I’m one of the founders of Yellow Tail Tech. We’re going to talk about cloud computing, which is what I do now and have done for the last 4 years.I’m still an admin, an engineer, but I do most of my work on servers that run in the cloud instead of in data centers that are controlled by my organization.
I’m going to start by the last slide, cause people like to know this first. This class is for you to learn about the cloud, to start seeing the cloud as a big opportunity for people in our field.
It certainly was a huge opportunity for me. It allowed me to grow really, really fast. I can really attest that people with cloud knowledge and cloud experience are being sought out aggressively, just because there’s hundreds of thousands of jobs that cannot be filled. So every position I’ve held has been open for more than 1 year, so they’ve been looking for people for over 1 year and they can’t find them.
A little bit about my journey, I am going to add a 0 there. First of all, I changed careers. Before I was a sys admin and I knew anything about Linux, I was a social scientist. I worked with statistics, public policy, and other things like that. I had one of those midlife crises, and I decided that I needed to learn something else. That’s how I ventured into IT.
The first thing I did was I decided to become a great sys admin and I’m saying great, just because I like to take things far. I figured out what a sysadmin needs to know, but also what makes a sysadmin valued. Having an RHTSA and an RHCE was gold. Everybody told me that I could find a job immediately, that salaries were good, and that after you had those credentials, you wouldn’t be asked most questions because you have already proven yourself. Once I actually had the RTSA and the RTE.
I made sure I understood that I had a very good basic understanding of networking, because I wanted to do cloud in the future, and I knew cloud is network heavy,I decided to go and and like study the basic network and stuff, I got a network plus.
In the beginning of my career, I worked in environments that did not use cloud, but I started to learn about it anyway, and my organization allowed me to participate in discussions that they were having about migrating things to the cloud because in the past 5 years, everybody has been talking about the cloud, moving to the cloud, trying to balance.
How they run their cloud operations in those environments, I was able to kind of figure out how to move backups to the cloud and how to do disaster recovery. Disaster recovery is just what you do when a disaster happens, and that disaster can be that your team wins the lotto or that your data center gets flooded.
Anything that can cause outages or a situation where you can’t offer your services any more is considered a disaster. I led those things in my previous job and I chose to learn about one cloud, and I think this is very important. I’ve tried to get that to you guys and some questions that I’ve answered in Slack. It’s way better if you focus on learning one thing, you learn it well, you can speak about it, than trying to learn about multiple things at once, and then you cannot really stand on your feet while being interviewed about it.
So I decided to learn about AWS and even though Azure and Google Cloud and all these other clouds seemed attractive, and people were telling me if you learn Oracle, you can get a $250,000 job in San Francisco like next week. I was still, I don’t want to do that now. I’m gonna focus on this one. I created a free AWS account, which they still offer.
They give you a free account for a year, not everything is free. There’s certain workload sizes that they allow, and I took advantage of that. I chose a first certification which was the AWS Solutions architect associate. I booked the exam. I always do that. I say I’m going to get the cert. I booked the exam 6 months ahead.
For me, even if I have to push it out, it keeps me more honest to a timeline. If I just sit here and say, yeah, I want to do that next year, that may happen 3 years later instead of 1 year later. When I had 2 certs, I was hired as a cloud security engineer, even though I had zero cybersecurity training. The reason why that happened was because a couple years ago, there was no cybersecurity training that was designed for cloud. Anybody that knew cloud, especially if you had some certs, most likely you knew how to secure some AWS environments.
They just felt that it was easier to teach me security than to teach somebody cloud, and I was given that chance. I still work in that environment. I work for Sony and now I’m a cloud security and IT manager for them. I have 5 AWS certs. I am always learning. There’s a lot to learn about. I have seen the bonuses after every one of those certifications, so it helps me keep going. Not only the bonuses, I’ve seen how companies just pursue you because they know there’s very few people like you and it’s very funny because it took me a long time to understand that that was the strength and not something that was holding me back.
Juvie always tells me to talk to you guys about this.One of the things I struggled the most at the beginning of my career was something that they call imposter syndrome.
All it means is that we constantly feel like we don’t know about a lot of things, and that the other people in our environment do know about those things, and we don’t feel like we fit the typical IT nerd looks, or like, it’s just not how you see yourself. I didn’t see myself that way, but I think I’ve definitely become that person and I had to learn how to see that I brought something different to the table.
The same way you feel like you don’t have things that your coworkers have, or that even your, your peers in this class have, you have something specific and you have to figure out how to make that something that you can showcase. That’s the little bit I had to share about myself.
I would like to know what is, what is cloud computing and what is AWS.
This whole class is gonna be about that. How much Linux?
I already had an RHC so I was doing engineering work with Linux. Now as an IT manager and the person that’s constantly interviewing people, I would not consider any cloud person that doesn’t have a Linux background. A Linux background is key. Linux people can make it into cloud, cloud people can’t really make it into Linux.
I’ve been remote 100% since the pandemic started, but I was remote 3 times a week before that. I was used to remote work before COVID.
When I got into the cloud, all I knew was bash, and you’re learning that in this class.
When I got my first AWS certification, that was like 5 years ago. The cloud practitioner didn’t exist. That was the starting point. I think you could get like any one of the associate level ones, but the architect is the one that talks about every single AWS service.
AWS is why I’m in this class, why I came to Yellowtail it is because I get different people telling me, you start with the basics, start with the architect, which one do you think is good to start with? I play around with the, the console and I’m try I’m focusing on the practitioner.
If you’re just trying to learn the proper basics and you wanna start getting the lingo,it’s a great start and it’s free, but it’s not gonna get you a job.
What is the cloud and have you used any cloud service?
It’s Microsoft Office 365. I was the network engineer there, but I used to create users and give them access, whether they can modify things in, Office 365 or not and I believe this is a software that is not directly in somebody’s server, locally in the server, but it is in the cloud. I would take that and Google Drive and I use that a lot for my work that I used to share with my students. It’s easier to share it with multiple students in one setting. That’s how I use the cloud.
