Episode #103: Differing Serverless Perspectives Between Cloud Providers with Mahdi Azarboon

May 31, 2021 • 51 minutes

On this episode, Jeremy chats with Mahdi Azarboon about why understanding different serverless perspectives is important, what challenges you'll face across providers, why you should take a more holistic approach when embracing serverless, and much more.

Watch this episode on YouTube:

About Mahdi Azarboon

Mahdi Aazarboon started working as a serverless specialist and evangelizing it through blog posts, conference talks and open source projects. He climbed up the corporate ladder, and currently works as Senior Manager - Cloud Presales at Cognizant. He helps big and traditional corporations to move into the cloud and improve their existing cloud environment. Having a hands-on background and currently working at the corporate level of cloud journeys, he has matured his overall understanding of serverless.

Linkedin: linkedin.com/in/azarboon/
Twitter: @m_azarboon

Watch this episode on YouTube: https://youtu.be/QG-N3hf1zqI

This episode sponsored by CBT Nuggets and Lumigo.


Jeremy: Hi, everyone. I'm Jeremy Daly, and this is Serverless Chats. Today, I'm joined by Mahdi Azarboon. Hey, Mahdi. Thanks for joining me.

Mahdi: Hi. Thanks for having me.

Jeremy: So, you are a senior manager for cloud pre-sales in the Nordic region for Cognizant. So, I'd love it if you could tell the listeners a little bit about yourself, your background, and what it is that you do at Cognizant?

Mahdi: Yeah. Just a little bit of background, I started as a full stack developer, then I joined Accenture as a serverless specialist, and over there I started to play with AWS Lambda specifically. Started to do some geeky stuff, writing blog posts, and speaking at conferences and so on. Then, I was developing several solutions for multiple corporations in Finland, then I joined another consultancy company, Eficode, which are known for DevOps. It is very good, they have a good reputation for that in Nordic region. I was as a practice lead, AWS practice lead driving their business. Then, I joined my current company, Cognizant, and here I work as a pre-sales capacity. I'm not hands-on anymore, but basically I do whatever is needed to make our customers happy and make them to go to the cloud. So that means high-level solutioning, talking with the customer and as a senior architect, I comment about stuff, I make diagrams, And I translate business and technical stuff requirements, basically as an interface between the delivery and the customer side. Yeah, that's all.

Jeremy: Right. Awesome. All right. Well, so you mentioned in some of the blog posts that you were writing and some of that was a little while ago. And it's actually, I think there is some interesting perspective there. So I want to get into that in a little while, but I want to start by this idea or this post that you wrote about sort of what you need to know about Azure functions versus AWS Lambda and vice versa and it was sort of this lead-in to this concept of multi-cloud and not cloud-agnostic like being able to run the same workloads, but being able to understand the differences or maybe some of the nuances in Azure versus AWS and of course, that got extended to GCP and IBM cloud and some of these other things. But I'm curious why understanding different serverless services or different cloud services across clouds in this multi-cloud world we are living in now, why is that so important?

Mahdi: Yeah. That's a good question. First of all, I would like to clarify that whatever I'm telling in this podcast is just my personal opinion and doesn't reflect my employer. This is just to save myself.

Jeremy: Absolutely. Like a standard Twitter handle route.

Mahdi: Yeah.

Jeremy: Views are my own, right? Yeah.

Mahdi: I don't want to answer to my boss after this podcast. Answering to your question, the thing is that multi-cloud is inevitable and even AWS which was ... In the best practices, I remember like a few years ago, they were saying that, no, try to avoid that. They started to even admitting through their offerings that they are trying to embracing that multi-cloud with their Kubernetes offerings. The thing is that, well, whether AWS fans like it or not, Azure is gaining a lot of market share and it depends on the country. For example, in Finland at least AWS is really popular. But now I'm dealing, for example, in other countries like Norway or UK, Azure is very popular. I mean, you can just exclude yourself to be only with one cloud, but in my opinion, you are missing a lot of opportunities, both to learn and just as a company to embrace the capacities, because whether ...

Well, Azure provides some stuff which are better than AWS. I mean, I heard from a corporation that they really like AI capabilities of Azure much better than AWS and they do a lot of analytics. So it's inevitable whether many people like to admit it or not.

