of experience around building API compatibility for open stack and getting that to work well with with public cloud providers So the very first question that I have for our distinguished panel is Around application portability So let's say that I build an application and I won't be able to move that from one provider to another is that easy How what do I need to look out for? Well, it's just drag-and-drop right so All right next question I mean, there's certainly a lot of marketing out there that leads you to believe that it is as simple as it sounds and in some ways When I was at Zynga my colleague and I were often professing kind of what we were doing it with hybrid cloud and very much So it's kind of more like teenage sex. Everybody's talking about it very few people were doing it in reality for us as a gaming company It was very hard even Considering our applications were already cloud native so, you know when you can especially as an operations person when you care about uptime and You know the fluidity of moving games between clouds is in our case games between clouds was very hard It would take us about seven weeks Once we got good at it, you know the first few tenants were very challenging to kind of a get comfortable with moving into a Private cloud There's there's often trust issues and fear, you know, everybody's happy Especially if you already if you most of you are already consuming a lot of public cloud It's hard to get your your tenants to move so there's some tricks to to doing that, but it does take a while Randy Sorry, I'm glad you asked about application portability because I feel like that's at the At the center of what people are asking for when they ask for hybrid cloud Like I you know what I see is that that's the hidden assumption like I I've got my folks So I'm moving over to a devops process and doing continuous integration to continuous deployment and I just want to figure out You know kind of a Build and release cycle for my application or application back in that makes sense And I wanted to work on this cloud or that cloud and maybe you're doing dev test and QA inside and maybe you're doing an outside Maybe it's production inside maybe it's production outside people kind of want to mix and match But then they're hoping that it just works they're hoping for the sort of magic bullet and then I think the problem is that is That there really is no magic bullet like clouds have to be built with intent if they're not built with intent Then what happens is you wind up where one cloud has dramatically different performance characteristics or dramatically different cost from the other one And then you don't get what you hope for which is that application portability where it just works You're you also you also talked a lot about repatriation of workloads Does that like is that something you have to like specifically art Architect for or is that what you're touching upon? Yeah I mean repatriation is just a term that we came up with with folks who go out to public cloud They start there and then they want to come back and you know they go through that repatriation process And you know that's where we got some of our early traction with companies like Ubisoft They're deploying games on Amazon then at some point the games flat line They're not growing anymore And so they want to bring them back and cost optimize them and they need to make sure that as they target You know from their development deployment systems from Amazon to their internal private cloud that you know The right thing happens when they push the button Anthony you have a lot of experience in repatriating workloads Can you talk about that? Yeah? No, absolutely. We do so at Moz. We were about a Run rate at Amazon between 700,000 up to close to eight to nine hundred thousand dollars a month And so at some point we started to pack made sense to do for us and As as they said before I think it's a lot harder than people think and I think even when you're inside of your own Cloud is private. I think you have to pay attention to the boxes that are coming in right so one release of a Dell box Isn't the same as the next right and same with Cisco We have both types of gear and so you have to test those ensure they run it because each one of those actually do different things and For us we actually rip those box down to be nothing but bare metal and ensure We're putting our stuff on it and even then that it's still very hard to actually do so it takes a bit of work And for us given our size we already had a DevOps team So managing that Amazon and managing it on site didn't necessarily matter to us It was really how could we find a way for our developers to push your button and deploy regardless of where it was So they can just provision work Cool. So are there any other myths or things that you hear a lot about in the press that that are your pet peeves some things that you That are very incorrect that that That you'd like to clarify I mean one thing that kind of drives me a little nutty is when people in the enterprise vendors I guess I'm an enterprise vendor The a lot of the sort of traditional vendors they have a They get really sort of They come up with these ideas that don't really make a lot of sense like I'm gonna take an Ethernet segment like a layer to Ethernet segment I'm a stretch it across the wide area network and you know put my databases over here Oh, this other stuff in the cloud without any like, you know taking in the count that like hey Latencies latency the speed of light speed light, you know, it's not going away And there's just sort of this pretend now I've got you know hybrid cloud because I got the database in