On my phone, I only have an extra 2GB for 299 on my phone that extra space in the cloud. I can store information. That’s all I know about the cloud.
Google Drive and 0365, the 365 is more than storage, and cloud services that uses on phone. Those are very common cloud services that are used by regular users, like all of us. Most of us Interact with the cloud. We’re using the cloud in several ways right now.
Zoom, believe it or not, is a cloud provider for us to host these meetings. We do not have to run a server that to manage our meetings. All this is happening from Zoom servers directly, and if we want to record these sessions, we can, and they store it for us. We don’t even have to have this to handle that story.
0365 is Microsoft’s enterprise solution for storage and their Microsoft suite. Think about like, Microsoft Word, Excel, all these things tie into that platform. Right now, everybody is using, at least every major enterprise is using O65. There’s like, no way to run from it., that’s something else that’s interesting about cloud.
Even though we hear a lot about AWS, there’s a lot of other vendors that are in this arena, and there’s jobs around all these tools. There’s other enterprise cloud services like Google Cloud, they abbreviated as GCE, Amazon Web Services, AWS, Microsoft Azure, Salesforce, which was the first cloud provider, WebEx, Zoom, so there’s hundreds or thousands of cloud services right now.
This is how we tend to imagine the cloud. Tend to imagine like some digital cloud that’s up there in the sky, but it’s not really that. It’s a little more simple than this. There’s no meteorology involved. What cloud actually looks like is this. The cloud is just a computer or server that’s running somewhere else, and that we’re accessing through an internet connection.
When we talk about the cloud, first, it’s not one cloud. They’re multiple. Many vendors that have their own clouds, and not every Think like you can’t access every single service in one cloud just because you’re in that cloud. Separation is guaranteed in a cloud. We access the cloud through an internet connection and most likely we’re accessing a server that runs in a data center like the ones you see on the right. The one in the bottom right is just a corridor in one of these buildings.
Jerry can tell you how many servers are in each one of these racks, because he worked in a Amazon environment at one point. There are thousands of servers in each one of these corridors, and any cloud provider tends to have multiple compounds that are separated enough to have a certain degree of independence, but connected together through low latency cabling.
You’ll see one of these, but 4 miles away, you’ll find multiple more and we live in an area that’s flooded, like Loudoun County in Virginia is like the mecca of cloud computing, or at least of data centers that serve cloud computing. There’s many of these environments. The cloud is just made of thousands of these data centers that host different customers’ data. To see this from a different perspective, we’re going to define cloud computing based on pizza.
Why pizza? Well, instead of going to solve and having to do all these things, you just have somebody else do it for you and put it to you with a click of a button or a phone call. Comes in sizes. Believe it or not, the same reason why we buy pizza or why we order pizza is the same reason why we use cloud services. The cloud is a lot like ordering pizza.
You tell the waiter you order, you tell them, hey, I want a medium, I want pepperoni, sprinkle some mushrooms, and then you don’t have to worry about anything else. The magic happens in the kitchen, the those fins in the air, and in no time you are served or even delivered that pizza. You could always make that pizza at home, and we all know that. Like they even sell the dough at at the supermarket. If you’re a little handier, you can make the dough from scratch. But with the cloud or a restaurant, you don’t have to.
So you guys said convenience, I don’t want to do it. All these things are reasons to use the cloud. Think about what happens when we make pizza at home. You need the right ingredients. You’re spending a lot of resources. And let’s say you’re like, it’s a Friday and you want to make pizza for your family, but you only have enough resources to make pizza for a party of 4 but then your cousin calls you and says, I am coming with all my kids. What do you do? You may not be able to scale, and that’s a problem we have in IT.
I can handle a certain demand, but can I go beyond that? Not necessarily. We need to have the ingredients, we need to have equipment. O the left, it fell through the rack.
Clearly they didn’t have the same, the right equipment, and that could have been because they were trying to do more pizza than they had the capacity to make. That did not work out. If they would have ordered it, it would have come out perfect. Like, it’s very rare that you have a problem with a pizza delivery. Also, you need knowledge. Like, I can have all the right ingredients, I can have the right equipment, but if I don’t know how much water I have to mix with the flour or eggs or whatever, I may screw it up, and I may still end up looking like the picture on the left.
With IT services, it’s the same idea. Sometimes you’re an organization. Imagine if you’re an organization that provides medical care, like hospital management. What does your organization know about IT? Is that your main mission? Why would you be making pizza or building your own servers or maintaining them all the time? The first principle of cloud computing is like, you could do this yourself, but you could also outsource part of it to another organization that doesn’t know how to do this.
Let’s get into a more proper definition of cloud computing. Cloud computing is the on-demand delivery of IT resources over the internet. I can’t have a cloud service if I can’t access it through the internet. If your picture is just on your phone and it’s not getting backed out of your phone, can’t be in the cloud. The other key point is, is that it usually uses a pay as you go pricing, and this is another one of the advantages of cloud computing. Remember when we had phones that had like very little storage and to upgrade it, we had to buy another card that was larger and and put it there and then sometimes the phone just didn’t have the capacity to go beyond that.
Remember pre-Apple days? Now with cloud services, we don’t even run into those issues with our phones because we have the capacity to have some of our, our storage in our phone and use the cloud when we’re running out of space. It’s the same principle, and we’re not paying to have all that storage all the time. We’re just paying it when we use it and when we need it. Instead of buying, owning, and maintaining physical data centers and servers, you can access technology servers, such as computing power, storage, databases on an as needed basis from a cloud provider. Again, We’re not running a Zoom server.
It’s pretty much what part of I is saying is like your phone is running out of memory and you buy a 256GB of storage for $50 like SD card, but you don’t have enough content to put on that $256. The cloud, you use it as much as you need. If you need 2gb, you pay only for 2gb. You don’t have to have 256gb of SD card on your phone, but you don’t have data to put on it. So that’s the PIO model. You have to use 2 GB for 2 gigs. You don’t have to get a bunch of storage that you’re not gonna use.