Jeremy: Right. Right. But so even the fact that it's inevitable and we talk about, multi-cloud is one of those terms ... I just talked to Rob Sutter about multi-cloud a couple of episodes ago and it's so expansive. I mean, everything from SaaS providers to, obviously the public cloud providers, to maybe even on-prem cloud, I know that sounds weird, but like your hybrid cloud and things like that. So the problem is that there are a lot of providers, there are a lot of SaaS products, things like that. I mean, are you advocating that people will try to become experts in multiple clouds or how do you sort of ... What level of knowledge do you think you need to have in order to work across multiple clouds?

Mahdi: I haven't met a single person who can claim to be expert in more than one cloud provider and I have talked with many experts because I have been running serverless in Finland and so I have been talking with many experts. None of them dared to claim that they knew it. I mean, even keeping up with one single cloud provider is a lot of work and I don't consider myself expert in any of them either, because I'm not hands-on anymore. The thing is that ... No, you don't have to be experts to work with different stuff. Of course, at some level you need some ... For example, you might need an Azure expert to work with Azure, AWS expert to work with AWS. But in my opinion, if you really want to keep up with the technology and so you need to be good in one provider, really good with that and then, know the fundamentals of the cloud, the best practices which are, I would say, it's irrespective of which cloud provider you are using there and be willing to learn.

For example, it happened to me. At that time, I mean, when I wrote that blog post, I was only working with AWS. Then they said to me that, okay, you have this project on Azure, go for it and I never touched Azure before. It was a lot of pain, but I learned a lot. So I mean, as I said, the fundamentals are same and now be expert in one and be willing to learn. In my opinion, that should be good enough.

Jeremy: Right. I'm curious, I think that's good advice to sort of be well-rounded. I mean, that's always good advice I think for technologists, going a mile wide and an inch deep is usually good enough. But like you said, being able to be an expert in a specific field or a specific technology or something like that can really help. So you think that's certainly a good career choice to sort of start to broaden your perspective a little bit?

Mahdi: Definitely. Actually, I was one of those AWS fans that really was following this Hero, Serverless Heroes, and so on, basically was parroting whatever AWS was telling and I was saying that I just want to come to work with AWS. Actually, it happened to be like that, but when I joined my current company, my manager said that most of the opportunities that you are filling, I mean, in my department, so is mostly Azure. So basically they said that it is as it is, and cope with it. And I felt very happy actually. When I, for example, see ... Well, I'm sure that anyone who is in the cloud gets many job offers from recruiters. I was thinking about it, at some point when I was AWS guy, at least in my experience, half of those job ads were irrelevant and ...

Jeremy: Right. Right.

Mahdi: ... depending on the country. For example in Finland, if you are Azure ... AWS is very popular at least and if you are Azure expert, you are going to miss a lot of opportunities. But at least in my experience, if you say that you are with that, you have worked with the other one, you know something, a lot of career opportunities opens up. This is my observation.

Jeremy: Right. Right. Yeah. And I think actually, you made a really good point and that's certainly, in terms of AWS heroes and so forth. I'm an AWS Serverless Hero and we get inside information but we spend a lot of time thinking about things the AWS way. AWS is very good at what they are doing with serverless and they have an interesting perspective in terms of what they believe serverless is supposed to be and what that roadmap looks like. But even just hosting this show and talking to so many different people in different clouds and different ways that they do it, getting that different perspective of how other people or other clouds think about serverless and how they are building it out. I think that's actually really good context to have.

Mahdi: Yeah, I agree. Actually, you are one of my heroes also, I was following you. But I should say that it has its own advantages and disadvantage was that I was in a kind of AWS bubble. But when I started to see that, okay, even AWS itself opens up having this multi-cloud offering and some serverless heroes start to write about that, I was like, okay, that's time for opening of your thing. But I mean, by that time actually, I already started to use Azure. So again, I mean, I would say that what you have been doing, actually heroes are doing a great job, really doing a great job.

Jeremy: Absolutely, totally agree.

Mahdi: Azure also have similar. If I remember correctly, they tried MVP, something like that.

Jeremy: I guess, that's MVP, yeah.

Mahdi: The thing is that, at least based on my observation, they have more or less same level of dynamics or a narrative between themselves. They also consider Azure more and AWS more and so. But I was lucky, maybe by the choice and so that somehow I had to join or use or attach to both communities. Yeah, it has been a very valuable experience.

Jeremy: Yeah. Yeah. So you went through that process, you were sort of an AWS convert or I guess, an Azure convert from AWS, and you stayed connected. But I know, that idea of transferring your skills and transferring the concepts and you mentioned sort of the pillars are the same as they are in AWS and you sort of have some of the general concepts, but as someone who went through that, what were the challenges that, what were some of the, I guess, challenges and the barriers that you faced going from AWS and that way of thinking into the Microsoft world?