my private cloud and I got the web app servers here But you know, that's never gonna work There is a case for that layer to Ethernet bridging thing, but it's just it's it's really a niche thing Yeah, one my pet peeves is cloud bursting and that that's kind of that like let I've got my databases here But that's burst out to the public cloud for the spike and the problem is that you're paying for bandwidth between public and private Cloud and like what you're the few cents you're saving by putting your your your extra compute on on Amazon You're getting hit by the amount of traffic that has to go between your database cheers and applications I was gonna pick that one, too. I've heard of it. I heard that referred to as spillover spillover. Yeah Wow spillover. It sounds like an accident And again, you know, you mentioned, you know, I have seen a couple companies get into a situation where they've had to put Something that was workload intensive in a either cloud or non-cloud bare metal situation to get around, you know Surprises in the cloud which are good problems to have And so they've been kind of left with that that situation of effectively being hybrid, but even the term hybrid to me We we at Zynga did actually do one application that was split across Both our private cloud and our public cloud just to say okay, we know how to do this in case We really run out of capacity, but for us that the more appropriate term that we Centered around was multi-cloud so you want to have those same tools those same front-ends to be able to operate both your private and your public in the same way and To enable your developers and we are very decentralized operations organization. So enable everybody to operate both of those equally but When while we were doing this experiment with this one app where it was in two clouds It was we just realized how insane it was there are so many moving parts You have so many more things that can go wrong and it's fundamentally more complex than it needs to be so for us It was multi-cloud put workloads that are known in one different workloads in another So so I guess like another myth would be that like workloads need to be either public or private But you don't have workloads that span to two providers is that what if you care about simplicity and if you care about scale and you care about Eliminating things that when there's an anomaly, which especially in public clouds There's anomalies more often than not because you don't know what else is happening. You don't have that full transparency Again as operations people you want to know what that anomaly is you want to have as few things to eliminate as possible Yeah, no, I know I agree with that as well for us bursting isn't bursting from one cloud to the other It's really what applications are running where and then being able to add capacity and so I totally agree So so did you have a workload that was the single workload you had was that did that span Amazon during that transition period? No, absolutely not We have really one running on one place or one running the other and then we also have backups We also have backups that are running in two different places, right? So but they don't really span two things so first is I Just want to caution us all to not paint with to broad a brush right if you say something like, you know Spanning workloads cross-class doesn't make sense it Kind of pens on the workload because if you walk away from compute having an object storage in two clouds makes a lot of sense and do I mean, that's a very good one You know you could if you have your system your stored data record in Amazon s3 And you want to have it in Google as well So in case you need to do a disaster recovery fail over and that's that's a great application, right? And that's what we do right and then the second thing is is that you know bursting could work But it's kind of like I said you have to kind of build with intent So you would have to put your private cloud very very close to the public cloud like you know and have like Metro ethernet very low Latency very high man with for it to actually make sense and again It kind of comes down to the question of do you really have an application these that are can you plan ahead so that you can Have that parts in different clouds and then as Anthony was saying kind of expand each one You know scale each one independently To back up to your original question as well Some of the myths that you often hear when people come talk to you if you're private or public or one of the other Is that as a as individuals we actually care actually? We don't really care what we care about is cost and stability and performance, right? Yeah, and as long as you have those things it doesn't necessarily matter where it runs And and your point mark was that if you have a workload that that's not really suited to be on both Then troubleshooting becomes very difficult and that that affects stability. Yeah, I'll reflect on cost as well You know it Especially for those of you who are big public cloud consumers and you're aspiring or beginning to build those private clouds Again getting that first tenant to want to move in is tough But ultimately you're accountable to the CFO and the fact that you've invested so much time and capital in that private cloud Will eventually force the right thing to happen which is to get your tenants moved into it? That kind of thread alone may not be enough and for those of you who are working to get chargeback or showback That's actually the tool where you know if you if you demonstrate the cost savings that you've invested in that in that private Cloud compared to your public cloud