 Hi everybody, welcome back to Caesar's Forum. You're watching theCUBE's day one coverage of Snowflake Summit 2023. We're going to get right into it because we're tight on time. Tony Baer is here, he's the founder and CEO of DB Insight. He actually has more than one insight, but he's, and Doug Hetchin is the vice president and principal analyst at Constellation Research in Sanjiv Mohan, principal at Sanjimoh. Guys, great to see you, we've done this before. Let's get right into it. Tony, what did you learn yesterday that you can talk about that's not NDA? What were your big takeaways? Well, I think it was pretty, you know, when the Christian, you know, finished the keynote this morning, it's container services that kind of blows the lid off of Snowflake processing. It basically addresses all the limitations that data scientists, you know, were complaining about, which is that, you know, they don't want to work through UDFs, they want to execute directly on the data. And with, you know, container services, you can do exactly that. Anything you'd add, your big takeaway, Doug? Yeah, I mean, absolutely snow park container services is the linchpin of almost all the announcements here, the ability to run ML, the ability to run the native apps, APIs, et cetera. Interesting that, yeah, one of the big organic applauses was on developer capabilities, a new command line interface, logging and tracking, auto sync to get repository. Right, it was like, oh wow, how boring, but people were excited. That's, you know, it's the simple stuff that appeals to developers. If you're going to be a platform for apps, you have to appeal to the app developers to make their life easy. And you would agree container services were, I mean, I think they're even containerizing the GPUs with the video. Yeah, exactly. So I thought container services would be my number one choice of the new announcement. And interestingly, what it does is, LLMs, which are, of course, the big thing these days. So now there are four different ways of accessing LLMs. You can containerize it, an open source, you can, you streamlet, you have native LLMs, document AI, and you can call LLMs through an API, like open AI. So this is how the platform is expanding. So the high level messaging, all data, all workloads. Okay, I get that. What doesn't, I think it's there, but I want your guys opinion. They got a unified data platform, they got a lot of different ways to query now, and they can store and manage different data types. So there's like this magic integration. Am I getting that right? Is that coming across in the messaging for these guys that I can query different data types and have it return kind of consistent, or is that, like, pipeline? I'll put it this way. I think we also have to put this in the context also of the Lake House. And while it doesn't get all the decibels in the spotlight that LLMs do, the fact is that you want to take the lid off of whatever data, and part of it is all the unstructured data, but also all the data that was in the data lake, with the Lake House, basically you get trusted data, and Snowflake is basically is essentially making Iceberg a first class citizen. That's, you know, it's pretty profound. It's all data and it's all workloads. I think that's the big take with the containers, all workloads. It's all data and all code. Yes, yes. That's what it is. In fact, it's a brilliant strategy because what Snowflake has come to a conclusion is we own the data, we should now own all the apps. And they're an underlying theme of this entire conference is security and governance. If you run things within our platform, no data movement, secure, govern, shared, and scalable. So you guys are on your way to San Francisco to the Databricks event. So you got Databricks is really, I see it. Tell me if you guys see it differently. There's a data engineering, data pipeline company, Snowflake's a data management company. They're each trying to go in and out of the direction. Is that how you see it? No question about that. I mean, the thing is you got to look where their natural constituencies were when they, where they got started. Databricks was always the company of data engineers and data scientists. And, you know, Snowflake originally came as the data warehouse. I mean, even admitted that this morning, even said those words, obviously. So basically I think through container services, they're really trying to make a very major play for the constituency that felt that going through UDFs and store procedures was just not going to be adequate. I thought data warehouse was like multi-clouded Amazon at this conference. Yeah, despite their efforts to encroach on each other's turf, their strengths are still in their base, but they're steadily pushing outside. I think Databricks' move to acquire Mosaic ML was a good move to shore up the data science and make sure they're not, they're still leading on AI and this generative capabilities. They spent the last few years really talking mainly about building out the house capabilities of their lake house, but they got to protect their flank in their data science supremacy. Was that move by Databricks a Neva FOMO, in your opinion? No, I don't know if they knew about the Neva FOMO announcement at that point. I think they're just pursuing what works for them. I mean, Mosaic is there in San Francisco. I think it's an aqua hire in many ways. This talent is scarce and they're bringing them into the fold. I find container services is a game changer. That moves them away from data warehouse. Like for example, one example that showed us was Pinecone. Pinecone is a full flash vector database that runs on its own, but now you can run the entire Pinecone as a container inside a snowflake. So I don't see them as a data warehouse company anymore. I think they've really blown apart. Well, they started a data warehouse company years ago. So you got Vector with Pinecone. Now you have a graph database with relational AI even as kind of a hybrid. So you have all these different data types, pluggable storage essentially. And how unique is that? I think that is quite unique. In fact, I would say snowflake is going down exactly the same path AWS has gone down, which is there are many different ways of doing it. We will give you native capabilities. So with document AI, they now have a native vector database themselves. Yeah, but you can't integrate them on AWS. That's the difference. That's the magic secret sauce of snowflake, right? Correct. Huge, that's powerful. George was just saying Werner Vogels was saying, wait, it's your fault. The customers two years ago at Reinvent, you guys asked for all this stuff, but they just threw out primitives. Snowflake's done the work, it sounds like, to do the integration. Is that true though? Have they really done that? I mean, we're hearing a lot of pre, we're hearing Iceberg, we're hearing, I didn't hear much about Unistore. Yeah, we got to