0:00
Hello, everybody. Welcome back to the Cloud Show. This is the show where we talk to leaders in the cloud or in cloud projects about things that are important to know or important to have a handle on when you're running a project towards the cloud
0:18
In this show, we have plenty of great guests, and today is certainly no exception
0:25
We're going to talk to the founder of C-Shop Corner, Mr. Mahesh Shand, and we are going to talk a lot about how to manage costs for the cloud
0:35
I think that could be number one question for any cloud project that's going to the cloud
0:41
It's about how do I manage costs and how do I keep control of my costs? So today will be a great show here at The Cloud Show
0:49
Let's take it away. Mahesh, welcome to The Cloud Show
1:03
Hey, Magnus. Thank you for having me here. Thanks. Yes, my pleasure. Definitely
1:08
So good to see you and so good to talk to you again. Yeah, good to see you too
1:12
Yeah. So you're back home in Philly. right now? I am. I am. I was in India and now I'm back home. It's nice. Summer is starting here
1:22
It's nice weather, beautiful. Everything is green and yeah, this is a good time
1:27
Absolutely. All right. Cool. So on the topic of cloud leadership, I know you have significant
1:34
many years of experience running projects for cloud and in various contexts for different
1:39
customers and so forth. I'd like to talk to you about cost management. This is the number one
1:47
question, right? It should almost be the first question. How much is it going to cost? And the
1:52
challenge we're facing is that costs in the cloud are calculated a little bit different than before
2:01
IT budget used to be this multi-million budget item that people would have per year
2:08
but now with cloud it's kind of different right it's it's a pay as you go so um maybe we could
2:16
start by talking a little bit about what kind of factors factor into the cost of of cloud services
2:24
yeah definitely definitely so you know a little bit before that even you mentioned that
2:29
why most people migrate to cloud is right look at the five top things in my list is
2:35
reduce cost that's number one then scalability is another issue right because when your traffic is
2:41
customers are growing if you have in-house hosting it takes forever right to scale that
2:47
and then faster to innovate and go live and then security and compliance is big big thing it's
2:53
already built in the cloud so you don't have to do anything and then disaster recovery so those
2:58
are five top things why people move to cloud but as you said cost is one of the biggest factors
3:04
That's why most companies' business move to the cloud. Yeah, and this scalability thing is also related to cost
3:13
Having an appropriate scale is what can vary your cost according to the workload or the user load that you have right now
3:21
Right, right. And keep in mind, by using cloud, first thing, if you look at the older way, you used to have full-time people managing your servers, right
3:32
Your networks, your security. and your databases. So now just moving to cloud
3:38
out of sudden you don't need that full-time, you know, full-time resources
3:42
Yep. That automatically saves you money. Yes. Right? Then second thing, if you look at the scalability
3:51
and if you don't use cloud, then you have to buy the hardware, then you have to set them up
3:56
then you have to deploy the application. Cloud is already ready for you. All they have to do is push a few buttons
4:02
and it's ready to go, right? You can add, remove hardware. You can add, remove resources
4:08
You can have more compute powers, bandwidth, storage, anything you can do
4:12
Everything is already there. It's just pushing a few buttons, right? That's all it is
4:18
Yes. So, yeah. And so we were looking, in fact, at the end of last week, we were looking at a migration project
4:25
And one of the things that stood out were things that we can probably migrate to the cloud
4:31
and really shortcuts certain things. Like, for example, before you would have a SQL server and databases
4:40
and you would have a separate setup and you would copy things between and so forth
4:47
What really strikes me as a massive optimization of costs are the sort of out-of-the-box pass services that basically
4:54
or literally have all these features just included at a click of a button
4:59
Yes, that's exactly it is. And, you know, you know, we I have worked on a project where we have using we use all three
5:05
You know, we use infrastructure, we use PaaS, we also use SaaS. Yeah. So that's one thing is when you are as a company, you are migrating to the cloud
5:15
You have to make sure that you're you have a clearly defined requirements for the cloud
5:22
Yeah. OK. You cannot just say, oh, let's move everything to cloud right away
5:26
You have to have a strategy You have to have planning make sure as a business you have to have somebody expert who has done that so what i see a lot magnus is is that a lot of these existing it company they go to cloud but they just
5:41
shift lift and shift from their local hosting environments to cloud and that's when it
5:47
can end up costing a lot because cloud doesn't work that way yeah yeah definitely i mean it's