costs. That's that's kind of a carrot or hammer However, you want to look at it to get that change Yeah, absolutely for us the developers They were kicking and dragging not wanting to go And I think once they saw how easy it was and how much better the performance was once we got out and the overall cost They you know No one really wants to spin up a box at Amazon or else at this point It's really if you have to do it you'll go do it and I think for us One of the things we're still working out is the DevOps side of it So, you know, anyone can go push anything at Amazon and just have it run and cost us tons of money It's a little bit harder and internally right? Yeah only so many boxes. So we still have to manage that piece as well By the way, this is really a session for for you guys out here in the audience If you have any questions either raise your hand or go to a microphone, we'll be happy to take that So one thing I've seen a lot in the last Maybe last 12 months or so is that hybrid cloud really seems to mean in among practitioners I mean it seems to mean one public one privates It we don't really run into the use case. I run hybrid between Four different public cloud providers and two different private cloud providers like that that seems to be all Too prior it doesn't make as much sense, but but hybrid cloud I'm getting the died the the sense of the hyper cloud really means two clouds one public one private What do you see there and to I? See people who are doing their strategic planning aspire to want to do multiple public clouds to avoid the you know The lock-in and be able to potentially negotiate on price though my experience in seeing Negotiation attempts on public cloud providers on price is that's all commodity They don't want to deal with those special cases I mean you have to be the Netflix's or the Zynga's or or bigger to perhaps get special treatment there So I see companies aspiring to do that But but again due to the number of that domain knowledge that you have to acquire about what how your private cloud works and Operates and behaves same thing with each of the different public clouds That again adds to the complexity. So I think people end up getting more satisfied just with the two So is cloud brokerage a myth or well, so for us we were a little bit different So we we run probably at three of our own data centers And several other clouds right and largely that's because of the types of things we do right because if you were in search In our case in order to get serfs or search engine results around the world You kind of have to be around the world see some of those things you can't always get them from the US And so that then requires us to do so also you can imagine if you tell someone you stop spending 800 grand at Amazon A lot of people knock on your door and say hey, I'll take a piece of that so So so therefore we we have the ability to run in multiple places I Think it's true that that's starting to become the perception, but I think part of What's happening is that as a lot of enterprises wake up when they're starting to do their first private cloud like Whether it's the vendor or the IT teams They realize that there's already some traction in one place or another and they're trying to set it up So that the the folks who they're going to you know deliver to are kind of see it as one unified system That's sort of the dream scenario right like you know if we come in and bring this hybrid cloud for you Then you know you've got one single pane of glass you hear that all the time now that allows you to sort of like control and manage both And it's a way to try to get things back under control if you're an IT team that has been dealing with shadow IT And they're the line of business is already out on the Amazon web services You want them to repatriate and you have to give them a really good experience And so that I think that's where it comes from is it's that tug of war between like should it be external or should it be Internal and the answer is it depends and it's the right tool for the job type thing For mark you can tell us probably better than any of us because of the two sides You've been on what there's kind of an inflection point which you start looking at either war And what was it for you guys at seeing a and when do you see it now? There's a pretty long list that that Encourage us to want to move back to private cloud But so again my experience was like a hundred bare metal servers when I started a few thousand as we were running out of space And couldn't build it fast enough and then we exploded with Farmville and to Amazon got to very many tens of thousands of Instances there and then that experience it was a cost It was a performance because we had very much experience knowing how well our bare metal worked and it felt about half as performant in the cloud the lack of transparency in the public cloud the We had disproportionate experiences of the scale of outages that would impact us and again. We were running our business on this right so Those all translated into if we because of our scale we could get wholesale data center space We could make it a third the cost and twice as performant in private cloud So that was that's that's what we took to the CFO and the other actual inflection point too was that as a pre IPO company Investing in capital is actually There was interest in doing that from the CFO based on the valuation that that would kind of influence in the company So that was kind of the ultimate Catalyst there, but again we built this thing and then couldn't get anybody to move into it because everybody was still happy with it With dancing with the devil they