Ep 10: AWS Expert Samir Karande on How to Solve the Cloud Cost Puzzle

Matt Brickey
Welcome to the Cloud Currents podcast, a place where you can hear insights from experts who are revolutionizing how businesses thrive in the cloud. I'm Matt Brickey. I'm senior vice president of professional services at TierPoint, and I am responsible both for our consulting and public cloud business units. This episode is going to explore cost optimization strategies and how benchmarking performance can help enterprises innovate their it operations. Today, we're joined by Samir Karande. Samir is the founder and CEO of Zopt, a company providing a SaaS solution to optimize public cloud utilization and spending. Samir has over 20 years of experience leading engineering teams and cloud infrastructure projects across startups and large enterprises like AWS and teradata. Thanks for joining us today, Samir.

Samir Karande
My pleasure.

01:00 - Samir Karande's Background in Cloud & Technology

Matt Brickey
If you wouldn't mind, I'd like to begin. Thank you. If you wouldn't mind, I'd like to begin kind of on your personal journey. Right. Let's get a little bit about you and your experience and what makes you tick. So Samir, how did you first get interested in technology and engineering? Did you start in a technology background or is this some transition for you?

Samir Karande
Yeah, I've been at the technology from the get go, did my engineering, did my masters from Boston University. I was fortunate to work with pioneering technology companies. First ever b two b e commerce server, first ever voice translation system, first ever telecom grade voice response systems. And that just got me hooked onto the technology world, working with pioneers, industry leaders, somebody who was defining the new way of doing things. And there was no looking back after.

Matt Brickey
You know, in your career, Jerry, and kind of looking through your history, you founded a number of startups. Can you walk us through kind of those moments where you were like, I'm tired of my nine to five job, or probably more like nine to eight job, and you wanted to start your own firm?

Samir Karande
Yeah, sure. So you can be tired of nine to five jobs and choose to do 24 x seven jobs. That's how startup journeys are. So I spent more than a decade in corporate world, in technology industry, and then got started on my own startup. This is my third startup. The first startup was very interesting. That was more than a decade ago. We ran it for a few years. It was acquired successfully, so got some test of success. Once you are written by the startup bug, there is no way you can be cured out of that, right? So you have to go back and keep on doing it over and over again. So that's who I am. So first startup was called Mobi Primo. It was mobile, data management. We did pretty well there. As I mentioned, it was acquired. Second was called Msoda.

It was mobile commerce for masses where the masses could be on the wrong side of digital divide, who are not having the privileged big screen interface or Internet bandwidth, and how to make ecommerce like experience available to them. So that was the second startup. We did not do all that well. We couldn't raise enough money in time and we had to pull the plug. So went through both the bands of experiences, good and bad. And the third startup is probably something where I have learned out of my mistakes and trying to build something better than the first two. So this is EOP, as Matt you mentioned in the introduction, and I've been building it for a little over a year now.

Matt Brickey
So when you founded these startups, was it because, you know, something that you saw that the entities like AWS and Teradata weren't really addressing, or is it just something like you said, that you had a couple instances there, firms that were interested in mobile devices and then now you're interested in cost optimizations. They're just kind of curious how these ideas spawn.

Samir Karande
Yeah, absolutely. So when I started my first startup, I used to work with a company called Persistent Systems. I founded their mobile business, I ran that mobile business, grew it to a sizable business, and saw a lot of interactions with users of mobile technology, mobile device manufacturers, enterprise mobility vendors, and enterprise mobility development companies. And I was kind of expert. IPhone, Android were new and people were just coming out of the Nokia symbion, Motorola and Samsung proprietary platforms. And I got into the mobile part because that was the domain which was exploding at that time. Now we have billions of devices on iOS android, and that time were probably 100th of that population in mobile world. And so I saw that as a big opportunity and that's why I went after that.

Coming to the current startup, I've been associated with cloud industry in different forms. My first two startups ran on cloud, on AWS. I was customer of us and that was in year started with 2010.

Matt Brickey
Right.

