Leaders in Hybrid Cloud - CTO Advisor Episode 108

Machine-generated from the episode audio and not hand-corrected, so names and technical terms may be imperfect. The audio is authoritative.

Hey, welcome to episode 107 of the CTO Advisor podcast, the first podcast of 2020, so happy new year to everyone. I have another great guest. We've been pretty fortunate to get practitioners on the line for the past couple of episodes and this tradition starts, I mean, continues with a buddy of mine, Joe D'Antonio. How's it going? Good. Good, Keith. It's interesting you mentioned 2020. We saw our first 2020 bug with a customer yesterday. Oh, the welcome to, welcome to after the break.

Yeah, they were doing some database code stuff that was really bad and it turned out when they were rendering the year 2020, they were only using two digits of a year, it was bad. So I saw a post that recommended that if you sign and date a legal document, you should not abbreviate 2020 as 20 because someone can easily come behind that and put whatever year they want, 2018, 2017, and I thought, you know what, that's not bad advice. Social media, you did not fail me.

Facebook, you did not fail me this time. Yeah, seriously. So Joey, tell the audience, what's your background? Yeah, I'm a principal consultant at Denny Cherian Associates Consulting, which means I do stuff for money basically, but to give you a little bit of history of what I've done, I was, my last real job before I started consulting was I was principal consultant, or excuse me, principal architect for SQL Server and big data at Comcast. So I worked across the infrastructure stack, we built a private cloud, we did a bunch of cool stuff, analyzing hardware and software to try to improve our really large scale environment at the cable company.

Before that, I did a lot of database work in the healthcare industry. And I've also, I've also worked with cloud centrally in the beginning, I was pretty early on in working with Azure and a little bit of Amazon. And fun fact, I interviewed for a role at Comcast, this is a few years ago, outdated by saying that I was in the field for or in the running for managing the OpenStack team at Comcast. Interesting enough, that whole team wholesale a year later, was at another company, the whole team.

And they and when I talked to them, they love their job at Comcast. This is just to highlight kind of the the the talent shortage when it comes to building private clouds and how valuable that Comcast experience was. Yeah, it was interesting. So we worked with that wasn't my team, but we worked with that team a lot. And they were, they helped us build a private cloud. And we built it, it was, I still call it the world's most expensive cloud, because it was on it was on Cisco UCS running VMware with with EMC VMAX storage behind it.

It was good. But man, it was expensive. But yeah, they quickly, one of the problems was that at the time we built it, which was like 2012 or 2013, OpenStack didn't really support Windows. And eventually that happened. And that made that more feasible. Wow, that was a good inside baseball and why my Comcast bill is so high. So, Joy, the last time I saw you, we were at we did a podcast together on on Tech Field Day, we talked about Kubernetes and whether or not Kubernetes is a real thing and whether or not practitioners should go about getting certified or trained in it.

But then I ran into you at Microsoft Ignite. How was Ignite for you? Ignite was good. It's a it's a big show. It's a little crazy. And I kind of both attended I was media and I also, as an MVP, have some booth duty. So I was also working the SQL Server booth for Microsoft. So I got to talk to a lot of customers as well. So I've called 2019 the the year of hybridity, the year that all of the major cloud providers kind of acknowledge, if not multi cloud, at least hybrid cloud, and we saw a lot of that at Microsoft.

AWS came out with Outpost. Microsoft has kind of led the major cloud providers and hybrid cloud, which included Azure Stack and Microsoft announced a pretty nifty capability of replicating Azure Stack up to AWS. And then Google has Anthos. And one of the things that I wanted to talk to you about was, you know, multi cloud. And, you know, I've had this whole stance on multi cloud from a control plane isn't a real thing. It's a good question. So I was recently in Europe, I was speaking in Slovenia, and I was talking to a bunch of my friends over there.

And one of the topics that came up, and I also talked to a friend of mine who works for a large consultancy in the UK, is that I think in the UK, and I know for sure Switzerland, if you're in the financial sector, and you're in one public cloud, you are required to be multi cloud, which I find really interesting. There's no requirement that if you have an on prem data center, you have to have a second on prem data center. But if you're in somebody's cloud, you have to be inherently in another cloud.