knew then the one they didn't know so what it took was Actually, our first tenant was an acquisition So we forced them in there and again, that's that's the other experience is you build this thing for a specific workload This game company had a completely different Profile it was we were not no sequel in our public cloud and they were very much sequel shop But again, it worked and then ultimately the thing that worked is we found you know a willing participant Who then became our evangelist once they had moved there they demonstrated it was twice as performant It was a third the cost and they were happy, you know with with the transparency was there So the the question is what were some of the difficulties was it the culture, you know, what causes the resistance? In my case at Zynga, there really wasn't at this concept of legacy I didn't have to transform these applications to be cloud native They were already cloud native and the same tools could apply to both public and private the the challenge with with at least the gaming company was all of that engineering talent is there to develop new features and and things in the games to drive new revenue to ask them to Spend some amount of cycles to migrate their applications because we couldn't do that for them Each game was it was different enough to require their involvement And so we had to collaborate as dev ops really to do this and so that took away their ability to innovate And so we kind of had to start with the smaller less revenue generating games to make them more of an evangelist And then what worked later was the big games were so costly in Public cloud that they their P&L and we were doing charge back their P&L Would look much better once they moved because now they had data and again being data driven is another key is show you know doing the show back on the on the costs around things as well as Kind of having your developers Participate with you on how you're doing the infrastructure as well. We got a lot of great ideas about how to do our availability zones that really made them more Confident in what we were building for them. Yeah, I would agree with that So I think what you see is there's the there's the cost And then there's just the number of features and people want to build throughout the quarter Right, so if you have in our case, we started off small There's five 15 25 people right and you're all heads down and you're just coding You got maybe a hundred boxes all of a sudden one day you wake up and you have 150 people at the company and you have thousands of boxes, right? And so to tell people hey you're gonna stop you're gonna pull all your stuff off and you're not gonna develop new stuff And we're gonna do this because we're gonna save costs. They don't necessarily see those cost further What you see also is we have devs that are coming in for engineers that for them being in the cloud is kind of a cool thing Right, it's good on the resume. It looks good. I know how to manage it And so now you have to then tell them well great You're not doing that anymore You're gonna go do this other thing and you're no longer in control of it that dev ops people are and so there's a few hurdles You have to try to get over in order to get that done and one of the things It's really tricky is that you know developers walking into cloud land maybe with the only experience with the Amazon You know, they're just loving the fact that they have these API's that they can call They're sort of this relatively deterministic behavior. They know what they get when they do it And you know on Amazon and networks and networks and network mostly EBS is EBS is EBS mostly And but then when you go to try to hybridize and you're trying to move an application from one place to another they get They're not infrastructure people they have a disk in another place And it suddenly doesn't operate the same way or have the same performance or the networks aren't the same You know, then it gets very very confusing and you start to be asking these guys to can you build an abstraction layer that hides all of that? From your application and then it gets trickier and trickier and trickier because these guys want to focus on the app capabilities and feature function Not on how it talks to networks and storage and compute and those kinds of things And we just haven't really made sort of a common substrate so that you have a guarantee as you go from cloud to cloud that it's the same Yeah, not totally so we were Dell's first one of their first open stack implementations when they first started shipping of what three years ago and We spent maybe 120 days tracking down a sub-second response time And we just tore her out trying to figure out where that sub-second was and so you know those those sub-seconds matter And so that was four months of dev time. You just won't get back, right now that we have it It's a ton of savings. However, that took us four months to figure out Yeah, this this right with something Randy said made me think of this other kind of portability thing or the abstraction layers That's what a lot of the public cloud providers are trying to provide is those passes and those higher order functions that once you start consuming Those you're stuck like There may be private cloud equivalents, but they're gonna be nowhere near the same So the data gravity problem not more of a vendor lock-in proprietary protocol problem is another piece that that can lock you in Can you say that is there any? Is it the future to have hybrids hybrid cloud? Oh Yeah, well, I mean, I think Docker and Other types of things that allow you to deploy anywhere are definitely the future You know, we