5:52
it's probably gonna work right it's gonna run it's gonna be functional it's gonna be running
5:58
and it may even be slightly better or or maybe quite a bit better than what they had before
6:03
depending on whatever deal they had before but it's not gonna be great is it there are some
6:09
some things you need to do to get to make it great so there is a large company here in the
6:13
us and they migrated their 30 000 applications to cloud and their cost was only 20 times than they
6:20
they were paying before now now that they got panicked their ceo got panicked like why are we
6:27
moving to cloud it's because their team what they did they took exactly they they literally created
6:33
virtual machines exactly same configurations start deploying databases on the same so this is
6:39
sometimes cloud can be costly if you don't plan it right and you don't have right resources okay
6:46
So the key is to plan the migration. Sometimes you migrate things as is from a virtual machine to another virtual machine
6:55
But the key is to find out which things you change in the migration to move to platform services
7:02
Is that correct? Yes. I say it all the time. One side doesn't fit at all
7:07
Some applications and products or some, you know, for us in our case of C Sharp Corner
7:13
some we have virtual machines because they are cheaper for us. rather than pay per use
7:17
Some are good because we only use so much limited resources. Then you put them as a resources, like Azure cloud resources, right
7:26
Some, then you use platform. It's already there. If a SQL, Azure SQL is already there
7:31
all we have to do is $99 per month rather than buying that annual license
7:36
So you have to look at one application, one resource at a time and see what fits best
7:42
That's how you can save a lot of money. Yeah. And that's probably where they went wrong when they had 30,000 workloads, because it's so much that they felt that we'll just migrate it first and then maybe we'll look at optimization later
7:56
The company made a mistake is that they said, oh, we have to migrate everything in 24 months. You can't do that. This is not how it works
8:05
Well, I mean, if you're willing to pay for it, but that maybe wasn't the case
8:11
So let's talk about scaling here because that is such a key item
8:15
Of course, you can use virtual machine scale sets to scale your virtual machine instance count
8:20
If you have a large workload, a lot of users, you can scale to more machines and you can scale in to less machines when you have, you know, during the night or something where fewer people are using your system
8:32
But of course, scaling, it works even cooler with the platform services like app service and various things where scaling is just built into the system
8:42
So let's talk about what the scaling bring to us in terms of that scenario for cost optimization
8:49
Yeah. So one thing, you know, any business, they are migrating first time to cloud
8:55
They have to understand that cloud operating system is designed and architected differently
9:00
it's designed to scale from day one it automatically knows that your application
9:05
needs more resources it's already compute powers you need more storage it will let you know and if
9:11
you configure it right it will automatically add those resources right you don't have to do much
9:16
or you can do on demand however you want to set it up those settings are already there
9:21
so this is one thing we have to understand is the scaling is already there sometimes
9:28
when architects are doing they just don't know all the options so knowing all the tools and
9:34
options is very important I'll give you an example when you know we come from this background where
9:40
when we first time say what kind of machine we need or what kind of storage we need or what kind
9:45
of compute we need we all goes we always go for the top range like oh when we are at the top in
9:51
peak time we will need you know i don't know 120 gb ram and we need you know but in scale in that's
9:58
how we configure it we can start configuring our new machine or new resource based on that
10:03
in in in case of cloud you always want to go to the reverse side you want to start with the minimum
10:10
okay whatever minimum you need so that's the that's the off time and then you automatically
10:17
say it's auto scale, it's auto all these features. So when your, you know, your application is going
10:22
to need more resources, it's going to automatically take that and then it's going to release that
10:28
Exactly. And that's, I think, it's kind of the reverse of how we were doing IT planning in the
10:37
past, right? Yeah. Because in the past, you would say at peak load, we need this much server, right
10:44
This is the amount of server we need, CPUs, RAMs, whatever, right
10:48
And then capacity and things. So we have to buy hardware or provision for that
10:56
And then that was what you had for all the time, like even during the weekend or the night when you had just a fraction of your users you would still have the same amount and you couldn scale it to less You always paid for the whole thing
11:12