Like in the past, if you were a company that wanted to have an application on the internet, you would have to buy physical servers, pay for them 100%, buy licenses and pay for it 100%. This was a startup, so that startup would have to invest heavily into creating infrastructure that they didn’t even know if they were going to use because it could depend on if their product survived or not. With cloud, it has those benefits too. Like you don’t have to make an investment of capital to get something started, you just pay as you go and if it took you 200 days to learn if your product was gonna survive, you pay for 200 and not for 400, or more than that. I like the startup analogy because if we go to like a real case study, if you look at a company like Quiby, Quiby was one of the, one of the reasons why Quiby burned through the money that they did, even if they were valued at or the valuation of that company was like in.
In upwards of billions, the reason why they burned through the money that they did is because they spent so much money on storage. They didn’t have access to data centers. I invite you guys to see this as even more than the technical cost, when we talk about a data center, we’re also talking about real estate. Remember Blockbuster? They had to have a store on important avenues in every city. When Netflix came, Netflix basically got a data center in the boondocks or got cloud services or created a cloud. They initially started with a cloud and now they’re 100% on on AWS but if they had to pay for the real estate to keep their operations, they probably would have been out of business, same as Blockbuster.
In other words, they save a lot of money in resources like energy, maintenance, and that kind of stuff. That’s why they can’t survive in the market right now.
As you can see, cloud computing is nothing but a revolution, and it’s not the first revolution in IT. For us to value cloud, we have to understand how things are in non-clouds. Basically when you have your own infrastructure on-prem, and by the way, I have to mention this, just because the cloud is cheaper and I would say better, it doesn’t mean other types of operations are gonna go away. The hybrid is what’s gonna survive and for multiple reasons, sometimes regulation doesn’t allow the use of cloud. Sometimes you have and and not only like, because it’s like fed data or PII like personal identification information and stuff like that, it’s more than that.
Sometimes regulation says like, you cannot have your data stored in a place that’s more than 200 miles away. If your cloud provider doesn’t have a data center around you, you just can’t put it there. In other cases, some things do cost more on the cloud than in other places. At work, we had to cost out a project. It was going to be like a log collection implementation for a very large environment, and it would have cost us something like $15 million a year if we did it in AWS and with like heavy discounts, but it would cost us like $12 million if we bought our own infrastructure. So for that use case, it made sense to keep that operation on prem.
How are things on prep? On prep, things can look like the picture on the bottom right. Many people work in environments like that. Fortunately, I haven’t had to manage it. You have to host your physical servers on your real estate. It depends on the organization, what they’re doing, and where they are. If you’re an organization with only one building, like one on 5th Avenue in New York, and you want to keep your IT personnel in that same building, then you’re paying really expensive real estate to put those servers. You can also rent space in a data center and put your servers there, but you’re still paying for it. Renting a rack is usually very expensive.
You also have to provide physical and network security for those workloads. It’s not just about giving a huge space in your building; you also need to install special fire controls in that room. You can’t just pour water if there’s smoke—that would fry everything. You have to invest in that. You also need firewalls to prevent someone from putting a USB on the printer and gaining access to your HR tools. You need people who know how to do this, which adds another cost.
Temperature control is another factor. Servers are very sensitive, and it’s common knowledge that wherever there’s a server, the temperature is kept really low. Servers generate a lot of heat, especially if you have thousands of them together, so you need strong AC systems to keep the temperature optimal. If not, your applications could fail because the servers crash.
Finally, you need to make large investments to buy equipment. A server is not cheap. Think about it—if your MacBook Pro is expensive, imagine the cost of servers.
A supercomputer will cost significantly more money. I’ve been in environments where servers were purchased for over $35,000 each. That’s what you’ll be managing, and it’s why you’ll be paid well. The challenge is that not only do you have to make a significant investment to buy these servers, but you also need to replace them every three to five years. The moment they’re installed, they start depreciating, and their performance begins to decline. If you get five years out of any enterprise-level equipment, you’re lucky.
By year two, you need to start planning ahead—evaluating whether the current equipment is meeting your needs and considering alternatives. For instance, if you bought Dell servers, you’d need to assess their performance and decide whether to stick with them or explore other technologies. This requires skilled people who can evaluate capacity and technology options.
In contrast, with the cloud, you don’t have these issues. You’re provided with a server, and if it crashes, it automatically migrates to a different one. On-premises setups also have limited compute power. Compute resources—such as memory, storage, and network capacity—are finite. For example, as Jerry mentioned, you can add 50GB of storage to a phone, but once that’s used up, you’re out of space. If you want to do more, you’ll have to consider upgrading.
The same thing happens on-prem. You just have so much capacity, you just have so many servers. And remember, if you’re in an environment that has 100 servers, prefer using very concrete examples. Let’s say you’re Macy’s. Every year, your sales go up by like 200% or way more than that on Black Friday and Christmas and certain holidays. But then the 355 other days of the year, you are running very, very low because your traffic and your demand is very stable.
So you would have to have a data center that’s equipped to handle the peak load, which you only use 10 days a year. So imagine you’re investing for the day where you’re at peak, but the 355 other days of the year, you’re running out of very stable low usage. So you’re investing unnecessarily on that. And in some cases, you can hit your limits, which we all know happened with Obamacare when they opened.
So again, as I was saying, you have to invest in that peak performance. So if you’re a startup, you need to know that you’re gonna be able to operate if your product goes viral. So if you’re investing on that, that’s gonna take a big chunk of your budget. And then you also have to handle the maintenance. When I asked, like, why would somebody order pizza, you guys talked about convenience and not wanting to do things.
Well, if you’re on prem, you’re handling your patches, you’re doing upgrades, you’re very involved in this type of operations.
And as I’ve been saying many times, you need people who know how to do all these things. You need people who know how to patch, people who know networking, and people who know how to install all those fire hydrants. You need people who can evaluate if your servers are performing well or not. There’s a lot of work that comes with managing your own infrastructure, and cloud basically saves you from all those headaches.
