Shop Talk: Snowflake Summit Recap

Eric Dodds 00:05
Welcome to The Data Stack Show shop DOT Costas, we have talked with people who built amazing data technology at companies like Netflix, Uber, and LinkedIn. But you and I actually don’t record our talks about data very much. But we actually talk about data together a ton. And so Brooks had this amazing idea of just recording some of the conversations that you and I have, before and after the show about data and our opinions on it. And really, this has been my favorite thing that we do. So welcome to shop talk. It is where Costas and I share opinions and thoughts on a personal level about what we’re seeing in the data space. And it really is simple, we ask one another a question, and the other one tries to answer it. So without further ado, here is shop duck. Welcome back to The Data Stack Show, we are going to talk shop. Costas. This is one of my favorite things, we just sort of get to go off script and talk about whatever. And I think we talked about AI last time. And I may have brought that subject up. So I think it’s your turn to ask the question.

Kostas Pardalis 01:19
Yeah. And so I think it’s like a great time, because I think you just came back from Iglu Snowflake Summit. And I wasn’t there. So I’m really interested, and would like to learn more about what happened there. Both in terms of like, what Snowflake, let’s say, had to say about Snowflake and the industry in general. But also, I’d love to hear what you’ve learned and like how what you experience there might have changed or not, like your perspective on where the industry is going?

Eric Dodds 01:59
Yeah, I think I’ll try to give a high like a high level summary of the sense that I got, you know, as someone who works in product marketing, I think I can tend to be cynical, maybe I don’t know, maybe I’m just a cynical person, in general, when it comes to marketing, but you know, for a long time, I mean, relatively long time. Snowflake has been kind of pushing this concept of a data cloud, and didn’t What was funny about that, in practice was that everyone just still thought of it as a data warehouse, and even called a data warehouse, right practically on the ground, just I mean, stuff, like data warehouse is just the vernacular that people would use data practitioners would use when they talk about it, right? Because that was for the first you know, however, you know, for the first, you know, since it launched and everything, it was used as a data warehouse, and still is largely used as a data warehouse, right? I mean, analytical workflows are, that’s, you know, sort of the most common thing you run into. And so there was a little bit for a while there, there was a little bit of dissonance between talking about the data cloud, and practically, people just thinking about Snowflake and using Snowflake as a data warehouse. And I really felt like this Snowflake summit was where you could feel it shift to actually there being a lot more weight to the concept of a data cloud. Now, obviously, there are really smart people and visionary people sort of establishing that as a concept. And, you know, product roadmaps are driving towards that. So I’m not saying that, you know, it was fake beforehand. I’m just saying that there was, I think, more dissonance for the average person who probably categorized Snowflake as a data warehouse practically in their day to day job. But there were just a lot of things where it really made it feel like a platform with lots of options, right. So the NVIDIA announcement, obviously, was huge, right? So that’s going to be pretty significant for the development of really large scale, really large scale AI models, which feels very different from the way that people would traditionally sort of categorize Snowflake as part of their data infrastructure, which is really interesting. The other big thing, I think, is container services. And so, you know, several companies announced, you know, actual sort of native integration with container services. So you can essentially run sort of these products within container services within the Snowflake data club. Which is really interesting. Especially when you think about SAS apps that have data in them, but then you can actually sort of operationalize that within operationalize that data within. Within containers, they get very interesting, right? And so now you all of a sudden you have all of these ideas, I think rushing to people’s heads around things that you can do and build that maybe seemed more theoretical, more theoretical before. So I don’t know, I don’t know if I’m sure I’m missing some things. But I really left the summit this year with a strong conviction. There’s a lot of infrastructure, and there are other things that they released that were, you know, really neat. With a strong conviction of like, man, we’re gonna see an explosion of people building really interesting things on Snowflake, far beyond the bounds of typical analytical workflows. So I don’t know, there’s my high level summary.

Kostas Pardalis 05:59
Yeah, let’s, let’s keep it interesting. And like, I think, like, kind of makes sense. That’s, yeah, you know, like, it’s the data cloud. And it’s the cloud at the end, right? Like, it’s my infrastructure to go and like, build in general. It’s not a dashboard. Yeah. Man. It’s very interesting. Like to see the path and the journey that Snowflake has, because yeah, like extending, like a cloud provider in a way, right? Yes. Which probably is like, okay, like the only way to justify also the multiples? The market, right? Sure. But, okay, so you mentioned, like, Nvidia. What’s about Nvidia? Like, what is the announcement there? Like? What did they describe? And what is like, how’s it like the Vizio and what they have working like, and providing access to Nvidia hardware?