Mahdi: That's a very good question. The thing is that in the department, at that time I was working at Accenture and actually all of us were big AWS fans because at least Accenture owned Avanade, so Azure was very separate, we were in an AWS bubble. Yeah. I'm sure that definitely AWS is much more mature in many aspects than Azure, no doubt. At least it was like that and I'm sure it's still like that. Their gap has been narrower, but that still might be the case. I remember at that time, many of my colleagues were really bashing down Azure, really bashing down and they were right. I mean, some of their services were really immature. But then again, I had the chance to ... Actually, it wasn't quite choice, they said to me, okay, this is an Azure project. Basically, it was a team, I would say quite junior, developed something on Azure, something that you never probably want to hear.

They developed everything in browser, infrastructure as a code nothing at all, they were junior, so they made quite many mistakes also, but they just made the app up and running. It didn't matter how or what, it was just running and that's all. So they told me that, okay, we need some little improvement, this was little improvement and that little improvement basically forced me to reverse engineer whatever they had done, and that required me to upgrade the whole application, because as I said, there was no infrastructure as a code, if I want to use it I had to use ... If I wanted to do local development, I had to use Windows, I had only Mac, so I had to change the complete platform. It was a very tedious process by itself. On top of that, I had to start to see how Azure functions work and that was another pain for that.

The thing is that I had AWS mindset and I was thinking that, okay, AWS is the best, they came out first with the cloud and Lambda, so Azure should be something like that. As I elaborated in the blog post, no, actually they are different and there are some small patches or nuances that makes some even days to find it out, but you need to find it out, otherwise, your app doesn't work. After a while when I reflected the things, I realized that, okay, of course, I was angry and pissed ... I was really bashing down Azure, it was fight of the dynamics over there, but after a while when I reflected through my whole process and actually I wrote in the blog post, I realized that part of the blame was on me because I was expecting Azure to work in the AWS way. No, that's not how it works.

I mean, when you look at, for example, authentication or the mindset, it's different. That requires a learning curve, I mean, you need to find out Stack Overflow and actually, the Azure community is really supportive. I really like it. They have their own community which is really supportive. So the pain basically was that ... Yeah, I had to find out how things work in Azure and what's different. But now that I'm working basically pre-sales in both of the cloud I can say that, again, fundamentals are same.

Jeremy: Right.

Mahdi: And these AWS architecture framework, there are five pillars. You can see that Azure has copied from AWS, it's obvious. Even they haven't changed the name. The naming is similar and you can find that it's just a bad copy. At least like few months ago that they had to implement for that. But at the end, I mean, Azure is catching up fast.

Jeremy: Right.

Mahdi: It's undeniable. And fundamentals are more or less same. I mean, if you want to make your app ... For example, you want to innovate, you should have shorter time to market. Basically, you need to use infrastructure as a code If you want to make your app really high-level appeal, you need to follow best practices, do maybe SRA. At the high level it's same, but when it comes to the detail level, it can be very different. Even the documentation was really confusing and it wasn't just me telling that.

Jeremy: Out of curiosity, was the documentation for AWS more confusing or was the documentation for Azure more confusing?

Mahdi: This is a million-dollar question. Actually, I thought that maybe it's me. I found the Azure doc very confusing, but I thought it's me, so I asked I think nine of my friends who are AWS experts that, "What's your opinion? Have you worked with Azure? Do you find documentation readable?" I think all of them said that it's confusing.

Jeremy: Yeah.

Mahdi: So I was like that, okay, then it's confusing. Then I talked with a few Azure experts who, they breathe in Azure, they are Windows guys and they never touched AWS and they said that, "No, documentation is good. Everything is fine." Actually, if I remember correctly one of them said that, "Actually, I find the AWS documentation confusing." It seemed like two different worlds, you know?

Jeremy: Right. I find them both confusing, actually.

Mahdi: Maybe now it has changed.

Jeremy: Right. Yeah. So, that's interesting. I mean, I think the documentation is a good ... Well, first of all having good documentation is important and I think they both have good documentation, but I do think it's organized differently, right?

Mahdi: Yes.