Samir Karande
So within 14 years into the cloud industry, worked with companies like Aerospike and Teradata where I had significant responsibility for cloud. And were cloud ISV partner and those were across all three clouds, Google Azure and Microsoft, Google Azure and AWS in between. I worked with AWS for over five years and could see a lot of customers, literally hundreds of customers, firsthand on how they onboard adopt cloud, how they mature in cloud, how they make mistakes, how they struggle, how they figure out problems to their initial struggle there are certain problems they able to fix. There are certain problems which are perpetual in nature. And after spending close to 15 years in the cloud world, I understand this space well enough. And that's where after seeing customers struggle firsthand, we decided to form this company.

Matt Brickey
You mentioned that the opportunity to form Zop really came from customer struggles. Now and being in the industry, that struggle is absolutely real. Where do you see the biggest waste or inefficiencies? However you want to define that with companies that are using public cloud platforms today.

Samir Karande
Before that, let me just give you a broader perspective. In all this last 15 years or so, I've been working on public cloud. After talking to probably 500, 700, maybe 1000 customers, I haven't yet met a single customer who's happy paying their monthly bill. That's a very big statement, and I hope I'll be proven wrong someday, right? That's how big the problem is, right? So almost every customer gets onboarded on cloud. Because cloud has a lot of advantages, right? You can deploy infrastructure in few seconds, few minutes, you can have global reach, you can have elasticity, you can pay as you go. And those are all good things. That gives you a lot of empowerment, right? But every empowerment comes with responsibility. And if the responsibility is not handled right, then it could lead to certain problems.

So cost overruns is a problem because people have been entrusted with the responsibility and because the domain is so complex, they don't know how to use that responsibility in a responsible way.

Matt Brickey
Right?

09:01 - Complexity of Cloud Cost Management

Samir Karande
Too many words of reputation responsible. But cloud is not becoming easier with every day. There are more and more newer features, newer services, newer products that are getting added. The breadth is mind boggling. All the three large public cloud providers take pride in saying, we rolled out so many new features, so many new products. And these are, if you consider it as Lego pieces. If you go out and buy a Lego set in market, you have a manual how to put those thousand pieces together. Now, when you're building something on cloud, you have those thousand pieces. The pieces themselves are changing. They are evolving. There are newer pieces getting thrown into the mix, and you do not have that manual. You don't even know what is the end goal is, right, how the end model should look like.

And when you're on that journey, you are going to make mistakes. Even if you're not making mistakes, certain pieces outgrow your requirements, outgrow those pieces. And there are newer pieces which are better suited for your requirement, right? So this is what makes it very complex. So it's a totally dynamic environment where customers use cases are changing on a regular basis, customers usage patterns are changing on a regular basis, and the services, features, products provided by cloud provider are changing on a regular basis. So when you make this as three dimensional dynamic equation, it is very hard to tame the beast. That's where the problem is.

Matt Brickey
Do you think that's more of a function of how organizations are structured in their IT firm or, sorry, their IT department, rather than necessarily the cloud itself, or some combination of both?

Samir Karande
The way I look at it, the root cause is because before you are called expert in any domain, I would quote Malcolm Gladwell, the outlier's book where it says you have to spend 10,000 hours before you are expert in anything. Now, one, if cloud is not your primary objective or you're using cloud for your business, you might be more focused on your business not focusing on the cloud as a technology for that 10,000 hours before you are expert. Before you spend 10,000 hours, the puck has moved. So before you are expert in something, you are required to pick up newer things, either because your use cases are different, your traffic patterns are different, your demand patterns are different, or the cloud provider's technology pattern had changed.

So it's a dynamic amoeba kind of equation which keeps on changing its shape and form and understanding that. And nobody can really say I'm expert in cloud because cloud is evolving so fast.

Matt Brickey
So if you're a CEO and you're in a position now to where your bill is two to three times more than what your CIO or your CTO said, I think a lot of them that we experience is that they don't really understand the difference between traditional it, which was like, well, we used to run data centers and had all kinds of servers and storage arrays and all of that in there. And that changed all the time. They were always asking me for money. How, if you're the CIO or the CTO, do you explain your amoeba analogy to them? That was a good one, by the way.