try to use those underpinnings for or something similar for everything we do So is it the future? Yes. I think it is I think being able to deploy anywhere is is Definitely something that we'd like to do I think if you look at open stack and what it provides or Docker is others You we wanted to have the ability to not only manage the platform But then also come up a level and manage the OS types of things that we have along with databases So therefore the devs can then just put application on Well, you know, and it doesn't really necessarily matter where it's going. It's a matter of provisioning at that point So yeah, it's a deployment. It's as a deployment and configuration tool It's awesome because you have a very controlled environments Now we if the question is what is it the future for everything? Probably not but for at least for deployment and configuration I've seen a lot of people stop using as much on the chefs and puppets and And some of the configuration management tools in order to use a to consolidate around a container type of deployment I mean the thing is is that I see some people think of Docker sort of like Completely encapsulating the application and isolating it from everything around it You know, that's basically impossible, right? I mean Docker does a little bit more than than that and then it'll abstract the network away and Makes the network a little bit easier, but that introduces its own problems, which I won't go into right now But you know an example of this is that if you had Docker running in two different places And there was a 10x difference in IOPS Like Docker's not going to do anything to give you, you know, it's not going to manage You know make your disk drives faster. It just can't do that. It's impossible, right? You know one thing that I always think is really funny is that and we build these abstractions to make our lives easier And we like to pretend that we can kind of hide things But the reality is is that as you move down the stack you get closer and closer to the physical reality And once you hit the physical reality the physical reality is what it is I mean if you're in a location like Japan you're gonna prioritize space over power in a data center Which people won't do here and that will feel its way up the stack because if you're prioritizing space over power Then you might go super super dense because you can put, you know 25 kva in a rack and that means that you're gonna use, you know potentially much higher speed processors So your cloud may wind up having a lot more compute, you know cycles than some other cloud And and so that's the reality is that you just can't get away from the physical Situation I mean no matter how much virtualization and abstractions you put in Look C groups is not gonna add more IOPS to your disk drive your disk drive has x IOPS It's got x IOPS you can't add y IOPS to x magically with Docker That's all you have if you go to another cloud or using Docker with the same application and it's got z IOPS It's got z IOPS 1,000 versus 100. That's just all there is to it I mean you can't fix that if you run Docker on one cloud and it's got one gig of ethernet coming out of the Boxes the Docker's running on and on another cloud. It's got 10 There's nothing you can docker cannot fix that problem now Docker does do nice things like all the containers on the box look like they have their own port 80 for example And then those are mapped out to you know different ports on the host box so that you actually get some Abstractions at the network level and sort of hide certain things But there's limitations to what you can do fundamentally Yeah, no, I mean I agree with you So for us as we move from place to place or talk to different vendors It's always same hardware same types of drives same Networks types and even down to network hardware because that's the only way you can actually measure that they're the same right? So if you tell me that you're gonna give me emc over net app or something like that Then I it's harder for me to manage that and actually know what's going to happen and predict So we always try to make it apples to apples It's the only way around it for us. It's the matter of deployment for the for the devs All right, there's another question. Yep So the question is how do you do live migration on the hybrid cloud and how do you keep the topology was less one consistent, okay Zynga's migrations were live, but they were not using the feature that you're talking about your migration So the way at least gaming companies commonly move their workloads is you duplicate the environment in Wherever it's going to end up you slave your sequel or your no sequel there So that it is concurrent with the activity that's going on you populate all of the front end you warm your cache and You quiesce the first one and you cut over to the other one In this way you have the ability to move back I would be worried about using something like a real feature for live migration if it even exists between clouds to To move forward and move back because if there is a problem where you end up now You need to move all that data back, right? So anything that changed once it was there where you're talking about VM live migration or just live application migration Yeah, I think I think there's there's I sense there's too much risk assuming it's a reasonably complex if it's one Server one instance, okay that make that there's a reason reasonable approach to that But my experience is moving Applications that are involving thousands or multiple thousands of distinct instances, so you can't do it that way I mean the thing is is that I think we're all assuming that