That's how we as an industry have been taught to do things
11:18
Now we're doing the reverse, right? Yes. Now you start with minimum and then it automatically scales
11:25
That's the advantage of scalability that you don't have to do much. it's going to automatically use and you only pay as much as you use
11:31
That's the best part. Yeah. Yeah. And I think in that, in that one other thing you can add is that one
11:40
one thing most importantly, businesses miss, especially the larger corporations. I work with larger corporations where you have several thousand employees
11:48
They have say 500 developers working, there's DBAs working, there's architects working. So in the beginning
11:54
there's the IT team that sets up the cloud requirements. Okay. And then they give access to consultants like us and developers
12:01
And they say, oh, just this is your account. You can keep adding as many resources you want, however you need for your applications
12:08
That's where businesses get wrong because not everybody is educated on cloud
12:14
And because they're not educated, you know, I see some consultants. They just go like, oh, let me add another resource
12:20
They say 8 GB RAM, 12 GB RAM, 16 GB RAM. And they use it or not
12:26
It's going to cost you as a business. Business has to pay for it as soon as you provision that
12:32
But so in this case, what businesses have to do is they have to have an enterprise-wide strategy and awareness
12:39
Even you are a tester, even you are a DevOps, even you are a finance guy, you should at least know how cloud cost works
12:48
If you build that and I'm logging in as a developer, I always have in my mind, okay, make sure I don't use these resources because it will be added automatically when I need
12:58
it. Yeah, yeah, that makes a lot of sense. And so I've had at the back of my mind what you said
13:04
there that you should be provisioning for your minimum sort of requirement. And of course
13:10
what you'll need to do is scale to your peaks, you should be able to automatically tune your
13:17
application and scale to the peaks that you have in your workload. Now, a fun anecdote there is
13:24
or a saying that I've had for a number of years is that, you know, I've asked sort of a half rhetorical question
13:31
what's the worst thing that can happen to your application when you put it in the cloud or when you run it
13:37
The worst thing that can happen to your application is that it's successful. Because it wasn't architected for that
13:45
It was architected for a small amount of users and then suddenly you're a huge success
13:50
and you don't have enough resource. But that's something that the cloud can really help you with
13:57
Yes, and faster, like launching your things faster. So another advantage of cloud as a business is that cloud is always innovating
14:09
You want it or not, right? So when you use Azure cloud, Microsoft already upgrading their security
14:15
their provision for new processors are being added, There's new security and then even like speed performance, everything is already being upgraded and it's already there for you
14:28
You don't have to do anything. So you get that as a cloud. But then what happens is sometimes companies get stuck on that
14:37
Oh, I'm a Microsoft shop, so we're just going to use Azure. I'm just Amazon shop, which is going to AWS
14:43
So another thing businesses should rethink about their strategy is that it's like you go for a shop
14:48
You buy jeans from somewhere else. You buy a shirt from somewhere else. You look for pricing. You look for your brand. You look for what you like, what fits
14:56
Same thing here. Businesses have to think what fits them the best
15:01
That's why multi-cloud strategy and hybrid cloud strategy is very important because some things are better at AWS, some are better with Microsoft Azure
15:12
some are better with Google Cloud. So this is how we have to think
15:16
that's interesting and and now with i guess with containerized technology deploying applications
15:23
kind of anywhere doesn't actually matter because containerization is essentially a huge abstraction
15:29
over whatever you're running it on it doesn't matter yep yep and not only that you also have
15:36
to think like on you know some some of the services are not best suited for cloud right
15:43
And one of the examples is CDN services. CDN services are always costly on the cloud
15:49
So, for example, on C Sharp Corner, we have this terabytes of terabytes of these files and articles and PDFs
15:55
If we put that on cloud, it's going to cost us $25,000 a month. But now we are using CDN, it costs goes to $900 a month
16:03
So, you still have to rethink that. Not everybody cannot go to cloud
16:07
You have to always look at that, that there are better solutions out there
16:12
because some companies are just CDN, that's all they do. Yeah, yeah, yeah, definitely
16:17
And so there are so many things to talk about in this fascinating
16:21
and very, very important topic. I'm concerned also with strategies in relation to whether you migrate
16:31
or you innovate right And in relation to reserved instances that you can buy you rent a long contract on compute instances and you get a discount How does that factor into cost management