Any questions?
Yes, I have a question. You talked about something about parameters, like regulation. Is there a given parameter where you can actually assess a server, just like you have a cell network communicating with each other? They have it closer in different locations, so can somebody from here kind of get the services of a server in Africa?
I suppose what you’re saying is can you limit that capacity so that not everybody, or so that only people in a certain place, can access that server?
Partially, yes, but I’m also talking about the parameter of a server. Like, let’s assume you are within Virginia or Massachusetts. Does the server need to be closer to you, or can it be somewhere in San Francisco?
OK. So it depends on what your server is doing.
If you’re FIFA, the company that does the soccer games, and you’re streaming live, you wanna have content close to the users so that they don’t get latency, like so that the users in one place don’t miss that goal. But if you have something like, hey, it’s just a website where you download these forms, then it doesn’t matter where that server is most of the time. But even with AWS, I think you can do it easier than the regular data centers because I think they have edge locations where you can deliver your data to people that you want, even if the data center is not at the location.
So with cloud, it’s easier to do. Yes, and regardless of that, it’s just easier because if you were handling all your infrastructure, imagine you needed to have your data here and in Europe. You would have to have two data centers, one here and one in Europe. So all these things that I’ve mentioned that generate costs and headaches and stuff, you would have them twice. But in AWS, you could do that with a click of a button.
So it’s way easier in the cloud than when you’re managing your own resources. Yes, I think, Paulina, you had a question.
I was mentioning a comment about the people that you need to have to maintain those servers in those rooms. Not only the people who know about technology, you need to include the people who do the maintenance in the whole building about electricity, the guy who deals with the AC unit and all that stuff in order to maintain those rooms at the right temperature in order to maintain those services running.
Because remember, if the temperature goes too high, what happens with the servers? The servers crash because it’s too high, the temperature is too high for them.
So talking about electricity, my previous job was in the government. I worked for the National Institutes of Health. We were 3 engineers, and every time the building where the data center was was performing any electrical maintenance, one of us had to be physically there. So even if it was a Sunday at 3 a.m., we were asked to be there.
Yeah, I think we have two other questions. I think Naos and Martin have the hints of virtually.
Yeah, I’ve got a question for you. So, you know that, in regular server setups, you can have like a RAID setup. Now, do the data centers use some kind of RAID setup, so all of your data is necessarily stored at the same center, or can it be stored in different places?
Yes, you can do that in the cloud as well.
Was that your question?
It is, yeah, you were talking about the regulations a little while ago, but, you know, you can’t be over like 200 miles from your data center. Well, in the case of an outage, I mean, if your data is stored in several different centers, you can’t, it’s not like you can actually go next door and get your data if you have an outage. So what’s the reason behind the mileage?
Well, again, they’re different regulations. So sometimes state governments will say, if you have money from the state of Boston, you cannot have your Boston data outside of Boston, you get it?
Like, it’s very common with military operations, like confidential data and stuff like that. In some cases, it’s not so short-sighted, like 200 miles. I remember, at one point, I worked with the European Union, and they had all sorts of nonsense. Like, if they were funding a project, you had to buy equipment that was made in Europe, and you had to store the data in Europe.
And that’s just how it works for the EU. They’re trendsetters. They even have a privacy law that’s very popular and everybody goes by it, even if you’re not in Europe, just because you want to make sure that users in Europe can use your applications.
Is that the Digital Privacy Act?
Yes.
OK. Well, that brings up another mind. I’m sorry, when it comes to regulation and, and say doing business overseas, when it comes to the regulation piece of it, is there like a standard when it comes to, say, a methodology or a process that other countries recognize versus the US?
I’m speaking more so on like the ITIs or the agiles or the DevOps, kind of modes of operating.
Not really. It, listen, to be honest, every organization is totally different from the other one.
I have not seen two organizations that do their planning in the same way. I don’t really see a consensus there. I do see principles that are more important every day, like agile principles, which are really popular in the sense that you divide work into little chunks. If you have an application, instead of saying, “Hey, in November, we’re gonna release version two,” you’re constantly releasing features. In the back end, it’s the same thing.
If we’re fixing our data center, like if our data center looks like the picture on the bottom right, we’re gonna divide that project of reorganizing that rack into multiple little steps, assign it to different people, and try to say, “By this day, we should look better.” But that doesn’t mean we’re going to stage all the work to happen in one day.
This type of principle is common, but when you get into the technical stuff, like what tool you’re using to do that, it changes everything. The way one tool works is different from the other, and things get lost in translation.
Earlier, you mentioned the Obamacare crash when people started applying. Would a larger cloud have prevented the crash?
Yes, that could have been prevented.
Every time a startup launches a service, it doesn’t matter what size it is, you’re always guesstimating what the capacity will be. I’ve done things at work where I thought a certain server size in the cloud would work, but when you start getting traffic, you realize it’s not enough, and you have to adjust.
When you’re on-prem, if the server doesn’t work, you can’t return it. You’re stuck with it. With AWS, if something doesn’t work, you can fail fast. You can change the size of the server or try something else without the heavy capital investment in purchasing servers, licenses, etc.
Cloud pricing is more complicated. You pay for the server, but you also pay for network traffic, making it difficult to compare directly. Even if the cloud server is more expensive than buying a physical one, consider the maintenance costs: hiring staff to manage and maintain the server.
Of the electricity, cooling system, and if something breaks, we have to take into account opportunity costs and capital savings. You’re not parking $50,000. You can have that $1,000 to do other things. It’s not an apple-to-apple comparison.
OK, so let’s move on, there’s a lot more content. Why should you care about the cloud? The main reason is that if you’re heading into an IT career, you need to go where IT organizations are putting their money. The total size of the cloud computing market has grown exponentially in the last 12 years, and it’s not going to stop. Millions of companies are still trying to figure out how they’ll get there, what to prioritize. There are a lot of opportunities in that.
This is the public cloud computing market, so it’s not just AWS. It includes service providers like Salesforce, who rent you space to host specific applications.