you're operating cloud native applications The request that you have about like would you do live migration? You know wide area network migration of a VM, you know, that's a request That's you know really for non cloud native kind of legacy applications Like I need to move my SAP from on-prem, you know to off-prem and I need its network model to go with it That's that's a whole use case that you know, some people would argue that that's not even cloud SAP isn't cloud the stuff that's under it, you know as virtualization And so, you know, I think that you know from our perspective, you know That that's just doesn't make sense the question you're asking because people don't solve the problems in that way You know most of the times when you do wide area migration of VM It's just a one-way, you know thing you're doing. It's not something you would do dynamically, right? I mean, why would you move databases around like that that are many terabytes or petabytes? It doesn't make any sense and if you've got a cloud native application. It's a shared nothing architecture So you've got all these tools that like, you know autoscale your arrays, you know Whether you've got 10 app servers or 100 app servers and that same tool you just point at the new cloud And it spins up 10 app servers. It actually that takes much less time than trying to move 10 app servers over the wire And your second question One more question. Yeah, go ahead. One of my favorite articles about cloud was about Zynga and the articles suggested that you started new game development in Amazon then as the game developed and Stabilized and gathered popularity you would then shift it to your private cloud kind of getting the feeling that article is bullshit now It was true for a brief time because the reality of building and investing all that capital in a private cloud meant I had to fill it, right? We there was not any tolerance to continue to buy reserved instances or pay-on-demand fees But the time that was true was right as it was becoming available My customers our game studios were still nervous and not kind of clamoring to you know try my new mousetrap, right? So once they did every new game Launched in private cloud so long as I had predictable enough capacity and the thing that you know that that was became then the only thing I didn't want another Farmville to happen We were lucky actually that Farmville launched an EC2 the only reason it happened is that my data centers were full When the request came in we had been playing with it But we hadn't really done anything substantial So and at that time Farmville was the fastest thing that had ever grown ever It was it exceeded our expectations by at least two orders of magnitude So yes, but once we got stable and predictable with the behavior of our private cloud and there was enough capacity We'd always launched there correct. Yeah, I like that the good and the bad of Farmville going away as my Facebook is not updated But to your point I think that that is the case for most of us is that you end up starting there And you start growing and it's just cost I think one of the things we should point out is since then Amazon's probably cut their prices 25 30 times, you know the reservists is just this year have been cut You know 65% in there somewhere and so the cost is a little bit different now. So now it's a matter of Not only just cost but reliability right and then also within their SKUs often You end up over provisioning because you want X amount of memory or X amount of disk space Therefore that starts to cost you more money. So that leads me to one more one more myth that that that we've been here We heard about like maybe three or four years ago Which was public cloud is cheaper and of course that depends a lot on the workload It depends a lot on how you're gonna build private cloud. It depends on many things. What are your opinions there? What's what's the least expensive and What types of workloads work best where yeah for us it depends There are some things that we still you could we use Amazon for just because it makes more sense, right? So we have thousands upon thousands of micros at Amazon and those are types of things I don't want to build out myself and they're all over the world and that just makes more sense to do there and They're from their CDN standpoint. It's a lot easier to deliver closer to customers There it is for me to do it so we push our web apps some of our web apps and our website to Places at Amazon all the way around the world. So to make it faster. So those things makes more sense for us to do that way I don't know that there's a simple answer to that. I mean if you're gonna ignore a lot of performance variables You might be able to claim public cloud is cheaper But there's so many taxes and Some of those taxes are around just it's such a large environment that as soon as you're dealing with multiple nodes talking to each other Even within a single zone The overhead you're paying on the microseconds between Instances it's such a huge environment that again a private cloud is gonna be smaller tighter more performant just given it's it's relative smaller size The nickeling and diming on paying for provisioned IOPS paying for all data in and out Some of those things depending on kind of what you're expecting to pump through your your infrastructure Those things can make it more expensive Um You kind of have to be careful and you got to make sure that you're measuring because if you I've seen Enterprise shops that are extremely dysfunctional from an IT point of view and Amazon and public cloud is almost always cheaper for them across the board there was a Big medical