Samir Karande
Yeah. I think there are other problems too in the mix, right? And I'm a person who gives the analogies, right, so that I can convey the points. So think of our kids, right? We give them credit cards, and those kids start spending that money using those credit cards, right? Let's say you give them credit card in Chuck E. Cheese, right? It's typical seven year old, eight year old you are talking about. It becomes a game for them, right? They are not worried about who's going to pay the credit card bill or they are not even worried or they are not even aware that somebody has to pay the bill, right? It's just a credit card swipe and like magic, they can play the next game. So think of developers and DevOps as those kids in Chuck E.

Cheese where they are responsible for spinning up infrastructure and they are not responsible for the final build that comes at the end of month. Now if you just expand that and see what the priority for CIO or CA for CEO is and what the priorities for the developers and DevOps, there is a C difference that leads to the cost overruns. That's one of the problem. Second is the easy approach to fix any problem. I have additional 20% demand in my customer pattern traffic pattern, right? What do I do? Do I go back and change my architecture pattern? Or do I throw more money at the problem and say, okay, let's fix it right now and we'll deal with it later? And then that later may not come in near future or ever. Right? So it's combination of multiple problems.

14:48 - Understanding Cloud Costs Like Utility Bills

Samir Karande
And that is because I'll give you another analogy. If you think of cloud as electricity bill, right? The utility bill that you get at the end of month, it's one line item. You don't know how much is the cost because of your ac or washing machine or the microwave and other things. And if you get that breakup, probably you know that I have to focus on something which are the top two or three line items in that bill. Maybe you're using old dryer, maybe the AC is running when it shouldn't be running. Maybe you have to change the mode, certain temperature controls, there are a number of ways you can optimize it, but unless you know what is costing you won't be able to do it.

Matt Brickey
And is that where Zopt enters the picture? Is the concept of why you founded your current organization? Is to bring some transparency to who and what is driving up that cloud spin?

Samir Karande
Absolutely. That's part one of the problem. Part one of the problem is getting the transparency on where am I spending, how my patterns are changing on a daily basis, hourly basis, monthly basis. Then the CIO or the CFO or CEO can say, is my business increasing in the same manner? My costs are increasing 10%, 20%, 30%, but my business is probably increasing at slower pace. So there has to be something that I have to fix. If you're lucky, you can say my business is increasing at 100% but my cloud cost is increasing only 30%. Right? Then that should not be a big worry. So that's 1 second is once you know your costs are increasing, what is causing that cost increase, what is causing that cost overrun? That is where the feasibility comes into picture.

If you are given, let's say, a Yahoo finance dashboard, right? Unless you have enough technical know how on how the stock market and trading works, you won't be able to decipher what is your next investment decision. So that's the second part of what Zop does. Zop not only gives you visibility, but it also gives you automated identification of opportunities that do this. And it is going to improve your cloud. Either performance, availability, reliability, or cost or combination of those four.

Matt Brickey
How do you stand apart from the native solutions of AWS and Azure? I'm sure you field that question probably every day.

Samir Karande
If you look at where the world was in 2020, 2021 versus 2024, or started in 2023, the cloud providers have realized that cloud cost has become a big concern for their customers. If you look at the reinvent speech of Amazon AWS CTO, it was focused on the frugal architecture. How do you build your system architecture, application architecture on AWS with frugality in mind, essentially cutting down the cost, the number of features that are launched. I'm just talking about AWS because CoC only supports AWS at this time in cost Explorer, the cost and usage reports and so on, the pace has accelerated so much in last twelve months, something that we haven't seen in last probably five, seven years, right? Because customers are a lot more worried because of the macroeconomic situations.