Jeremy: And again, it's organized more towards I think maybe that different mindset. But let's just talk a little bit about the maturity of those, because to be fair to Azure, I mean, Azure or Azure Functions, it has come a very, very long way. I remember way back in 2018, way back, I mean it seems like a long time ago at this point, seeing very early demos of Durable Functions and I remember thinking like, oh, that's just a mess, like that is not the way that you want to do that. Now fast forward three years, Durable Functions are pretty cool and they do a lot of really interesting things. It does take time to catch up. So certainly I would think your criticism of Azure Functions back then in terms of what it is now, that's probably there is a huge gap there.

Mahdi: Yeah. I'm sure that most of the criticism, the detailed one that I mentioned the blog post, I'm sure that many of them have either been fully addressed or they have been improved a lot. So that's why I don't want to focus that much on detail and I would focus more on the high-level things. Yeah.

Jeremy: Right. So speaking of the high-level things, let's go there for a second. So you mentioned like a well-architected framework, sort of this idea of their being something very, very similar, maybe even a carbon copy in Microsoft. But what about getting down, you said that your individual skills are kind of when you get into the weeds there, that is certainly different, so I mean for the most part though, event-driven, stateless computes, things like that, do those skills transfer over?

Mahdi: Yeah, they do. It's just a matter of implementation. For example, I can tell you, yes, those ones ... Well, there is some caveat. For example, I remember in Azure community, I was at that time, this probably has been changed, but I think it shows some kind of mindset. I was struggling to find out the observability tools of Azure, if I remember correctly it's what's called Application Insight, one of the tools, and they had some event driven insight, something like that which was, they call it near real-time. I remember that basically when I want to get the logs from the functions, it took three minutes to come up, three minutes. At the same time CloudWatch, for example, it was coming in 20 seconds, something like that, 10, 20 seconds and I mentioned it in their community.

If I remember correctly, it was a notable dude, either one of the product team, or he was a very notable dude and he said that three minutes time is, in my opinion, is near real-time. He said that and I remember we made a lot of joke out of that sentence with my colleagues about that.

Jeremy: I can imagine.

Mahdi: But that shows some kind of mindset. I mean, three minutes, I don't think is near real-time. Most probably this time has been reduced, but I just wanted to tell you their mindset about that. But, yeah, event-to-event stateless stuff, they are transferable. But when it comes to implementation, it's different. For example, as I mentioned that blog post, there was some stuff that you can do with an authentication with some, certain some, environmental variables in AWS, but that same thing in Azure, if I remember correctly, is done through something like service principles, it's different. So if you try to play with environmental variables, it turns out no, it doesn't work that way. It gets to very detailed stuff, that gets different. Yeah.

Jeremy: Yeah. Right. Right. Yeah. I'm curious to hear about like another sort of interconnectivity of what you would connect. I'm now trying to remember what they call bindings or triggers and bindings in Azure functions as opposed to events or actually event sources, I think we call them in the Lambda world. So would you look at the way that you connect to other services? Is that another thing that is similar between the two?

Mahdi: Okay. I should say that I don't remember that much of these details anymore, but as far as I remember, again, the high levels were more or less the same. Okay, they call it three gears, but I don't remember now what does AWS Lambda calls it. But it was more or less the same.

Jeremy: I can't even remember what it's called, it's like event sources or something like that.

Mahdi: Yeah. It was more or less same. Yeah, yeah, yeah.

Jeremy: Yeah.

Mahdi: And they had something like a bus, events bus in order to have a centralized event driven thing. It's same I would say.

Jeremy: Yeah.

Mahdi: Again, when it comes to poor person who has to implement it if he hasn't done it before. But the person who is doing the high-level architecture and so, I can easily see that, I mean, I don't see that much difference. But I know that if someone has to implement it and hasn't done it before, he will go through the most pain, because he has to find this small configuration things that, unfortunately, you need to make them. Otherwise, it doesn't work out. But high-level, it's same. It's event ...

Jeremy: Yeah.

Mahdi: Yeah.

Jeremy: I think the nuances are always those tough things. So thinking of the overall mindset here and sort of maybe the approach to serverless. So I know you went from AWS to Azure, but I'm curious, do you think it would be easier to go from Azure to AWS or easier to go from AWS to Azure?

Mahdi: Well, I came from this part of the river to the other one, so I can just speculate about the other part. But I would say it's more or less same, because again when I talk with a few Azure people who really have been breathing always in Azure and never touched or barely touched AWS, I felt that they are feeling same thing about AWS. So I would say it's more or less same. They need to go through the same pain, they will find AWS stuff very confusing, especially that they will not have that great community support of AWS, but they need to either do the Stack Overflow thing or have a enterprise support of AWS. I would say it's more or less same for them.