And I think, you know, as we define multi cloud, there's a few ways you can look at it, right. And the first being just having IaaS stuff, or infrastructure as a service, VMs, network storage, running in say AWS, Google and Microsoft. That's the most rudimentary way you can do it. You don't have that single control plane. I think one of the interesting approaches, and it's something Microsoft kind of had a launch of at Ignite, with a product called ARC, that ironically runs on Kubernetes, where they're going to be able to present you, and the product's really limited as to what it can do right now, but they're presenting you effectively the Azure data platform control plane.

So Azure SQL database, Azure Postgres database. They're presenting you the ability to run that control plane anywhere, whether it be on-prem, whether it be in AWS or Google Cloud, possibly even Oracle Cloud. Anywhere where you can run Kubernetes, you can run their control plane. And I find that, for a couple of reasons, really interesting. First of all, from a Microsoft perspective, it's a big divergence. Everything that had been developed for Azure before had been developed using Service Fabric Mesh, which is Microsoft's open source proprietary mesh networking software that they built to build Azure.

And secondly, it gives the customer ability to run anywhere they like, and getting the same experience. There's still not a whole lot of detail about how much it's going to cost and what's the pricing model going to be, but I think it's a compelling solution to give the notion of a single control plane, the ability to run platform as a service offerings in a different cloud. I think that's something, I think that system design, whether Arc sticks or whether it's the product that dies pretty quickly, I think the concepts they're introducing there are pretty interesting for customers that either are forced to go multi-cloud or want to go multi-cloud for various business reasons.

So, first off, the regulatory standard or the regulatory direction that you have to be a multiple clouds if you're in the financial industry, but not having an equivalent requirement for on-prem data centers. I have to say as so you're like that is it doesn't surprise me at all for a European regulation. But what I'm really intrigued about is this arc, because, you know, Ignite was so big that it was really hard to capture a lot of the nuance in some of these announcements.

So I spoke to some of the product guys for the Azure Stack team and we talked about Arc a little bit from their purview. And if I remember correctly, the primary benefit of Arc, at least I walked away with for Azure Stack, was that was the platform that they were going to use to replicate data up to AWS and your multi-cloud experience. So since you seem like you're a little bit more aware of this topic than I am from a control plane, when you say that I can run my control plane in multiple clouds, are you basically saying I can take what's effectively the Azure, lack of a better term, the Azure dashboard and run that in AWS or Google Compute or some other cloud?

Your dashboard, what I don't remember is what components, if you have to have anything running in Azure or not, I'm not 100 percent sure on that, but you do have the Azure experience for the portal, the PowerShell command lines, the CLI, the API, you have that experience for resources that are deployed either on-prem or in another cloud. So effectively, let's go back to the banking and the financial services example, how would that solve the problem for those entities? Right. So my thought, if you're having to go multi-cloud and your choice is to go Azure and another cloud, right, to do that, and let's face it, Microsoft does have a strong edge in the enterprise space specifically, you might elect to use Azure as your primary portal and run everything through that.

I mean, obviously, this product is a long way to go, but in order to support that. But notionally, that most functionality would be supported. You'd run everything in Azure, even though you'd have resources sitting in an AWS data center and it would be presented to you as though it were an Azure resource, even though it's sitting on top of, you know, Amazon Kubernetes. So what's really interesting and we'll take a look at, I think, the other leader in this space, I think.

Microsoft is probably the leader in multi-cloud or at least hybrid cloud capabilities with Azure Stack because they've had such a head start at being at least consistent in what they think the world view of hybrid cloud should look like. That's pretty solid, you know, in theory, in theory, I'm not saying to go out and do it. But I mean, I think I think Stack was I don't want to say it was a failure. Like they're selling it still. Right. Right.

Stack is pretty compromised in terms of the products it offers. It's fairly limited. And the other the other issue I think with Stack that I don't like as opposed to this is Stack was a, you know, a hardware appliance offering. So you had to you had to buy it through a hardware vendor and all the stuff that buying an appliance through a hardware vendor entails, you know, those extra support costs that you wouldn't otherwise have. And I think that was the notion of building this on on an open platform.

And I don't want to I don't want to hype too much that it's on Kubernetes because I don't think that's a big deal. I think it's just a platform that's available everywhere. And that's why they decided to go in that direction. And that's where I was going with this is that what I like about VMware strategy is that it's vSphere and I can manage vSphere, you know, kind of anywhere. I have some portability, even though vSphere itself is limited when it comes to cloud native capabilities inherently.