And the cloud providers, all three of them are worried that if they don't provide that handholding to the customers, they are probably going to see customers moving to alternate solutions, like going back to on Prem. And we have seen some examples of customers moving back to on Prem. That is not because they don't like cloud, it is because they cannot tame the beast. That is the nature of the problem. What AWS is doing, what Azure is doing, what Google is doing, is necessary but not sufficient. That is where what we provide is going to help customers a lot better than what the native solutions can do.

19:35 - Cloud Repatriation and When It Makes Sense

Matt Brickey
I'm curious. I've heard the same thing you have. So 2023, especially the first half of 2023, was very much around optimization work. The well architected framework. Stay where you are. And then it seemed like some of the media pundits, I'd say towards the end of Q three and into Q four, really started using the cloud repatriation phrase quite a bit more than they had been previously. But when does it actually make sense to shift some of those workloads back to on prem versus just staying put. And I say that because an organization would have invested a lot of time and energy in the effort to move to the cloud. And there's some cost to move, literally just to move, migrate to the cloud, you do not get to erase that cost when you come back out. So I struggle with that.

It's large scale, I understand it potentially where you can run your own data center. But for an organization that is moderate or even high scale, it just seems like the cost to actually implement the move would be a deterrent. Do you see that the same way?

Samir Karande
Let's go back, right. Why cloud started, why people started using cloud. So cloud's underlying principle is stop forecasting your infrastructure requirement. Now, if you don't have to forecast your infrastructure, you can say infrastructure is available on tap. I can leverage it as and when I need. That is the underlying principle. Right. And for that, you have to have certain ways your infrastructure is architected so that you can get that on tap demand handled. Now, if you take cloud versus on prem and if you have to move back and forth, right, it's like moving your houses and it's never easy. So just the moving cost and the cost of opportunity, that could erase the benefits that you are going to have. That's part one. Part two is for only certain kinds of businesses.

Moving back to on prem might make sense because what you have to do is you have to do forecasting of your capacity again. You have to build that capacity in your data center, deploy your architecture, deploy your infrastructure, make sure that it's functioning right. And when your next growth peak happens, you should be ready to handle that peak. Now, if you are, let's say a B two C handling Black Friday traffic, you are turbotax handling your March April tax season. You are Uber, which is handling unprecedented demands because it rained in couple of cities, right. It rains in New York and Uber demand peaks up. Right? How do you predict that? So certain businesses cannot use that on prem forecast model unless you want to start paying for your 100% capacity, whatever may happen as a peak, right.

But then you are losing that capacity and that cost for remaining 364 days in a year. So it's never an easy solution. If you are a traditional b two b solution where the traffic patterns are very predictable, maybe moving back to data center could work for certain businesses, for smaller businesses, as you rightly said, Matt, they may not have the necessary relations with the hardware providers, relations with the data center providers, colo providers. And doing that is hard. Also, let's not forget, you require different skills if you're managing cloud infrastructure, you require different skills. If you're managing on prem, you require different skills. Now you're stuck between wall and a hard rock. So that is the nature of the problem. Yeah.

23:58 - Skills Gap in Cloud and On-Prem Management

Matt Brickey
And I'd say that skill, you hit the nail on the head. I was actually going to ask you a question about that. But that skills gap is, I think in conversations that I have with executives is they're not really understanding that. Is that if you repatriate some workloads, you remodeled your team to be able to run in, let's just say, aws. Those, those skills that you had in running data centers, or even if you outsource that part, they're running a vmware cloud or a nutanix cloud or whatever you want to pick. There's an alternative there. Now you have tools for each one of those, most likely, right? So your software costs go up, your people costs go up. Like I said, it seems a much more difficult problem than just saying you're not happy with your bill.

So let's move you back to where you were before because they weren't happy there either. So give me a little bit of context around, like cultural shifts. Or is there a way to better architect clouds to prevent organizations from doing, I guess, what would be a mass repatriation event.