All this data comes from a survey asking IT organizations how they use the cloud, where they put money, what types of applications they use. One important question is, what are people using? Are they 100% in the cloud or still on-prem?
Are they using multiple clouds, all that stuff. In a 2019 survey, 94% of respondents said they are using the cloud. Companies of all sizes, 94% are using the cloud. However, of that 94%, 22% are using exclusively public cloud, like AWS or Google Cloud. A higher percentage of 69% is using hybrid cloud, which is where you should see a bigger opportunity.
Most companies are in the cloud, but most will stay in a hybrid situation for reasons we’ve already discussed. Some operations are more expensive in the cloud. A lot of applications the government uses were written in languages that are now outdated. COBOL is one example. At work, we’re paying more to migrate an old database that isn’t cloud-friendly.
Nobody can deal with it, yeah. So, if you have these applications, they are called legacy, and by the way, I said the government, but not only the government has them; enterprises of all sizes have them. These are applications that were built in 1989, and they made a couple of mainframes. They made a couple of updates to it, but since 1992, they haven’t found one person to work on that application. So, guess what? Nobody can migrate it. For them, it’s easier to find people to write that application from scratch than to migrate it to AWS. So, guess what’s going to happen? That’s going to stay on-prem until that server dies, literally, and they’re going to babysit that until they can. So, there’s still the cost of maintenance and everything else that goes around it.
So the other thing is that there are certain types of tools that are not cloud-friendly. If you’ve worked in any office, you’ve seen that there’s usually a printer server. It’s just easier to have that thing on-prem. So why would you put your printer server in the cloud? It just doesn’t make sense. Your phones, if you have voice over IP, there’s probably a server that controls that. That’s also going to stay on-prem. So, there are different workloads, and they fit in different places. Raymond said he worked with Office 365. I work in an environment where we’re like pro-AWS. We don’t want to know about Azure, but we use Office 365. If we don’t find a different tool than Excel, Microsoft Teams, and Word and all that stuff, it’s very unlikely that we’ll be able to move away from Office 365 also because of email. They provide our email service, so we’re never going to put that in AWS, and that’s where it’s not really showcased here, but there’s a different category called multi-cloud.
And that’s when you’re using more than one cloud. If you have 365 and AWS, you’re multi-cloud. If you have AWS but also have things on-prem, that makes you hybrid. Some companies have private clouds, building their clouds for themselves. Alan, our other teacher from Jira One, specializes in private cloud. He is a private cloud expert and builds them daily.
Was that like VMware and all of that? Not really. VMware is used as a hypervisor, and it’s probably used in AWS, though they won’t tell us. It’s also used on-prem in data centers. Other tools allow you to treat your data center as a cloud. You access everything through the internet. Sometimes, there are billing models where you can say, “Charles’s team deployed a server here, so we’ll charge him this many dollars per hour.” That’s usually the private cloud model. However, that’s smaller than the public cloud model because not all companies want to do this.
Think of this like something the Department of Defense (DOD) would do because they have many regulations to handle. Before trying to put data in AWS, they might decide to build their own system. To answer Gabriel’s question another way, VMware is just virtualization of servers, but cloud computing has other components besides virtualization, like networking.
Private cloud is essentially the same thing as having your own data center. The difference is how the infrastructure is presented to the company. In a private cloud, the company still manages everything itself but presents it as a cloud-like service.
Regarding the cloud market, data from 2019 shows that Amazon leads in market share, with a huge number of users and more services than other providers, making AWS the most dominant. After AWS, Microsoft Azure holds 18% of the market, and it’s widely used for specific workloads. For example, if you had a server running Windows on-prem, it would be easier to migrate it to Azure than to AWS. However, for big data or analytics with multiple types of databases, AWS is usually the better option.
Google trails with 8%, while other players include IBM, Alibaba, and Salesforce. A trend seen in companies like IBM is that hardware providers now offer servers that automatically send data to the cloud, creating a link between physical servers and cloud storage. For example, Azure’s 18% market share comes from users purchasing Microsoft Office, where the company offers to store data in Microsoft Azure for free. This often leads to customers moving to the cloud without realizing it, and then Microsoft begins selling additional services. This strategy could inflate the numbers a bit.
I think so. But I may be biased, I’m not sure. As I was saying, there are several cloud computing models. The first one is the SaaS model. In the SaaS model, you are using a service that already exists. For example, think about Webex. We are using the Webex service; we’re not building anything on Webex or migrating anything to Webex. We are simply using something that already exists. Similarly, when you use Gmail, you’re using a service. Gmail is a SaaS. We don’t install anything on our computers to store our messages. Another example would be gaming platforms like the PlayStation Network. That’s also a SaaS.
This shows how these are different from other types of cloud providers. Then we have PaaS, which stands for Platform as a Service. An example of this would be Salesforce or platforms that allow you to build digital training.
They give you a platform where you can use different building blocks to offer something. For example, an LMS (Learning Management System) or QuickBooks is a SaaS—you’re just using it. If you’re building things on it, then it’s a PaaS. If you’re migrating things to it, then we’re talking about Infrastructure as a Service (IaaS), which is what we care about most.
IaaS clouds offer servers, storage, network, security, and system management. All companies are trying to expand into other models, but they are essentially infrastructure as a service providers. This applies to Azure, Google, Oracle, and others.
The pay-as-you-go model mostly applies to Infrastructure as a Service (IaaS). For SaaS, the payment structure can vary. Sometimes, you pay a monthly fee, or you may need to buy licenses. For example, with PlayStation, you pay a yearly fee of $50 to $70, which is a subscription, not based on usage. In contrast, with IaaS, you pay based on the servers you use and the amount of storage you consume, which is the true pay-as-you-go model.
Zoom is a SaaS, not a PaaS.
We’ve discussed cloud computing models, and now we’ll focus on cloud deployment models. This refers to where you deploy your cloud. There are three main models: hybrid, public, and private. These models allow for different levels of scalability and control. For example, if a company like Macy’s has data centers on-prem, their scalability is limited to what they can afford. However, for a company like Netflix, using a public cloud, scalability is practically infinite because they can scale across multiple clouds and regions.