Jeremy: Yeah. I mean, I think that's interesting too just, that it is different enough that there is pain there, right? I mean, it would be nice if there was some standards and I know there's like the opening, the Cloud Computing Foundation is like open events and some of those things whatever, not that that's all working out for ... I think Kubernetes and Knative and those and some of those teams are implementing it or those projects, but I'm not sure the same things fall into AWS. But anyways, go ahead. You have any thoughts on that?

Mahdi: Actually, that Cloud Foundation, I was working at Eficode and they are really working that stuff. They are so good in Kubernetes. I find that also another world completely.

Jeremy: Yeah.

Mahdi: This Cloud Foundation stuff. I never had to implement any of that for any of our customers in any of the companies that I worked, that they were AWS or Azure. Yeah, some of them they used Kubernetes also, but that CNC or whatever it was ...

Jeremy: Yeah, CNCF.

Mahdi: Yeah, yeah. I found it, that's a different world for me also, I should say. Sometimes out of curiosity, I played with it, but I never ... Nobody ever asked me that, do you want to use that?

Jeremy: Right. Right. Yeah. No, that makes sense. All right. So we talked a lot about, we've been talking about the difficulties in switching between different cloud providers, but also the value of knowing those different cloud providers. And more so, so that you can build serverless applications. So let's talk about serverless in general. I know you are a little bit outside of the ... You're not in the developer role anymore. But this actually, could be really interesting to get your perspective on the management approach to this and how other companies are thinking about the value of serverless at a management level as opposed to ... I guess, even as a sort of planning level. So let me ask you this question then. Are you seeing companies looking at serverless and adopting serverless and that serverless mindset and then maybe a follow-up question would be, if they are not, why do you think they should be embracing serverless?

Mahdi: Okay. Firstly I'll answer the second part. Basically, the thing is that nowadays the world is fast changing. Many companies, many corporations basically, are benefiting from their existing market share or regulatories or the monopoly that they have. For now, it works. If they don't want to change basically if they have the mindset that things are working, what's the point for change. Most probably within a decade or so they are going to die, their business is going to die. Because the world is fast-changing and they need to have them to adapt to the market.

So ideally, they need to go through the pain and disrupt themselves. Disruption always brings pain. You cannot disrupt yourself and feel that everything runs smoothly. Ideally, they need to disrupt themselves, go through the pain and so become really agile in order to understand the customer feedback and deliver the value to the customer, what really the customer wants. They can either have this phase or they can ignore it and say that, okay, things are working, we are making money through our monopoly, regulatory, existing market share, whatever and then, their business is going to go away. These two choices, that's all. Yeah. Painful process to become more competitive and be ahead of customers or assume that everything is okay, and then at that time that's going to be very late.

Jeremy: Right. So let me go back to that first question then. So you are seeing people not doing that?

Mahdi: Okay. The thing is that what I'm telling is going to be biased because I'm working in a cloud team and whatever opportunity that they are going to bring to me, of course, you have the departments and the companies that they are interested in the cloud. So my mindset is a bit biased, but what I'm seeing is that it varies a lot and I mostly focus on corporations, because ... Yeah, of course, for startups it's much easier to go for that.

Jeremy: Right. Of course.

Mahdi: At least in Finland, my observation was that there are two ways. Either they are very ... it depends on the executive leadership. For example, a major bank in Finland, they say that, we want to go to the cloud and be, we want to go for that. And once, one of these big ones goes through that, there is going to be a domino effect on others. But there are some other ones say that, no, it's cloud, who is going to take care of the data? We are not going to do that and they don't touch it.

There are some other companies and their departments, I would say there are departments who are interested in trying things out and then, they have to fight internally with the more conservative departments. So I'm sure that there are three levels of that. But mostly, I work with the ones who are inclined toward using the cloud.

Jeremy: Right. Right. So then, the ones that are starting to dabble in the cloud, is that something where you see ... I mean, clearly there's lift and shift, right? Which I think we probably all understand at this point, it is not the best implementation or the best use of the cloud, right? That it is better to maybe use more native cloud services or cloud native services, I guess, to do that. So in terms of people just rehosting or maybe re-platforming, are you seeing this sort of rearchitecture, or I guess, this refactoring or is that something where companies are staying away from that?

Mahdi: First of all, I respectfully maybe have to disagree with you.

Jeremy: Okay.