Samir Karande
So think of situation where you own a single family house, you are not happy with the monthly mortgage payment that you have to do, and you decide to move into a rental place, right? Completely different ballgame, right? Your square footage is different, your backyard, front yard is different. You may not have garage, right? So there are lots of things that come with that decision. Now you're trying to save your monthly payment cost by x, right? Your moving cost will put you back by three months, six months before you can get to break even. And then you probably are not happy with the new way of doing things. And if you decide to move back after a year again into single family, right, then you're incurring that moving cost again.

So that's where people are making those decisions or struggling to make those decisions because it's not an easy decision. Part one, part two is I talked to lots of companies, especially the younger startup kind of companies, who are given credits from these cloud providers. You go to AWs, Azure, Google, everybody has their credit programs, and you are running in cloud a versus cloud b. And the other cloud provider says, come and work with me, I will give you $100,000. They go through similar kind of struggle. And before you realize you have taken a setback on your business because you are focusing on moving that infrastructure, which is undifferentiated, heavy lifting, you could have used those resources for advancing your business rather than moving infrastructure from one provider to other provider, which is not getting your business in any better shape.

You could have some dollars saved or you could have earned some free credits which evaporate in few months, few quarters. So I think one has to take a longer term approach and evaluate how can I make the existing infrastructure better? How can I get to the world where I'm more efficient in using that infrastructure rather than trying to change the infrastructure because I cannot handle the current infrastructure. I think that's what it boils down to.

Matt Brickey
So I'm sure you've heard a similar thought as this is that if I just containerized my entire organization, now, I can move all of these applications between AWS, Azure, Google, and I can take advantage of spot rates, and I have not seen a lot of that materialize. So give me your view of how containers can help in the current situation with this kind of problem that we're talking about. And then where do you see that 510 years? So just get the kind of how do I use it now? And then where do you see that the container vendors could really help the user community in the future?

28:27 - Evolution of Cloud Infrastructure Technology

Samir Karande
Yeah, absolutely. So we grew, with my gray hair, you can say I have seen the bare metal world, right? Then we got into the virtual machine world, then we got into the container world. So we are dividing the bare metal capacity into smaller chunks, adding multiple layers on top of it. The advantage with virtual machine and containers is it makes it more portable. Definitely, right. Certain workloads can get benefited with the containerization. Your application tier, your middleware can run in a microservices environment, smaller components which are running in containers, and you are having scaling advantages at individual smaller component level. Your inventory lookup is hard or your pricing engine is hard, right? You can spin up more containers and you can handle that if you look at the other workloads.

If you want to have, let's say aiml job, if you want to have a database, those are heavier operations which may not be really suitable for containers. So containers is not the solution for every kind of workload. It works beautifully for certain kinds of workload advantage. You have portability, but that is only for partial of your infrastructure, not for 100% of your infrastructure. Second, if your team is very comfortable working with virtual machines, you are asking them to be comfortable with containers. Which is going to require time. You are doing re architecture from monolith to three tier to containers to microservices. That is going to be a cost, right? So I think it should not be a hype cycle. It should be something that one carefully thinks through, that this is an architectural pattern I'm going to adopt for next five years.

And if that is giving out to be all good green signal saying that, okay, this makes sense and this is the investment, this is the ROI, then one should definitely do it. But if that is because everybody else is doing it, so I should. Right. I think it has a limited upside to it.

Matt Brickey
Yeah, I agree. And we saw many organizations refactor that cloud first aspect. So this could be the same thing, container first aspect, just because my neighbor or the CIO that I know down the road is doing it and we've seen that definitely change. So this is your kind of core world. If I'm in containers now and I'm using kubernetes, for example, they're pretty effective, but they're not terribly efficient sometimes in the way that they're deployed. So where have you seen in Zopt where you could really optimize a Kubernetes cluster?

Samir Karande
Certainly. So we talked very briefly about Zop. Right. So let me tell you what we do first before the containers part. So we talked about getting visibility into what the provisioned infrastructure is. What is that you are paying for essentially? Right? So in the cloud world there are two kinds of charges. One where you say I'm going to pay for this, give me one virtual machine, give me one hard drive, give me one database, and so on. And then there are multiple charges, which are for your usage, how many I ops you're using, how much data is being transferred between regions, between region cointinet and so on. And there are charges based on that. There are charges based on how many disk I os you are doing now. That is the usage based charge.