company on the West Coast in California that was paying $15 per terabyte for their storage of radiology images I mean, you know, that's a little ridiculous, right? But yeah, so I so I it really does depend also I've seen people who like they come at it and they're like well, you know We're gonna get a V block and we're gonna put it on and then if you go that route It's just and I and we sell V blocks. They're great But you know if you're trying to cut costs, that's not the solution that you would go to right and so You just see people who kind of aren't really thinking through those those conditions one customer We talked to is taking Hadoop and they're running it on Cisco UCSB series blades on top of a fiber channel sand Right, you're not gonna see any of the Hadoop's cost economics from that So, you know, I think it's a little it doesn't make sense to really talk about which is cheaper It's just it's very dependent on workload level sophistication of the IT team You know, how good your procurement systems are whether you've got the stomach for commodity gear, you know And as Anthony was saying sometimes it just depends on the workload Let's let's give the audience a couple seconds to come up with a question. Yeah, go ahead It's it's not like hybrid cloud is impossible. It just there's you really have to stop and think about and I mean, I'll repeat the question So the question is, you know, let's say you could kind of eliminate some of these variables like API inconsistency between providers Minimize kind of the variance and latency and some of these other kind of undesirable still evolving parts, right? Then would hybrid cloud kind of be the best way to go and again, you have to be very Pragmatic with it first first. I think you need to be very strong with can you do private cloud? Well, can you get that right get that right first? Before and unless unless you're forced into this to solve some, you know, crazy workload because your app just took off, right? Solve for private cloud first optimize that you're gonna be expected to use almost all of that that you can and then If you if you really are as sophisticated to be able to put your app a single app in two different places and ensure active active Then then great. Okay. Do that A lot of people have the approach. Oh, I want to have a pilot light over here Well, you better be practicing that pilot light, you know If you're gonna do private and then have this little piece ready to go you better practice and make sure that works There's some kind of realistic questions of doing that Yeah, no, I totally agree. So it not only it has to be active active, but I think it's also about scale So for us for one of our applications there We have over a million five hundred thousand shards, right? And you really can't do that in two places and it's even hard to back those that many shards up And so then when you start talking about moving petabytes of data in multiple places, you know Hundreds of petabytes that that becomes tough too. So it's really about being active in both places and be able to serve data from both places so If you get a chance you should take a look at the Presentation I did at the Atlanta summit where it's called open stack architected like AWS And when I am talking about, you know, you need to build your cloud with intent I mean, it's kind of goes to what mark is saying You can't just pretend that some of these things are gonna appear you actually have to make them appear and when When I did that presentation, you know, I really sat down very hard And I said what what does hybrid cloud mean? What does interoperability mean? Like what are the all the elements and I came up with six areas, you know performance availability cost You know behavioral compatibility API compatibility, you know and really tried to break it down And I think, you know, there if you go and you see that presentation You'll see there's a lot of thinking there about, you know What are all the aspects so that you can stack them up and you can measure between, you know The two clouds you're trying to hybridize and understand if you're actually going to get the experience that you expect But again, it's gonna have to be with intent and you know deliberating this One last point on that too is that behavioral compatibility, right? Again, one of the other reasons why Zynga didn't end up putting single apps across two clouds is There's never going to be identical performance out of a web instance in Private cloud versus an EC2 or a Google anywhere and for something again This is workload dependent for something that we care about that consumers experience playing the game We want it to be performant reliable and consistent when you have two different things that are serving that out You cannot guarantee that I think the one thing they leave with is in the hybrid When we talk about being hybrid it sounds like neither one of us did it with intent, right? We didn't intend to do that, right? It was something that was a byproduct because something else happened to us And so I think if you go into a planning and knowing and you have the the bill in the luxury to plan it out It's a whole different thing. So what are the problems that hybrid plan solves? For us it's now we can balance workloads and the types of things and do it efficiently And we can do it cost effectively and so for us It's a matter of now we can plan because we have the capacity internally and we have it externally I'm gonna call that multi-cloud multi-cloud solves. Yeah kind of the operational Disparity that you would otherwise you would assume when you have two different things Hybrid cloud to me means I'm running a