The Power of Abstraction: The Case for Super Clouds with Sanjeev Mohan

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

Hey, welcome to another episode of the CTO Advisor Podcast. I am extremely fortunate in the quality of guests that I always get on the podcast. And today is no exception. We have Sanjiv Mohan, who is the founder of Sanjmo Advisory Firm. He is a former VP of Research at Gartner, one of the sharpest people I know in the data analytics space. And we're going to save kind of his specialty for the last part of the conversation. Sanjiv, welcome to the podcast.

Keith, it is such an honor to be on your podcast. I myself feel honored to have gotten to know people like you in the industry. I followed you for many years. So it's always humbling to hear someone so high up from Gartner has followed me. I'm like, I'm just this lowly independent guy. But now you're independent and you're doing some amazing work. It seems like I'm seeing you everywhere. I saw you not just in my session, in the super cloud session yesterday, but you're all over the place.

Congratulations on the exposure. And we're going to start the conversation at that super cloud. Is this, I know we had this conversation yesterday on theCUBE, but is super cloud a real thing? Let's get past the term super cloud. Is super cloud itself a real thing? I think it's a natural progression. It's a natural progression from being single cloud to multi-cloud and now having that kind of abstraction where my workloads can be running across different cloud providers. And I don't really care where they're running.

I just need an ease of use and span and pick the cloud where I can run my workload in the most cost-effective, reliable and secure manner. So I think it will happen. So this is something that we've talked about. It seems like forever. And it's interesting that people are getting pushback. Do you think it's because of the corningness of the term super cloud or are people resistant to the idea that at some point we'll abstract the clouds? So I think it's a combination of a number of things.

The term itself can be a bit problematic and it's not just the term super cloud. We've lived with these jargons and terms forever in IT. I mean, you can even say big data. What exactly is big data? What's a data lake? What's NoSQL? The list goes on and on when it comes to terms. So terms are always problematic. That's the issue number one. But the thing is that it doesn't matter what the term is. What matters is what does the business want?

And if the business says that I have a certain workload because of data residency requirements, I need to run it in a certain region and the only cloud provider there is X, but the rest of it can run in a different region, then it is going to happen irrespective of what we call it. A problem though is that like in anything that's new, it takes time to mature and we don't have any standards. And that I think is the biggest problem. Right now, if you ask AWS, what do you think about super cloud or Google cloud or Azure?

Well, what are they going to say? They don't want to share the limelight with other cloud providers, right? So they want all workloads should be, that's why we have egress charges. So you stay in a single cloud. So lack of standards is a huge impediment for super cloud to become mainstream, in my opinion. Yeah, so I think this lack of standards is one of the biggest factors. Whenever I think about the problem around super cloud is this ability to abstract away the clouds, but the clouds move so fast.

And I think some of that is starting to slow down, but the clouds move so fast. And when we think about abstracting serverless, you have Lambda, you have cloud functions, you have Knative, you have OpenFaaS, you have all these different ways to manage the control plane of serverless, but you don't have a centralized way to kind of have one way or one abstraction for serverless. So I think that's the opportunity for super cloud is to say I can do functions as a service in a singular way.

There's one way to do it all. And then the control plane or the super cloud determines which cloud to run that function on. And that's all handled. That's a super difficult problem to solve, but I think that's like the potential and I think the natural progression of cloud. Let's go on to the next topic that we wanna cover, which is the economic situation. I'm refraining from calling it a economic downturn, but it is almost impossible to ignore the mainstream media covering all of the layoffs in enterprise tech and tech in general.

How should customers be in these companies be thinking about the potential of an economic downturn? So economic downturn is real. We see people losing jobs every day. There's some news. I think there are ways to protect oneself. I talk to a lot of my clients. They're very small technology companies. By the way, first of all, Keith, I don't know if you agree with me on this, but this slowdown is primarily a technology sector. I don't see this in letter manufacturing or healthcare or banking, for instance.