But scalability is not the only factor. Control is also a significant consideration.
When we discussed government data, it’s clear that organizations like state governments often require more control over who can access data and how platforms operate. For this reason, private clouds are heavily used in sectors such as government, healthcare, and the tax industry, where sensitive personal identification information is involved. Private clouds offer greater control over data security.
Hybrid clouds provide a combination of both scalability and control. For instance, a public cloud can host a website like Obamacare, providing the necessary scalability for traffic, while the customer data is securely stored in a private cloud, ensuring sensitive information remains protected. Cloud security has significantly improved since 2014, with providers now complying with many more regulations, making it easier and safer to store personal information in the cloud.
In general, public clouds offer scalability, while private clouds offer more control. This fundamental principle remains consistent as organizations choose between different cloud deployment models.
The first one is the principle of agility. To be honest, this is where I think cloud makes the difference. Charles was asking me like, what’s one principle or concept that everybody adheres to. So I think this is the goal. Agility is always the goal. Traffic can really show what agility means. So in the current state of IT, like if we’re, if we’re managing stuff in our own infrastructure, we’re probably spending 80% of our time. Money just running whatever we have.
They need to run that, secure that, patch that, and maintain that. They need eyes on it to ensure it’s up. So, Try’s team probably only has 20% of their time to innovate. In another world, which is the desired state, it should be the opposite. We should spend more time innovating than sustaining and running. In a world of unicorns, we should all be paid to constantly figure out new things. That’s the ideal world.
The other change is that in the current state, we tend to see IT as a cost center. These guys are spending all the budget because they need to pay an electrician, improve their cooling systems, and handle other tasks. But if we’re innovating more than we’re running, we’re generating more income, so IT is seen as profit generators instead of a cost.
The principle of agility means that by removing some of those burdens of running and sustaining environments, we can focus on performing our mission. Think of it like the pizza example: Why make pizza when someone else knows how to do it right? Enterprises, like companies in the medical industry, are better off running their hospitals than designing applications. By focusing on their mission, they can innovate faster, reduce delivery times to business users, and release changes in an agile way. They can implement small improvements gradually, reduce internal costs, and fail faster.
If a startup notices an application won’t succeed, they can kill it quickly because they didn’t make a big capital investment. They only pay the AWS bill for a few months. The same applies to our work.
At work, I can spin up as many instances as I want. There was one day, in my second month, when they told me to stress test an application and asked me to deploy 3,000 servers. I was unsure how to do it, but it only took three lines of code. We had so many servers that we had to contact Amazon to ask them to increase our limits. We were able to deploy them, and guess what? It didn’t cost us $500. We ran the servers for an hour, powered them off, and that was the end of it.
This is something you can’t do on-prem. Cloud computing lowers the admin overhead. On-prem, someone would have had to give me an IP address, configure things manually, and ensure certain agents were installed. There are many maintenance tasks performed on a server before it can go into production. But with cloud computing, we can automate the process, which significantly reduces the admin overhead. We also spend less time on maintenance and upgrades.
In my environment, we try to build everything as an auto-scaling group. This means we don’t deploy a single instance but a group of instances, which can even include a group of one, which is a funny concept. The cool thing is that instead of patching the server, I can just click on it and choose “Terminate,” and it will terminate that instance and build another one that looks just like it. Instead of constantly having to patch and evolve a server, I get the latest and greatest every time I interact with it.
Another benefit of agility is that we avoid vendor lock-ins. There are legacy applications that are untouchable because nobody innovated on them for 20 years. Now, 20 years later, no colleges teach those languages, and they are effectively locked in. I remember in my previous job, we had applications built with programming languages that were unsupported, meaning the vendor didn’t even sell licenses for them. That application was mission-critical and would only run until the day it failed—there was no way to bring it back.
For the record, I know COBOL. In the banking field, we had this program written in COBOL that we tried to translate into other languages, but it failed, throwing a lot of bugs. So we went back to COBOL. Once you learn one language, you can learn another.
The second principle we’re going to learn about is elasticity.
Elasticity means the ability to acquire resources as needed and release them when they are no longer required. I mentioned auto-scaling groups earlier, and if Macy’s were running in AWS or any cloud, they could use this feature to automatically grow when there is more demand and shrink when the demand decreases. This reduces admin overhead because there would be no need to monitor every single system and approve adjustments manually. This ensures that the system will not crash if 300 extra users visit the website on a given day or use the application.
In the cloud, you want to be elastic and scale automatically, with the ability to scale in multiple directions. For example, scaling out horizontally means adding more servers. If one server is performing a function, you can add two more, and then a third if needed. However, there are situations where scaling out alone is not sufficient, and you also need to scale up or down. For instance, if a server has 4 CPUs, you might need to upgrade to one with 8 CPUs, similar to building a double-decker bus.
The other principle is high availability, which improves for IT organizations when they migrate to the cloud.
On-prem, there’s usually lots of outages. If your data center is in your building and they’re performing electrical maintenance, those systems may go down. Because you are constantly patching every month, the state of your server changes. Your ideal definition of the server and what it becomes starts to separate. If you run that same server for 5 years, you may install junk on it and never remove it. After 5 years, when you do a patch, your server may break, and you don’t know the actual state of the server that worked. In AWS or any cloud, because you set things up as code, you keep track of the ideal state and tell it, “I don’t want to log into that server to make changes. I want to build another server to replace it.” That takes less time. I can do that with a couple clicks instead of days of work.
High availability refers to a system designed to avoid service loss by managing failures and minimizing downtime. It’s a goal. We try to make systems highly available, but not every system needs to be. Think about critical systems in your organization. Imagine Amazon’s mobile market and search tool stops working. How many millions would be lost in an hour? Imagine the US Postal Service and the system that prints labels is down for a day. Companies identify critical systems and make them highly available by using redundant hardware and software. If one server controls printing, you might have one in every state, with redundancy in place so if the New York server fails, a Jersey server can take over.