Mahdi: Actually, I think rehosting is actually a good approach and that's what even AWS promotes for conservative companies who want to start working with the cloud and they want to get the fastest result in the shortest period of time, with the least amount of pain, it's better to do migration through the easiest one which is lift and shift. Easiest, everything is relatively.

Jeremy: Right.

Mahdi: And then, have a data-driven approach to see what really needs to be improved and then refactor or rearchitect or re-platform based on data. So in AWS terms, I'm sure you're right there with me, have that evolutionary architecture in a data-driven approach. So lift and shift, I don't consider bad at all. Actually, I consider it a very good cornerstone, stepping stone at the beginning, for the beginning.

Jeremy: Interesting. Okay.

Mahdi: Yeah. What was the other question?

Jeremy: No. I was just going to say, so you've got companies that are lift and shift, and, yeah ...

Mahdi: Oh, okay. Sorry. Sorry. Yeah. Sorry, I just remembered.

Jeremy: Yeah.

Mahdi: Sorry to interrupt you. Actually, I'm a bit careful about using the word cloud native. I remember, in a previous company that I was working, we had some philosophical fight about that and I'm sure that then everyone was dissatisfied and I had to have an authoritarian appearance that this is the definition of cloud native. I'm sure many of them hated me after that. But the word cloud native, I really struggle to find a consensus of what does it mean and if you spend some time, you realize that you will find a variety of definitions of that. So I'm picky for the word cloud native. There is a lot of fight can happen, what is exactly cloud native. Some consider Kubernetes cloud native. Some consider using AWS or Azure cloud native. So this is the picky ... this is a very controversial term, I would say. Yeah.

Jeremy: Well, let me interrupt you for a second. So when I think of cloud native, what I'm thinking of are services and components that are built specifically to run in the cloud, things like your API gateway at AWS or Azure functions or things that are like very much so built to run in the cloud environment where they do things. It's that serverless aspect. I think of it more serverlessly. I mean, I know containers and so forth fit in there as well. But that's how I think of cloud native. I think of cloud native as going beyond just your traditional VM and running everything on the VM and moving to the higher-level services that are more managed for you.

Mahdi: May I challenge you?

Jeremy: Absolutely.

Mahdi: So you just said that basically things that use cloud, like API gateway and so. And now I should ask more of a technical question. What is cloud?

Jeremy: Right. Well, that's another good question. Right.

Mahdi: Okay. I can tell you, based on these several definitions that I read and I reflected on them, I have this definition of cloud native, most probably many people I'm sure will disagree. So that's fine because it's very controversial. In my opinion, cloud native is very simple. If your application is architectured in a way that it can leverage the advantages of the cloud environment, then it's cloud native. Doesn't matter if it is on Kubernetes, if it's on AWS, if it's on Azure or so. If it can scale to zero and theoretically to infinity and you pay for only what you use, then it's cloud native. That's my definition of that and I read so many definitions, so I came up with this. But feel free to disagree with that, because many people disagreed with me. I'm fine with that.

Jeremy: That's all right. You are not the only one I'm sure, has differing opinions of what cloud native are. So let me ask this though because I think that's interesting, the way you explained the strategy of lift and shift of basically being able to say it's the, probably the lowest risk way to take an application that's on-prem and move that into the cloud and then to use data and so forth to kind of figure out what parts of the application might you want to migrate to, maybe again I don't want to overload the term, but more cloud native things. I think that's actually really interesting. I have found and I have seen many companies that seem to do this where it's more that they move things, they just rehost without really thinking through what that strategy is going to be and then they basically just end up having their on-prem in the cloud and not benefiting from some of those managed services and some of the benefits of the cloud that you get, they don't transfer on to them. That's what I have seen.

Mahdi: Well, you know it better than me. Your cloud environment is never perfect and it's always an ongoing operation. So I mean, going to the cloud ... Again, if you put your own frame in, put them I don't know, use EC2 or which VM or the AWS or Azure, that's a very good first step ...

Jeremy: Right. That's probably true.

Mahdi: ... but you need to be able to start leveraging that. At least get the data, which one is being used and hopefully, hopefully when you are going to the cloud, you have done some analysis and you have realized that some of the services even are not working with the cloud. Some of them need to retire, some of them cannot be rehosted. They must be rearchitected, because they are so legacy for that. But even again, assuming that you have done your homework and you have done rehosting, okay, you need to leverage that and go and see that all things that AWS or Azure provide, how much over-used or over-utilized or under-utilized are your CPUs and this kind of thing and according to that, do right sizing for that.

Jeremy: Right.