So that is what constitutes in your bill if you're provisioning something, that's easy to say, okay, this is what I'm going to pay at the end of month. But the usage based charges are very hard to figure out on how many HTTP calls I'm going to get on my system in a month. And based on that I'm going to pay so much for HTTP or API gateway. And then it triggers a lambda behind it. So lambda charges me certain things per call, then it makes a database call and I have to pay certain price for per database call. Right? And then it becomes very hard to imagine when your bill spikes up. You have no control or no insight into why, which of these components are contributing to this, right? So one is visibility, that's what we provide.

Second is we take your provision infrastructure usage patterns. We recognize there are certain over provision resources which could be adjusted. There are certain resources which are getting overused than what you have provisioned for, and maybe there are better resources that could be deployed. Third is based on your usage pattern, there are different classes of services that are available that could be deployed. So going back to my Lego model analogy, right? So there are certain parts which should not be part of your model. There are certain parts that needs to be replaced and certain parts that needs to be replaced so that you get better reliability, better performance, lower cost, right? So we do that identification and what we do is based on literally the founding team's experience of working with thousand customers and saving them more than $200 million over years.

Once we have identified that it comes with certain confidence score, that confidence scores tell us that this has worked with so many customers, so it should work with this customer too. Right? By the way, this customer has different architecture. This customer is in different industry, it is in regulated industry versus b, two, c or whatever is the pattern. And we make that adjustment for our identified optimization opportunity. Once that opportunity is identified, we guarantee that if you are saving cost with this opportunity, it is not going to affect your performance. So we guarantee that you won't see any player or maybe a couple of players who can guarantee that performance protection when they're reducing the cost, which is very hard.

And third thing that we do is even more beautiful is if you want to implement that optimization identified optimization, we offer a single click implementation. So we call it automated remediation. And that saves efforts, skills know how, human error in the equation, right? So we do this end to end visibility, automated identification, automated remediation, and you get that complete value from the equation. Now coming to the containers world. Containers charges are of different kinds. One is how many nodes or how many virtual machines you have configured. If I have to use AWS terminology, how many EC two s are configured for running your container. And that could be a dynamic pattern. How many storage volumes the AWS world EBS volumes are configured, you are paying for that.

Third is you are paying for the data transfer charges, you are paying for the control plane charges where you are managing the cluster. Fourth is within an EC two, within a virtual machine, you are deploying number of containers. Now you could have a virtual machine which has capacity of 100 you have containers which has individual capacity of ten. You have only deployed three or four or five.

Matt Brickey
Right.

Samir Karande
Leaving other 50% capacity unused, essentially wasted. So there are different flavors of where you are spending, what is that you are using? And our help to the customers is identifying, are you using the right Lego pieces? Right. Going back to the same analogy, can you cut down on certain Lego pieces that are not required in the model? Can you have different, better Lego pieces in the model so that you can get better reliability, availability, performance at lower cost? And the last one is, can you get better packing of those container parts in your nodes? And how do you achieve that? Right. And that requires you to look much deeper into the usage pattern on what is the per hour usage for your vm capacity to the pod capacity used. Right.

And then based on that usage pattern, you can recommend them certain strategies which can help them save money. Right. So it's a lot more surgical operation. Right. It sounds very simple, but in order to get a single data point, right, it requires a lot of effort. Right. And getting these 3457 things and then mapping it to what worked with other customers and why it should work with this specific customer requires a lot of domain knowledge.

38:45 - Emerging Technologies and Transforming Cloud Problems

Matt Brickey
Yeah. Unfortunately, there's no dial on these that say go really fast or go really cheap and you can just sit there and tune it with one knob. It's far more complicated than that in the lens of that, with AI being probably the buzziest buzword we've heard in a couple of years. Machine learning, all that, what emerging or cutting edge technologies are you excited about that could kind of transform these problems that we're talking about?