Do you see it? No, it's interesting. There's this big talk of slowdown in our industry and every industry that we serve seems to be humbling along just fine. Correct, yeah. So my point is that the producers of technology or the suppliers of technology are suffering, but the buyers are still pretty robust. Now, if the buyers went down, then we would have this global recession or a massive problem. So this is a time I tell my clients, you need to very carefully think where you are spending the money.

For example, a lot of companies I know have point blank five people frozen their marketing budgets. And to me, that's a kiss of death because right now, everybody's concerned about burn rate. How do I reduce my burn rate? But the problem is if you stop innovating, if you stop getting your name out, then the effects will be felt maybe two or three quarters from now. So I still think brand awareness is really important. It's really important to be in the news, cost-effective ways of getting your name out.

But don't retract, don't add. Like for example, sometimes I even tell companies, they're like, oh, we are product-led growth. So we are taking this time to add more functionality to our product. I'm like, well, but now you're spending more money and you're telling me burn rate is a big issue. You stopped everything else, your sales has stopped, your marketing has stopped, but you're adding more functionality. Why are you adding more functionality? Your focus should be on what do I need to do to retain my existing customers and add functionality only that adds to your depth.

Don't start looking into other categories because if you start doing that, then your product market fit is going to shift and now you'll have a new problem. So, yeah, and we've seen this play out itself time and time again in the industry, whether we're talking about marketing, engineering, et cetera. When you cut too deep to the bone, you put yourself at a disadvantage. There may be great short-term benefit, but long-term, can you meet the needs of the market, let alone the needs of your customers?

The case in point, I saw in the news last week that HPE was being sued by shareholders because when they spun out HPE Managed Services to form DXC alongside of CSC, the new company cut too far deep in the technical expertise. 7 billion, which is an amazing short-term savings. Yes. But long-term, they couldn't meet the demands of the users. Case in point, I wasn't obviously any longer with that company in 2019, but that company went on to another service provider. I'm told because DXC was not meeting the needs of the customer.

They lost that deep technical expertise they were known for when they were HPE Managed Services. So yes, I think this is a good warning point for both enterprise customers and those who create technology that yes, we should be prudent with our resources. But as you mentioned in our pre-banter, measure once. Yes. Cut once. Don't cut before you measure. Yes, and maybe in these times, you measure twice before you cut and you measure frequently. Don't wait for six months or end of the year to say, okay, how are we doing?

So constantly measure, see what programs are doing well and then readjust your budget and tell everybody, you know what, the budget is not zero. The budget is half. And if you do that and you really focus on your existing customers, retaining them and making them happy, I'm guaranteeing these customers will find the money. Maybe you were expecting a million dollar deal. It'll be only half a million, but at least it's not zero. You know, as long as you are saving the customer value, saving customer money or showing value, like for example, in my space, you know, companies are putting in data ops products so they can deliver faster, you know?

So whatever it's agility or cost or whatever it is, as long as you can do that, you know, the customers will find you the budget because large enterprises, non-tech enterprises are still growing. So let's go into the last topic, bouncing off of that, your area of expertise, data and data analytics. We're coming out of, both of us were at AWS reInvent. Yeah. And we're coming out of actually some pretty, I want to say they're low key announcements, but pretty substantial announcements out of AWS when it comes to data analytics and overall capability.

What's your impression of the announcements in the show? So the announcements are low key because we have reached a level of maturity in AWS's journey. For the second year in a row, we were surprised that AWS did not come up with a brand new database, for example. And that's a good thing. We don't want yet another database, already 15 of them or some large number from AWS. What AWS announced at last reInvent is bringing different component services together. So people don't have to manually connect them or write code to do it.

For example, zero ETL was a very big announcement. So you've got an operation database called Aurora, you've got an analytical database called Redshift, and don't worry, Mr. Customer, AWS is going to take care of it and sort of make sure that in a very performant way, you can connect these two or you can go to Redshift, look at Aurora data through virtualization, all that underpinnings of integration, they're taking care of. So that to me was a refreshing change in AWS's attitude. Yeah, when I was in enterprise software sales, I would talk to a CDO because I was in the infrastructure space.