16:50
So that's very important and that's leasing, right? This is literally leasing versus renting
16:54
When you lease something for longer term, car, for example, you get cheaper because you know
17:00
you're going to drive that car for three years but if you go lease a car for just the weekend
17:04
obviously it's going to be costly right they're charging it daily so same thing applies right like
17:10
every every time it's a weekend you always pay the premium right yes yes so same way with the
17:17
cloud so if you say reserve instance it's now cloud company like microsoft they will create
17:21
your servers and they know that they are already booked they don't have to repackage resell or
17:26
reservice to other clients so they get up to 85 percent cheaper i was surprised that when we
17:34
start using reserve instances for two because we know our needs are same like we have to run this
17:40
we need this storage regardless we need this compute so we created those reserve instances
17:45
two of them and then now we are saving 65 to 85 percent so as a business if your needs are fixed
17:52
you know that you need this was minimum it's always a good idea to go with reserve instances
17:57
yeah it really is and and so um i i'm i'm also you know taken almost a little bit in shock that that
18:04
the the discounts couldn't can be this this significant now which kinds of which kinds of
18:12
which kinds of of infrastructure can you usually get these discounts on yeah so again reserve
18:19
instances are not available for every service because some are running as a pass and some are
18:25
running as a sass so in those cases you may not have those available but when you create a reserve
18:31
instance depending on the zone and region what service they have available you will see different
18:36
options and i think they are working on it there i think eventually you should see more and more
18:41
reserve instances with more resources so typically which ones do we have today like virtual machines
18:46
Yep. You can get virtual machine. You can get the storages. You can get your computes. You can almost. Yeah, you can do SQL. I can get a VM and SQL install or part of the whole package as a reserve instance. You can say when you're creating your machine, you can watch your machine. You can say this is the configuration I want. These are the licenses I want. So that whole thing can be a package
19:11
Wow, yeah. Well, that's really, I would say, I mean, critical information for a company
19:18
If they have workloads that they know they're just going to move and there's no value in rebuilding them or redeploying them to platform services, we'll just leave them
19:29
Then if you know you're going to have them for a longer time, definitely need to look into those discounts
19:38
Yeah. But in reverse, what you're saying is that we should be looking at, well, two things, right
19:45
Migration to or innovation away from infrastructure services like virtual machines onto platform services
19:52
And then configure that auto scaling thing, which is going to make sure that your compute power is elastic
20:01
When you have a large number of users, it costs a lot. And then you pay for more instances
20:08
But when you're not paying for much, so those are like your kind of different axes to stretch
20:17
Yeah. So there's one more feature besides auto scaling. Sure, auto scaling, one feature
20:22
There's another feature called auto shutdown. So there are machines, like in big IT companies
20:28
if there are machines, for example, there's a testing going on, and you know your testing team is going to start from 8 to 5 p.m.
20:36
only 8 a.m. to 5 p.m., there's no reason for you to running those machines overnight
20:41
So you can configure them to auto shut down. So when they're shut down, all of a sudden you're saving your money on those resources
20:49
That's so, so useful. Because what I've seen in a lot of cases is that you need an engineer
20:55
to set up a testing environment. The testers cannot do it themselves. So first you have to like plan for when the engineer is going to set up the testing environment and then they do that
21:05
And then it's like a weekend coming or something, right? So don't touch the testing environment
21:10
Nobody touch it. And then it's like the whole weekend is just there, all of it
21:15
And then maybe the testers will start testing on Monday, right? So the whole weekend you're paying for, it's completely waste
21:21
Yeah, exactly. Exactly. There's a lot of optimization to take care of there
21:26
Well, so, I mean, I think we're at the end of this segment
21:31
It's been a great show because these are very, very powerful and useful tips that any cloud company, any leader in a cloud project need to take into account when they're planning for cost management
21:44
Yeah, I hope, you know, hopefully, you know, whoever is watching this show learned something from this and saved some money
21:51
Oh, yeah, definitely. I bet they do. so thank you very much for being with us today Mahesh and I hope to talk to you again sometime
21:59
soon thank you Magnus appreciate it thanks