Mahdi: That's a good step for that. Then if they want, requires refactoring, try to I don't know, do refactoring and use more managed services for that. So again, rehosting is a good first step, but cloud is a long journey. I don't know who came up that cloud is cheap, I really don't know.

Jeremy: Right. No, I totally agree. You are right about the first step and I actually loved your point about which services might you be able to retire and not move at all because I think in a lot of these big companies, there are a lot of services that you probably don't need anymore or they are redundant or whatever and you could get rid of those moving to cloud. Good point. All right. I got a couple of more minutes and I want to go back to an article that you wrote. Now, this I think is like three years old and in terms of reading the article now, it's not relevant, because so many things have changed. But what's relevant is, what has changed and this was an article that was about the worrying and promising signals from the serverless community. I think this was an event you went to in Germany, they did this, and you have a couple of different points that you called out.

One of the points was that users have ignored security and that was a worrying sign for you. Where do you think sort of cloud security or more specifically serverless security is now? Do you think people are still thinking about it or have brought it front and center like it probably should be or do you think it's still a worrying factor?

Mahdi: Since I have implemented cloud solutions for I'll say mostly enterprises and a few startups, I haven't seen a single one of them using, having a cloud security specialist. Most of the corporations when they, at least in my experience, when they want to go to the cloud, they must address the security of it and typically because of the customer requirements, so they bring a security guy who has worked with this, let's put it this way, all their security stuff and he has to come in on the cloud part and it's funny that actually, sometimes I have to teach them basically. I remember they had a head of security for a customer. I really had to teach him and actually, I had the Lambda functioning in front of him and he was like, wow, is it really like that? I had to teach him what are the attack methods and it was funny. He had to sign off my solution that it is secure, but basically, I had to tell him what are the priorities.

Jeremy: You had to tell him what it was.

Mahdi: They address it from a traditional way. Yeah, they do some kind of a test, automated test and this kind of thing which is, yeah, definitely ... Again, I'm not a security expert, but as far as I understand, again they have some fundamentals which are safe, that's true, but when it comes to the cloud especially serverless and functional service, you will see that there is a lot of more attack vectors and unfortunately, these security experts, I have not seen any of them who have any expertise in that. I learned about it because I was curious about it and I started to work with basically professionals, some startups which provide professional security solutions for serverless. So that's how I got that, but again I had to go through the pain. It took few months to read so many stuff. But I haven't seen any security specialists who have been working on cloud projects who have done this.

Jeremy: Yeah.

Mahdi: So I would say customers, they consider it, but no, there is still a lot more way to mature.

Jeremy: They are not addressing it. Yeah. It's funny because I remember that in 2018, 2019, there were a couple of companies that were in the serverless security space and they were all acquired. So now they are part of larger platforms which is ...

Mahdi: Exactly.

Jeremy: ... great for them, don't get me wrong. All right. So then another thing you said and I think this is important, because the biggest complaint that I always hear about serverless is, just the workflows are not easy. So you had mentioned that DevOps was finding its way and that was sort of a promising signal, you think that we've ... I mean, we have got a lot of tools for serverless now. Speaking of Azure, the way to deploy an Azure function right through VS Code now with the plugins is really, really slick and Serverless Frameworks, SAM, CDK, all these are there, Terraform and so forth. I mean, have we gotten to some stability around serverless and sort of mixing in DevOps there?

Mahdi: Based on my experience, at least the ones that I have worked with, I can say that, yes, DevOps is now a part of a solution that's provided to the customer and maybe it's correct because personally, I went through the pain whenever I proposed any solution for the customer, so they are always using infrastructure as a code and always try to have a DevOps-centric viewpoint about your solution. So I try to push for that and, yes, I find customers receptive about that. It seems to me that, now DevOps is not one of those buzzwords for cool kids who just want to do this stuff, even the corporate guys are more receptive with that. Again, there is more way to really do the DevOps stuff, because you know that many companies claim that they are doing DevOps, but in reality, they are not. You know this better than me.

Jeremy: Right. Of course.

Mahdi: But, yeah, it's good. I'm happy for that. I mean, a few years ago DevOps was one of those buzzwords, but now I don't think it's buzzword anymore.