And the CDO of a large Midwestern bank said that his data analytic, his data scientists spend 70 to 75 to 80% of their time on ETL and not the actual work of a data scientist. And for me, it seemed obvious that this is a infrastructure problem. If you're spending 70 to 75% of your time on something that's essentially infrastructure, you should invest in infrastructure. So I don't know if I was, I obviously wasn't the only one excited about this ETL announcement at AWS.

You obviously caught that, this being your area. And the engineering effort needed to go into reducing the minutiae of ETL for data scientists will effectively, in some cases, double the output of your data scientists. Correct. Yes, I mean, that productivity is critical. And by the way, there's yet another big announcement for me. I also cover the space of data governance a lot. Also, data governance is a very loosely defined term. It's just like super cloud, actually, to some extent.

A lot of people don't like using the term data governance because they feel that it's so many times it's a taboo to talk about data governance. So let's talk about metadata. We have all these things happening, whether it's data or even infrastructure. So Keith, you cover infrastructure topics a lot. So when you build these pipelines of moving data from, let's say, a SaaS application, use Fivetran, you load it into a database, you use DBT, then you use Looker, so you use a combination of these tools.

Every one of them produces metadata. That metadata is broken. It's siloed. It sits in everybody's different tools. We've tried to fix it with data catalogs, and I think data catalogs have come a long way. So AWS, which up to this point had stayed away from this entire messy metadata area. They had something called Glue, but Glue was very technical and very limited in its scope. So they announced something called DataZone. So DataZone is this new business metadata catalog that was other big announcement.

Now, it's still early days. I have not yet seen it in action. I only saw it reinvent in the demos, but it's a good step from AWS. Yeah, so ironically, I put this, I have this AWS every day, and I cover a different AWS product every day, and DataZone, we covered last week, it's in preview, and it is a really interesting solution. If you look at like, I call it the Databricks for AWS. It is their alternative to it. So if you never have to leave AWS, if you have a SaaS provider that's in AWS, if you have all these different data sources, and you want to get together your metadata so you can do analysis across these separate types of data, DataZone seems like the promise, right?

Yeah, in fact, AWS is, I would say, a little late to the party. If you look at Google Cloud, their Google Data Catalog, they announced it a few years ago, and in fact, they don't even sell it anymore. It's embedded. Yeah, it's not even, it's a value add of using the platform. And then Microsoft Azure has had its own journey. They had something called Azure Data Catalog way back, but then they rewrote the whole thing, and they came out with Purview, Microsoft Purview.

I've heard mixed messages about Purview, but it's a very difficult space to be in. So I'm happy that AWS finally has a product, and it'll be great to see how this whole space matures. Sanjeev, it's been great having you on. Wealth of knowledge across several different domains. Where can folks find your musings, writing, research, what you're doing these days? Thank you for those kind words, Keith. I have my website. com, but I have not kept my website up to date.

It's been many, many months. The best way to follow me is on LinkedIn, so I'm quite active on LinkedIn. I do have a medium blog, so I've already written two articles this year. So Keith, you might be interested because you are in the infrastructure space. I wrote my key takeaways from Oracle's Cloud World a few months late, but better late than never. com. And the last thing that I want to say is that I've been doing a podcast a lot.

Last year, I launched my podcast series on YouTube. I don't have the audio-only version. It's called It Depends. So the link is actually, I can send you the link if you want to put it in the show notes. I'd love people to take a look at it. It's mostly data and analytics. It's called It Depends on YouTube. So those are a few places. So we'll put all of that in the show notes. I love that you have so many platforms to engage with, Ian.

And I will co-sign. You really need to follow him on LinkedIn because he is a prolific LinkedIn poster and engager. He's always engaging in interesting content. com. You can follow me on Twitter at CTO Advisor and check out the new AWS every day. I think we're up to 30 different AWS products. We had Joe Peterson the other day review some security products. And I might even ask Sanjeev to do some of the data, the more data analytics centric stuff.

So we'll continue to have guest appearances. Again, so share the podcast with your friends, family and coworkers. Talk to you again. Thanks a lot, Sanjeev. Thank you. I'm just starting my day and you've already made my day, Keith. Thanks. Yeah.