Tanzu is a different view of being able to do what Microsoft is doing. But what I like about the Microsoft approach and what I dislike about VMware's approach is that I can go to Microsoft and go to Azure and have this experience across multiple clouds. And at least the control plane is all one control plane managed by Microsoft versus VMware. Yes, I can have vSphere everywhere, but the control plane is is is is is bifurcated. There's not a single central control plane unless I created it myself and having my vSphere across my vSphere cluster across multiple environments.

I have to do that myself. I'm not I'm not consuming that as a service. Yeah, for sure. And I mean, I think the other part of it that I find kind of interesting is Microsoft is starting this arc from platforms as a service from the ground up, as opposed to starting with IaaS and rolling out because like Azure Stack started mainly with IaaS and some of the platform services rolled onto it. I think it's I mean, and I'll share a little bit of inside of information that I have.

This was something that was developed from within the data platform team more than, say, the Azure Stack team. So we saw that development come from data platforms. So it was definitely coming more from a platform as a service mindset than an infrastructure as a service mindset, which I also find kind of interesting. You know what? You brought up a really interesting point with the birth of where this is coming from. I always thought in at least the beginning of twenty nineteen up to about the middle of twenty nineteen until the industry mark marketing buzz around workload management overwashed me.

You start with data, you start your multi cloud journey with data. And I thought NetApp had the best story when it came to multi cloud data management, but they're very weak in the workload management piece of it. VMware and Microsoft much better at the workload management pieces of it, but not as strong in the data management components of it. I like that this was born of the data team. Yeah, yeah. I mean, the SQL Server team has been doing some really, really interesting stuff that goes far beyond databases in recent years.

And I think. I one of the things that is so I've been working with SQL Server and frankly, Oracle for about 20 years, I think one of the things that I one of the most interesting evolutions I've seen take place at Microsoft is just how that team has evolved, especially since. So this makes me feel old, but SQL Azure, their platform as a service database offering has been running for I think it's close to 10 years now. And they've learned a lot that that development team has learned a lot in the fact that they've had to manage, you know, they've effectively had to be DBAs and they've learned a lot about how things work and how to operationalize things.

And I think that's given them some insight that some other teams and even other firms don't necessarily have. Yeah, I was on a line with Intel the other day, and I'm surprised that not just the Azure offering. But also what they're doing with in-memory databases, with Azure and the extension of Optane memory to all the way down to SQL standard, which I didn't expect. So, you know, maybe 20 years ago and I'd hate to date myself. But, you know, when you're looking at SQL Server 2000, which is probably the first version of SQL that I ever used.

Not just the you can tell not just the mindset around databases, but the mindset around solving data problems has really evolved at Microsoft. Yeah, and ultimately, like you say, that's the biggest multi-cloud problem, right? It's super easy for me to deploy a web service to Amazon, Google and Azure at the same time and put it behind a load balancer. But keeping the data in sync in those multi-cloud scenarios is really the challenge. And I think I think that's why this is such an interesting piece, because it potentially gives the ability to do that.

But like at the same time, we also have our old friend, the egress cost that can make keeping data in sync multi-cloud rather expensive. But, you know, if your business has that requirement and you have the money to do it, then you're going to do it. And I think ultimately that's going to be a problem that the market works out. I think the egress costing will be something that customers just demand become a reasonable cost. And I don't and I think Google and Microsoft, as the smaller cloud providers, will probably innovate around pricing when it comes to egress charges.

Yeah, for sure. I mean, I think it's a place where you have the potential to have a competitive advantage. Right. And that's that's always going to drive price reductions. So let's look towards the future and let's bring the conversation back to Arc. I get the high level concept that. You know, if you're not going to come to an Azure data center, we'll bring the Azure data center to you. Microsoft tried in Azure stack. For a set of customers, I think Azure stack solves a problem, but at scale, if I don't want to run hardware, if I don't want these OEM relationships, if I want to get out of the data center business, but I need a consistent experience across platforms, Arc is the way.

What do you what do you think the killer feature product set or service Arc can provide to help Microsoft really hit the acceleration on Arc development and uptake? It still pains me to say this. VMS. Yeah, I mean, one interesting thing I'd really love to see and we'll never get it shared is just what percentage of Amazon and Microsoft's businesses infrastructure as a service versus platform as a service. And I just know if a customer is icy, it leans probably towards 80, 85 percent infrastructure as a service.