Also, another interesting thing about high availability is that it encourages companies to be honest with themselves and identify things that can’t fail, because things will fail. I remember earlier I mentioned my work in disaster recovery. Sometimes, availability and disasters don’t stem from negative events. I’ve seen news stories where, around Christmas, teams decide to play the lotto and actually win. Nobody shows up the next Monday, and if they were the only ones who knew how to deal with the systems, there’s an availability problem. Becoming highly available involves figuring out who can support these systems.
Well-designed high availability systems avoid single points of failure, which can be technical, electrical, knowledge-based, or practically anything. Any hardware or software component that can fail should have a redundant component of the same type. For example, a hospital needs electricity, so most likely, it has a backup generator. The same applies to IT operations. If you have a data center, you need a backup generator. Hopefully, cloud providers have many backup generators, which they do, improving availability when moving things to the cloud.
This map shows AWS’s presence worldwide. As you can see, these are different regions.
Here they have the Oregon and California regions, on the east coast they have USC one and Ohio, then they also have the Gulf cloud with two locations. By using a cloud that has all these different locations separated geographically, if something happens to one, like a natural disaster such as a flood in Virginia, it’s very unlikely that the same flood will affect Ohio. By using two different regions, you’re improving the availability of your systems because whatever happens to one will not affect the other, and you’ll continue to run from the second region.
Are you saying that if AWS is down, we have a bigger problem? No, I’m saying that if they’re down in one place, most likely they’re not. If they are, we have a bigger problem. It’s like a whole energy grid being hit. Something crazy, like 80% of the internet traffic in the US goes through Northern Virginia. But if a crazy hurricane happens in Virginia
No, but I think that also brings me security. That means we’re in a very secure area where floods and such are unlikely to happen. Makes sense. Because they put a lot of thought into these things. I have a question. By the way, you’ve been doing a very, very, very good job with these explanations. The elasticity parts and the high availability, could you give physical examples between the scale-up and high availability? I understand that when you scale up, you fill up a gap, like you said, acquiring more resources, but I know that in high availability, we’re talking about networks and such.
OK, and elasticity, we’re talking about the units that are performing the work. The concepts are related, but they don’t mean the same thing. Elasticity allows you to become highly available, so it improves your availability. High availability is a goal, so you want your systems to be highly available to prevent failure. One way to get there is by making sure your workloads can be elastic. When you scale out, you’re increasing the number of units performing a job. For example, you need another fridge in your house because you have three kids who play football. You decide to get another freezer. But another option is buying a bigger fridge, like a two-door one.
I understand. High availability, another way to think about it, is like our presidency. We have a president, vice president, and then the speaker of the house. If something happens to the president, the vice president steps in. If something happens to the vice president, someone from Congress steps up. That’s a highly available system, and that’s the idea. Things that are critical, whether a computer system or a societal system, we create backups. We say, yes, something can go wrong, so what can I do to continue functioning, at a lower or same capacity, when things go wrong?
Elasticity relates more to redundancy, and high availability would be failover and backups. Is that safe to say? No, I think elasticity is more about scaling. Scaling on demand, shrinking or expanding as needed. Highly available means redundancy, which ensures availability.
If something goes down, you’re good, but elasticity is when the load picks up, you don’t have to say, “hey, I want two new servers,” that happens automatically.
I got you. They complement each other, but they’re not the same thing.
This is the principle of fault tolerance. Fault tolerance is the ability of the system to continue without interruption when one or more components fail. Let’s say only Charles knows how to perform one task in the organization. What should you do to make that organization more fault-tolerant?
Make sure other personnel are cross-trained. Exactly, so Charles can go on vacations in peace, or you can add redundancy. Or get a second child. As hard as that is.
You’re trying to prevent problems that would come from Charles going on vacation, or Charles winning the lotto. Fault tolerance ensures high availability and business continuity. This is the capacity to continue operating under any circumstance. I worked figuring out how to do disaster recovery. Disaster recovery ensures your business can continue during a disaster.
Let’s say we live in Maryland, close to DC. If the area floods, like all government organizations with data here were unavailable, what could they have done to prevent losing data?
Put them on the cloud. If you have a house in Florida and Florida may flood one day, what’s your plan?
Have another holiday home somewhere else.
You have another house somewhere else.
The furniture, the same design, so when you move to the new house, you feel like it’s the same. I would just get flood insurance.
I would just learn how to swim. A second place is always the best.
You would hope they invested in backup somewhere else. You don’t necessarily need another server in another region; you could just copy the data to another region, so you can be back up in hours or days. Organizations have plans that say, if we have a disaster, this application needs to be up in 6 or 12 hours, but this low-priority application can be done in 2 weeks.
What they say in those plans also determines the effort put into building fault tolerance. If they tell me I have 6 hours to restore, I know I can’t just copy the data. I need to also have a server in that other region, even if it’s off. I need a server that can use all the data, and everyone in my office needs to be trained to turn that server on if needed.
Fault tolerance sometimes goes beyond that. It even involves having people outside your department know how to turn on those servers. If it’s flooded and everybody in Maryland is affected, we need someone in a different area to perform the task.
Companies usually have a safe where they keep documents with step-by-step instructions on what needs to be done if something fails.
Imagine a database of customers.
Information can be replicated to another machine. If the primary database goes down, operations can automatically be redirected to the second database, either automatically or by a person. Of course, automatic is more elegant.
Is that where AWS or any cloud becomes better than the traditional data centers? If a disaster happens on the East Coast, all the engineers don’t have to travel to the West Coast to rebuild the infrastructure. They can stay at home and access the West Coast cluster, and with infrastructure as code, they can build it faster too.
As I mentioned with power sources, if your electricity goes out, hopefully you have a generator. Another non-technical example is tall buildings. Any tall building has an elevator, but what else does it have? Stairs. Sometimes, the backup system isn’t as good as the original; in the case of a power outage, the second elevator wouldn’t work, but the stairs would.