Samir Karande
So I have a slightly different take on this, right? Maybe a controversial take. So you see a lot of hype cycles, right? Let me go back few years and talk about blockchain, right? There was a lot of hype, a lot of investment went to it. A lot of vcs burned their money in that process. And what happened is the real use cases have started emerging in the recent past, right? It has taken multiple years before people could really take certain problem statements, solve it with the blockchain, right? I'll give you example of data Mart, right? If you have rich data and you want to lend it out to certain, let's say you have clinical data, you want to give it out to some research groups, you can rent it out rather than sell or get a perpetual license kind of revenue on that.

Right now with the blockchain, you can see that data is getting transferred, that data is getting verified. It is the right. Verified data that somebody has access to for a certain pre approved time. So you can have lots of checks and balances in the process. Now that is the kind of use case it took multiple years after the blockchain hype cycle kind of died out. Recent AI generative AI kind of hype cycle. I think it will take a few years before the hype cycle dies out and the real use cases emerge, because right now I see a conflicting message in the industry, whether your model is a differentiator, whether your data is a differentiator, and then there is a huge chatter about what data sources should be used.

What is the risk of I using my confidential enterprise data into Gen AI and am I risking exposing my data? Right. So there is a lot of work that is happening, but I think it's very early before the money making or the viable business models on gen AI are going to be here.

Matt Brickey
So more of answer, looking for a question. That's that.

41:30 – Outro: Seeking Solutions for Emerging Technologies

Samir Karande
Yeah. So I see probably hundreds of companies who got funded in last twelve months on Gen AI. Very lofty valuations. I talked to a bunch of pcs in my network and they have their own doubts whether they're going to make money on those deals.

Matt Brickey
So that perspective is always very interesting to our audience, right. Is that you've founded multiple startups, you've raised money, you've hit strides where you've been very successful, you've hit strides where it's like my startup didn't work, right, because I couldn't raise enough money. If you were talking to the audience directly, what key lessons would you say that continue to help you in your approach today without the magic recipe, what could you share with the audience?

Samir Karande
Absolutely. So there are two kinds of businesses, right? One is where you create a new category, which is very hard, right. EWS created cloud, which is very hard. Or I talked about the b two b ecommerce, the first ever b two b e commerce that I was lucky to work on. Now that's a very hard category because you have to educate your buyers, right? And that requires a significant expertise, significant money and elongated timescale. Second is where you know the problem is already understood, the problem is already recognized, and you are going to offer a solution. Now, that solution could be a vitamin versus painkiller. Vitamin selling is easier in good times, but when the economy takes a nosedive, like what we saw in last couple of years, the vitamin decisions would be always delayed.

But if you are selling a painkiller, probably that is not something you can avoid, right? You cannot delay if you need addible, you need edible, right? So that's how the situation is. So if you can identify painkillers, if you can identify existing recognized problem, and then if you have domain expertise as your founding team for that problem statement, either you have lived through that problem, you have experienced that problem, or you have helped number of customers solve that problem, then you understand that problem lot better than most others. And that is where your solution is going to be a lot more suited, lot more effective for that problem statement. So it's hard, but I think startup is never easy. Yeah.

Matt Brickey
Sameer, as we wrap up our session, if you wanted an audience member to remember one thing from our conversation today, what would that one thing be?

Samir Karande
I think there are decisions one can make. And going back to your comment, Matt, about people moving across clouds, people are moving cloud to on prem, vice versa. Right. There are certain decisions which are costly decisions. Do your homework before you make those decisions. And before you make those decisions, think about your existing situations and how that could be optimized so that you don't even have to make the difficult decisions. That difficult decision is no longer required. And I think that's one thing I would like everybody to remember. If you want to optimize your aws, we have a solution and would be happy to help you out.

Matt Brickey
Great, Samir, thank you for your time. I'd also like to thank our audience. And that concludes our episode of Cloud currents.

Samir Karande
Thank you very much, Matt.