They love my and my customers tend to be forward thinking and they love to take advantage of this platform as a service offerings when they can. There's just a lot of legacy software in the world. You're not running SAP in a platform as a service offering. You know, you're not running your financial software and platform as a service. So I think if they can get VMS to that platform pretty quickly, it's probably a really killer play is that then then they can meet the needs probably more so than just about anything else.

So it's funny that you say that and I have to I wrote it down in my notes from Ignite and I have to find it because I thought the stat was so interesting. It was something along the lines of 85 to 90 percent of the workloads in Azure are VMS. Yeah, that doesn't surprise me at all. Yeah, I thought when I heard that I would have I would have guessed 70 to 80 percent, but it was much higher than the 70, 80 percent that I had guessed.

So it doesn't surprise me that you would say, yes, VMS. It doesn't surprise me that cloud providers have embraced VMware when it comes to running vSphere in the cloud. Most of the challenges you've done in most of the challenges that I've seen in cloud migrations, and I think you've done enough of these two to see that the same challenge is that. A lot of the cost is from not being able to take advantage of the elasticity of cloud. I use fixed resources in my private data center and I provision storage and that model just doesn't work well or doesn't translate well from a cost perspective in public cloud.

So it's interesting. I think just we've had both had similar roles in the past. I think storage cost is actually lower in the cloud universally, even even allocating full store over allocating storage to VMS to get performance. I think you still end up spending less money on storage, but the compute is way more expensive. And if you're not in a situation where you can take advantage of, like you say, the elasticity, whether it's spinning things up and down or adjusting resources or or using auto scaling and platform as a service offerings, that's where you're going to spend your money in the cloud and ultimately you weigh the expenses.

And frankly, I think a lot of times it's a wash in terms of running your own data center versus paying more for the compute. But yeah, VMS are still so prevalent just because not everybody's a startup and is building a new application stack, right? They have to run software that runs their business and that software runs on Windows or runs on Linux. It's not easily portable to a platform as a service offering. Yes, ironic. I call the platform as a service the future.

Ten years ago when Microsoft released its platform as a service offering, when it entered the cloud space for going IaaS and IaaS is still primarily the the driver. And to just validate something you said around storage cost for lift and shift migrations, it also has been my experience that I think the last analysis I did for a customer, I didn't even bother to do storage. We mainly focused on compute optimization and saying, you know what, if you're going to do a migration, get your CPU and RAM counts in order and kind of storage would take care of itself.

It was, you know, where we ended. Yeah, I mean, so fast storage in Azure is roughly one hundred and thirty bucks a terabyte a month. When I was at Comcast, and this is a few years ago now, we got really good pricing on EMC arrays. We did reverse auctions and stuff. We were paying about eighteen hundred a terabyte. I was going to say about that's about right. Yeah. Yeah. And that's about that's really an annualized cost when you factor in vendor support.

And that's good pricing. So if you're a smaller customer, it's going to be a pretty big difference. All right, so let's end a note on not necessarily predictions, but what are you looking forward to in 2020? So I'm on a long term contract where I'm commuting between Philadelphia and Seattle, and that ends in June. So I'm looking forward to not being on a plane as much. But other than that, technology wise, what I'm looking forward to is I think I want to see where things kind of land.

I think we're seeing a lot of dynamic, dynamic things in both the cloud space and, you know, kind of the open open source platform space. I think one of the things that a lot of people are getting confused with, and there's a lot of development that's getting pushed by Microsoft and Amazon because it's good for their clouds that might not be good for everybody. And I want to say that's kind of where Kubernetes falls in. It might not be the right thing for every enterprise, but you're seeing a big push because it is the right thing for running a cloud for sure.

So I think that's all going to start to settle down. And the other thing I'd like to see that I think could happen this year maybe is a company that spins up to give enterprise level Kubernetes support. And I know Red Hat sort of has that with OpenShift, but the whole IBM merger kind of, I think, put a damper on that. So I'm curious to see, not that it puts a damper on it, you know, just customers and IBM, you know how that goes.

Yeah, yeah, I know how that is. I think there's a ripe opportunity in that space, and I think it would be interesting to see what happens there. All right, so that's it for this episode of the CTO Advisor. Joey, where can people find you online? com or you can find me on Twitter at jdanton, J-D-A-N-T-O-N. com. You can find me on the web, CTO Advisor on Twitter. Make sure to like and share the podcast with your friends. Talk to you next, CTO Advisor podcast.