Jeremy: Yeah. Yeah. And I think that serverless has actually opened up a lot of making it easier for teams to do automation and things like that, there's a lot that you can do because you have that little bit of compute power that you can do something with. So I think that's definitely promising. So speaking of sort of compute power and other things that you can do with it, one of the things you mentioned was that you saw as a promising sort of signal was, that serverless-based prototypes were on the rise, meaning different services, so whether it was cues or whatever or I guess Lex and things like that, all kinds of services that allow you to do different things that are specializing in different capabilities. So how do you feel about that now? Because there are a lot of those APIs out there.

Mahdi: Yeah. Actually, I also find that even from these legacy corporations that I have been working with, I like the idea that now, they definitely when they want to do migration especially or this kind of thing or do anything cloud, first they do POC. Yes, I find it good. Honestly, I was sometimes impressed that, oh, from some people that I would never expect them to use this one, first let's do POC, then see what's come out. Oh, really? Yeah, it's good in my opinion. It's finding its way.

Jeremy: Yeah. Yeah. No, I like that too and I think you are right about proof of concept, because it's just one of those things where even if it's expensive to use the Google Vision API or something like that, it's a really good way to prove out how that fits into whatever the business use case you have for that and then like you said, you can certainly take a step further and create more sophisticated or I won't say sophisticated but maybe more integrated tools or something like that, that would work around that. So I think that's interesting, allow people to fail fast, learn quickly, and just build out their applications.

Mahdi: Yeah. When we say POC, I should say that I wouldn't exclude it only to this cool new serverless or what the AI stuff that AWS and Azure provide. Even for migration actually, POC is highly recommended. Again, I was working for some period of time for, I would say, one of the most conservative banks in Finland, small and conservative, for consultancy, but even then as we are trying to push the cloud and even then they said that, "Yeah, first let's do a POC of migration and see what's going to happen." Again, there really I was surprised. I would never expect it from them. But the idea of fail fast and learn fast, I think at least that it requires some level of maturity to reach that.

Jeremy: True.

Mahdi: That really needs more room for improvement, fail fast, learn fast. Yeah. Just something, I don't know, I would like to address about this cloud stuff if I can.

Jeremy: Yeah, absolutely.

Mahdi: Yeah. Basically, when companies or customers decide to go to the cloud, I'll recommend that don't look at only the technical aspect of it, because I see that there is, at least there is lot of debate for example ... At least it was like that. AWS, Azure, or this kind of thing, at the end I'll say that most of the things it doesn't matter that much. I mean, it depends on their, sometimes company policy, how much discount you can get, how much funding you can get from the cloud provider. So it's not really the technical people who decide, sometime it's the executive who decides.

Jeremy: Right.

Mahdi: But even then, when you go to the cloud, in my opinion as much as the technology and maturity of the cloud provider matters, the amount that your company is ready to change its operations is also important. This is my favorite example, that I developed and I would say at that time at least a state-of-the-art serverless solution, DevOps, or CI/DC stuff for a major bank in Finland and I was the first one who managed to do that among so many consultants that they have. It was really good. I'm proud of what I did and actually, I open-sourced that. It was really basically we could deploy multiple times per day and we went to their release manager and I said that, "Okay. It's like that. Everything is perfect. DevOps, CI/CD, we can release multiple times per day." And she said that, "No. It doesn't work like that. We need to release once per month," and we have to go through a very painstaking process, fill out so many useless documents.

It didn't matter how much I tried to convince her that, "Well, the idea is different. I mean, you need to do small deployment. This way actually you have less risk. You deploy once per month. Still every time something goes wrong, but when you do a more frequent deployment, your risk is lowered." She said, "No. We are a bank. It is as it is. Sorry." Most of that effort that I made at least at that time went to waste basically, because the process was legacy, even though the technology was good. I'm sure that by now, they have changed because I was among those innovators basically or the early adopters who made through that. But in my opinion, technology matters but operation and processes and release stuff also matters and everything needs to change. So basically it needs to be holistic approach of going to the cloud, not just implementing from technical viewpoint.

Jeremy: Mahdi, thank you so much for having this conversation with me. This was a lot of fun and then I love people who have sort of experienced, from moving from one cloud to another. It's a huge shift, but I think your advice here is great, just to sort of know those basics on those other platforms and do that. So if people want to reach out to you or find out more about, follow you on Twitter, things like that, how do they do that?

Mahdi: Well, I have a Twitter account, but nowadays I mostly put non-service stuff, but LinkedIn is a good option for me.

Jeremy: Okay. Great. And it's m_azarboon on Twitter and then, I will put LinkedIn and Twitter and that in the show notes as well. Thanks again, Mahdi.

Mahdi: Thank you very much for having me. Bye-bye. Thank you.