Eric Dodds 07:03
Yeah, well, actually, so I wasn’t at the keynote. Um, I wasn’t at the keynote. So that’s a full disclosure. But I did. I did talk to people who were at the keynote, which actually is almost more interesting, like, I don’t know, in some ways, maybe this is more interesting to some people, maybe not. But like, I talked to several people who were at the keynote and asked them what really stuck out to them. And this may sound funny, but I think one of the biggest things is confidence that there’s enough horsepower there to actually do really large scale, machine learning workflows, and sort of develop, like really large scale. You know, sort of, let’s just say like, enterprise level, ml production workflows, right? Because like I said, Before, people just didn’t normally think about that, right. And so what, like, the people that I talked to, who came from the keynote, who were really excited, you know, who worked for, you know, some of these people work for very large, you know, sort of maybe like, fortune 1000 ish type companies. Right. And they didn’t really talk about Nvidia, specifically, right, or like, the technical undercarriage of like, what the partnership means. They were just like, wow, like, maybe we can build some really big stuff on a snowflake’s platform now. Right, which was really interesting. Again, it kind of goes back to what I was saying earlier is that they sort of it was almost like a confidence thing of the horsepower actually existing. I don’t know if that’s my takeaway. And I’m probably just showing some of that wrong because I wasn’t

Kostas Pardalis 09:06
like, I could be honest, and that was, like, a follow up question that I had for you was about like, the interactions that were you had like with the people there because you were not you know, as a vendor, you’re also going there like as a vendor, you also have like the opportunity to have like a very, like almost like, interesting in between a position right? Like you are not a potential customer like Snowflake there. And you’re not Snowflake, also. So you are like, as a vendor always you have a very unique kind of perspective and way of interacting with the visual world visiting. So what was the I mean, you already said some stuff about the confidence that You said that they had on the ML side of things in general, like, what’s your? What’s your take from what you had, like from people visiting there? While they were asking what they were looking for, how they felt? What is it like there? What, like the vibes that you got from them, as you know, like practitioners, right? They are not vendors, they are not Snowflakes. Yeah,

Eric Dodds 10:25
this probably isn’t going to surprise you. But I would, if I had to simplify it as much as possible, I would create two general groups of people. The first one is actually the bigger one, which we’ve talked about this before on the show a lot, which is people who are just trying to build a high quality data practice within a company, and who are trying to solve the basic challenges that you have, when you’re trying to do that. And that is, I need to get a lot of different disparate data sources into one place. Obviously, the people at Snowflake are, you know, doing that in a snowflake’s environment. And then I need to try to create some sort of value with that collected data. And in many ways, that kind of characterizes a lot of the traditional thinking about Snowflake as a data warehouse, right? It’s a data store that allows you to easily get all of your data into Snowflake. And then the separate separation of storage and compute allows you to, you know, make smarter decisions about how you actually try to begin creating value out of that, you know, for different use cases. And just sit, you know, when you talk to people who are coming by the booth and just ask them, How are you using Snowflake, it’s just easy for us to forget that like, a lot of companies, especially larger companies, it’s just really hard to get over the initial hump of doing the basic stuff, right, like collecting data. And even driving really good analytics is still a very difficult problem for a lot of companies. So that’s sort of the first group now, I will say one thing that was interesting was the ecosystem of tools provided by Snowflake to do that was talked about way more. So like the snow pipe streaming, infrastructure, and other things like that, where it’s like, you know, you’re seeing Snowflake actually now have the ability to replace what traditionally would be sort of a, you know, complicated set of Kafka pipelines, and maybe like homegrown API’s and stuff. So that was kind of interesting. So I think that some people certainly felt like they had more options from Snowflake that were really viable for sort of replacing some of those traditional data flows. And it was that sort of group one and I, again, I would say that’s a larger group, right? Because as much as we’d like to tell ourselves that every day to practice is super modern and sophisticated. A lot of them are still trying to do basic stuff. But again, that’s getting easier. The second group were, I would say, this is a really interesting characteristic about people in the second group, they were thinking about all the new capabilities of Snowflake. And there was a lot of discussion around consolidating workflows, right? That’s a huge problem. And especially with the traditional split between analytics, workflows, and ML ops, those I guess, maybe a good way to say it would be like, there are people who, in their daily job, are starting to see those things converge. from a cultural standpoint at the company. I think a lot of that’s accelerated by AI, right, and prioritizing machine learning, right? And so you’re starting to see things like analytics and ML nailed. And a lot of people on the ground there are there to figure out how to get more value out of their Snowflake investment, right? Like, how can we use this platform to create more value inside of our company. And so it’s really interesting to see them. You know, they may not have used this exact phrase, but if I had to distill it, and put words in all these people’s mouths, which is always very dangerous, but you see their gears turning around consolidation of workflows, which is pretty compelling, actually. Right. So if you think about, let’s say, there’s someone who’s ahead of data, and they have a really mature like analytics practice, and then a more immature and malpractice, but they can actually leverage a lot of the analytics work that’s already done. Like a running start for ML and the infrastructure is already there essentially, like, you don’t really have to do a big infrastructure project to migrate data, move that data, you know, run complex transformations on that data. It’s actually just there and you can start doing ml. That is very exciting to people. And I think it should be because, I mean, that’s pretty sweet. If you’re someone in that position.