We’re getting to the end of this. Someone asked me what AWS is, so I’ll tell you. AWS is the world’s most comprehensive and broadly adopted cloud platform. They have most of the market share and more services than any other cloud provider. There are things you can do in AWS that you simply can’t do in Google Cloud or Azure, and that’s the main benefit. It is considered a cloud services provider or infrastructure as a service (IAAS). It has over 200 fully featured services from data centers globally. If you have regulatory requirements where your data needs to be in Africa, they can provide that.
So, who uses AWS?
Right now, millions of customers globally, ranging from startups to small, medium, and large enterprises, government agencies, and nonprofits. They have something called FedCloud, which has government certifications, so it’s considered Fed ramped. That means government agencies can spend government money on them. The types of use cases in AWS are very unique, which is positive because you want to use tools that have a large community.
The reason for this is that you always need documentation. It doesn’t matter if you’re working on a Linux server or in the cloud; you want to leverage what others have done to accomplish your tasks. If you’re talking to a vendor and saying, “I need to collect logs from all your accounts,” and they respond with, “We haven’t had a customer who did that before,” you’re in trouble because you’ll have to design it from scratch. But when there’s a healthy community around a tool, it’s easier to build on others’ work and share information.
Why do customers use AWS?
Most customers use it to lower costs. They are moving away from investing in physical servers, maintaining cooling systems, and real estate just for those servers. They’re also becoming more agile by not having to manage these data centers.
They can shrink in a certain way, and can also dedicate themselves to innovation and not operations.
This is gonna be like a very quick overview of some of the services. As you can see, the services range in different categories. There are foundation services that are like the ones that we know about the most. Like, I’m sure everybody here has heard about S3 or has seen it on their, like, sometimes you’re downloading a file from a website and you see Amazon AWS and you’re like, wait a minute, I’m downloading a form from the government. Why is this an AWS? The reason is because it’s stored in S3.
They have other services that are capstone like EC2, those are their servers. So basically, the same server you would have in your environment, you could have it in their environment, and it’s called EC2 or any EC2 instance. Other common services are RDS and Dynamo DB which are two types of databases, but as you can see, they have other databases services. The VPC service is pretty famous cause any EC2 instance needs to live in a VPC. And then, as you can see, they have services that are very specific. So, just because I have 5 AWS certs doesn’t mean that I’ve touched every one of these services. It’s like the same thing as Linux.
You’re gonna learn a lot of Linux, but you can do a lot more with Linux, and it just depends on what field and industry you fall in and what role. There’s services like Kinesis that basically Handle like logs and information coming from systems. So, who has a smartwatch? Yeah. So that’s considered an IoT, an internet of things. Your TV if it has access to the internet, it’s an IOT device. So all those devices are constantly communicating back to a system.
And they need logs. If Samsung sells you a TV, they’re keeping track if you touch that TV or not. Your TV can be leveraged by hackers to perform an attack. All these systems are communicating back to other systems in the cloud, and Kinesis can help in those cases. There’s way more than that.
Other deployment and management systems include CloudFormation. CloudFormation is Amazon’s native, I’m gonna call it language, but it’s not really a language. It’s how you declare things so that Amazon can then build it. It’s infrastructure as code service. OpsWorks is the implementation of a tool called Chef and another one called Puppet, offered in AWS in a way where you don’t have to manage Puppet or Chef.
There are also a bunch of administration and security tools. If I had to say that I specialize in something, it would be here because this is what I focus on at work. IAM is how you grant access to your accounts, how you secure your account. Trusted Advisor scans your accounts and sees if you’re following good practices, if you have any security loopholes, or if you’re almost reaching certain limits.
If you’re using 50 instances of a certain size, and you have a limit of 60 on that region, it tells you, “Hey, you’re reaching that limit, you should consider increasing it or using a different server size.” Config tracks the configuration of things in your account. If today you deploy a server with 4 CPUs, but tomorrow, France changes it to one with 8 CPUs, it stores that change. If tomorrow, David changes the security group or firewall attached to that server and opens it to the world so that he can log in from his house to install some Bitcoin mining stuff, Config lets you see who did what, where, and how the server evolved over time.
CloudTrail tracks all API calls made to AWS. Every time you click somewhere, that’s an API call. You’re interacting with something giving you information, identifying yourself every time. CloudTrail tracks that, and if I click on “Tell me what are the users in this account,” I’m performing a list account operation, and it tracks that. If I go and create a user, it tracks that. Config leverages CloudTrail to put the picture together of what happened in the account over time. CloudWatch is just a monitoring tool, where you can see, “Is my server using too much CPU? Is it using memory?”
How much networking resources am I consuming? That would be there. Then you have other edge tools, like in application services, you have Elastic Transcoder. I’ve never touched that, but I know some companies use it to transcode documents. If you’re a media company that produces music and you want to have MP3s and different formats, you would use the transcoder. Or if you’re Netflix and you’re taking very old movies that weren’t in a certain format and putting them in a format where you can share them on Netflix, you could use that. Then there’s Cognito. I’ve only played with it while studying for certifications. Many applications with public users use it.
Imagine you have a game Cedric created, and what’s the one that was popular last year? It was a dude from North Carolina, and all the kids were using it. Was it Candy Crush? OK, Candy Crush, fine. It’s not that one, but that works. Candy Crush can create a pool of users. Everyone who creates a Candy Crush account gets registered in Cognito.
In Cognito, you declare that user Juvie has access to Juvie’s points, and if Juvie interacts with Martin, Juvie can start having access to certain parts of Martin’s profile. That’s what Cognito is for.
But as you can see, there’s a world of possibility, and as I said, there are way more than 200 services. I think now they’re closer to 400, but every one of these services has hundreds of features. So, the things I can do with RDS are not just 1 or 2, they’re hundreds. It’s a very large list. This was just a teaser so that you guys can see some of the services and their categories, as well as the most common ones.
cloud environments. So, it’s gonna be different and it’s gonna be a little more intense in this class, in my opinion.