single app and multiple clouds and again. There are cautionary statements for that Where it where that does come into reality though is you have that app that takes off surprise surprise I work with a customer whose back-end persistent data store couldn't operate successfully At scale in EC2 for them So they rushed to provision raw bare metal for their MongoDB back end and then they put that front end in Amazon continued that there and had the back end function for them So that for them was hybrid cloud although the private part was bare metal I'm Sorry, I'm a little punchy. I didn't get as much sleep as I wanted So, you know, one of the things that I think is really funny is that You know people are still getting a lot of value from public cloud right the reality of the situation is that a Lot of the time you don't know what's gonna happen and public cloud really reduces the cost of failure You know what I'm saying. I mean in the in the old days I'm sure people remember but in the late 90s if you wanted to launch a new SaaS service You you're gonna put you know millions of dollars of investment into like making it go and you look at somebody like Netflix all in on Amazon web services and you're like wow, you know, it's a little crazy. You could be running a cheaper operation Yeah, but then they just pointed over to Mia and they went bonk and Netflix is suddenly in Europe You know a few hours later like that's powerful, right? And so that reducing the cost of failure allowing you to have global reach. They're still really great use cases for a public cloud You should not be afraid of it and that's part of why I think hybrid matters. But again, you've you've got to do the practical realities No, no Go ahead So the question is what do we think of the the open-stack public clouds? I'll have a quick comment here Yeah federated open so one thing I noticed so again like this is the the project I build scalar Talks to all the different public cloud providers and working with the open-stack APIs. We've noticed that It seems that all the open-stack public cloud providers use different APIs So for example, we had we were forced to have different libraries Whether it were using open stack from trunk open stack from cloud scaling or open stack from rack space It that specifically of rack space like they're not working is different a whole bunch of things are different So so like when you when you do a quick glance at the rack space API is they seem they they use the same words As as open stack, but there's semantically. They're different Go ahead. I mean open stack Can't believe I'm gonna say this for like the fifth year running You know is not a cloud operating system. It's more like a cloud operating system kernel, right? And so people you know download it and then they build their own thing You know whether it's as a product or a bespoke cloud and when they do it's like taking the Linux kernel in building Android Which is running the Linux kernel and building Cray Linux, which is running links kernel and building rel Which is running the links kernel But I guarantee you have zero application portability between those and all the application Exactly are different, but it's still all Linux and it's all still the same code Exactly may even be the same version of Linux, but you've got no compatibility So right now the problem is in open stack public cloud land that we have lots and lots of snowflakes So federation is like very far away. I mean there's some there's some bright lights I don't know if this is the right form to talk about them I'm you know deaf core and ref stack and give us a way to sort of test interoperability Which may make this cleaner and then we can get to federation, but we're a long way away from it right now Yeah, that's a good point I There's a push from the consumer of the open set clouds to have interoperability and standardization The problem is is that open stack exists because it's got an inclusive attitude And it's it tried to have a big tent and have as many people around the table as possible with all those people around the Table they all want to take it a different direction And so you know if we continue to look at it sort of being a complete product Then you know we can't get it right But you know if you if you try to port an application from a rail box to a rail box It works if you try to port it from rail to suze it doesn't work So, you know, I think people need to be having their heads that what the right way to consume open stack is via product Yes, I have a product So you know take it with a grain of salt and you know, but you know raw open stack just leads you to a bunch of challenges I'd add to that for us the It's really not open stack to open stack for us is the swift stack So in order to let the devs just put data wherever they want and just have it work That's kind of the thing that works the same the open stack piece going from Rackspace to Amazon to others it really doesn't what happens is we just cut a ticket until it works exactly how we want it to We just sit there complaining and then eventually we get there and so, you know, and that takes time, right? I think it takes it took us about three to four weeks before we got to work at rack space We went over to Google. Of course, they work totally different, right? So Great question All right, I think we have a time for one more question. Who's the who's going to be the lucky person quick? All right, no lucky person. All right. Well, everyone. Thank you very much for for listening in on this panel Hope you got enough. I will stick around if you have got any questions. You can come up to us personally. Thank you very much