Kostas Pardalis 15:25
Yeah. 100%. And, and I think, like, okay, like a big problem, that’s data infrastructure, because right now, he’s like fragmentation, then there’s like, a lot of replication also, that is happening, like at the end, there are, let’s say, common patterns that exist, that regardless of what you’re doing, like if it is a male, or like, via, like reporting, or whatever. And I don’t think we have reached the point where, you know, there is like a robot mice in the architectures like to provide, let’s say, the best possible experience at the end, because people might think that it’s more about cost, because you don’t want like to, you know, like double gate things. But at the end, it’s not like, it’s not like the cost in terms of money. Like, what people don’t understand is that like these things, even if they were like, let’s say, for free, they just don’t scale to the size of the problems that they’re trying to solve. And actually, like sad, slow, brutal infrastructure. It makes Joma. Like, how down like the whole process, that’s why we ended up in getting this kind of fragmentation, like in most people, at some point, they just had like to move much faster than the rest of the infrastructure there because things were happening. And they couldn’t wait like for the rest of the infrastructure like to change, right? Yeah. So that’s why we had like, all these things. But at some point, if you want to operationalize all these things, you need to have like a common infrastructure like to work on top. And that’s where like they this whole concept of the data cloud, or whatever you want to call it, like, makes sense, right? Now, who’s going like to own these? And if it’s going to be one or like, multiple come? I don’t know. But that’s, I think, like, where we are heading towards, I think it’s going to be like really fascinating. I’d love to also get were like, close to the end here. We talked about Snowflake, there was another sound mixer that was happening at the same time, right? Yep. I’d love to see it, but I wasn’t there. But I’d love to find someone who was there to shop, talk and like, also give a quick update of what happened there. So let’s try to figure this out and like, make it happen.

Eric Dodds 17:41
Let’s do it. And I would say one, one other thing, just, you know, I know that there are probably a lot of data vendors who listen to this, but it’s always a really good reminder that there are so many vendors for doing very similar things. And it’s hard for people to sort through all of the options, they have to do very similar functions, right. And so, and that’s actually getting worse, because of snowflakes’ advantage of building Snowflake, native apps, right, your options are actually proliferating even within the Snowflake environment itself. Yeah. Which is a great thing for Snowflake, but is creating an interesting complication for people out there who are trying to decide, you know, which sort of tool sets to put together. So I think it’ll be interesting to see how vendors sort of respond to that from, you know, a communications standpoint, content standpoint, all that. All right, well, we are at the buzzer. And that was my brief overview. I’m sure someone will email and tell me about all the things that I missed. But we’ll get someone from the data AI conference on the show soon. And we’ll catch you on the next one. You know, we learned so much from the data leaders that we talked to, but I learned so much from picking your brain. And actually your questions really make me think really hard. So I appreciate shop talk. I think it makes me think sharper.

Kostas Pardalis 19:07
Well, it’s, it’s fun. What? I think it’s good to exhaust Seaton Chartio Boggs, the stuff that we experience. And yeah, I think like, I felt like people enjoy it. That’s why I’ll keep asking for people to reach out, please do this. Gah, pork, like, you can’t do that. Like, send an email. Yeah, let us know how you feel and like, what are your opinions and your experience with the show? So please do that to me, and then we can get a copy.

Eric Dodds 19:43
Of course. And of course, we try to take the same types of questions to, you know, data leaders from all sorts of companies, large and small, so definitely subscribe to the main show, if you haven’t yet. tons of really good episodes there. And a ton That’s a really good thoughts from data leaders really around the world so definitely subscribe if you haven’t and we’ll catch you on the next shop talk