remember that container services is private preview at this point. Last year we heard about Unistore. I'm surprised they mentioned it. It didn't come up very much yesterday, but we got to make sure that we don't think of, these container services, the latest shiny thing that makes it all available. They have simplified. For example, the container services underneath the surface is Kubernetes, but it's not exposed. They've taken all the complexity away. That's what they're saying is basically, hide the code and both of them, for simplicity's sake, but also to protect IP of all the creators. And then it's also something that they sell. They're selling the compute, they're selling the orchestration, they're selling the storage. It's something that is Snowflake product. Which is very different to AWS. Same, if I understand it. And GPU. Same with GPU. It's like Amazon's going to get paid. We'll give you this option. But I'm not going to spin up a GPU EC2 instance from NVIDIA, from NVIDIA, I'm going to buy Snowflake. So Snowflake container service is like EKS. But the choices of compute and storage are things that Snowflake will provide. It's not like on a cloud where you have lots of choices. I want to ask you guys the supply chain example. Use Blue Yonder, and where that fits in container services, because Blue Yonder is Duncan Angove's latest heavy lift. You guys know those guys. Is it, that's the manugistics, at least in part the manugistics. They got all these legacy apps. So you're going to containerize them, put them in and modernize them. Is that a viable play? Well, put it this way. I mean, look, it's also the same story you hear like with anyone, actually it's surprising coming from Snowflake, because that's the type of message you tend to hear more like from the IBMs and the oracles and the world and the terror data. So it's just that we have all these landed investments. And what Snowflake is saying is, we'll take those land investments, and if you can put it in a container, you know, haven't run in our environment, you know, haven't run under our security and all the wonderful stuff against our data. You know, I mean, Snowflake basically at the end of the day, you know, basically cries all the way to the bank. But I could do that in the cloud, but to me the difference is that you're actually, they're actually re-architecting around Snowflake and relational AI to actually truly try to solve this supply chain problem. But I don't know enough to know what to do with it. It's Blue Yonder's app. It's going to be a Blue Wanderer native app. And there's, you know, Snowflake is saying, we're providing all this infrastructure. We're inviting in the native apps. And that's another important topic here today, the native apps they now introduced last year. Now they have 25 companies that have built 40 native apps that are in production. You know, that's good, that's impressive. We were told that they're mainly analytical because they don't have the Unistore yet. Little quiet whispers behind the scenes, mentions that, you know, container services could run Spark or could run SQL, you know, O-L-T-P. They don't want to run Spark because they're saying Snowpark is two to four times faster than Spark. Is Spark, but you know, what they're saying is container services is very open to run many things. Could be even relational database. Yeah. They've eliminated the Delta in performance between sort of Iceberg Open and Iceberg on Snowflake. I remember last year I asked Benoit, but yeah, isn't there a Delta? And he looked at me funny because he's so far ahead, probably into the roadmap, like, no, no, you don't get it. You know, he didn't say that, but he kind of looked at me that way. I think they were saying 10% last year, a 10% penalty. Right, there was a penalty. And now there's ostensibly no penalty so you can actually update right in place on the Iceberg table. I'm not sure if this came out in the keynote. It's actually bidirectional now. So you can create a managed Iceberg table, but you can also go to Iceberg and access a Snowflake table, right? So it's bidirectional. Yes, exactly, and you can both read and write and so far in terms of the Open Table Form it's there are very few vendors that basically do it both ways. So they're stretching the Snowflakes already use these terms with fabric or mesh to different data types. Yeah, that seems like a very viable and bold strategy, but should we be concerned about these announcements and pre-private preview and then a year later we're still not in GA or is that- I like the native apps. I like the native apps because basically that's gone to public preview. So I mean there's some real progress there. Some of the other stuff like, I mean Unistir, I think that's going to be a long and winding road. At the analyst summit and today they were very good about following up on last year's announcements. At some events you don't see that. We were at MongoDB last week and they interested as much of analytical capabilities last year and we didn't see them talked about very much. They were very good about following up. They brought up Unistir. They brought up the native apps. We are seeing progress. There are things that are announcing here that customers want in a big way. Lots of cost controls and optimizations. And if I were a customer I'd be asking them to put on the coals to bring these to the fore. And in the FinOps environment we're in today with big interest in cloud optimization. They should have more of these capabilities. I'm surprised there's one of their customers, Capital One, has this Slingshot app. That's something Snowflake should have. Well their timing of that was pretty good. They announced it last year and now everybody's so focused on optimization. But it's really interesting about Slingshot. Whatever Slingshot is doing in the app Snowflake is doing it in the database. You can create a budget. You can create thresholds. You can monitor. You can alert. So the question that begs is like what is the reason of having a third party product if it's natively available? Well they have a lot coming but it's in private preview. The Snowflake capabilities are in private for the most part. But Slingshot had a massive boot last year and then didn't really hear very much about it. Well the other thing, the other place that battle may eventually be fought would be in ML Ops. And Snowflake has to do a very careful sort of, I guess just very fancy footwork with their partners like Datakoo and SAS and all those wonderful, you know, DataRobot and all those wonderful stuff in terms of that they all do ML Ops. So like where does one pick up and the other leave off? What about ETL? What happens? I mean they're kind of going to dance around that because you've got ETL vendors in the ecosystem and they're basically saying yeah, you don't have to move the data anymore. So yeah. You have amusing native apps frameworks. For example, Matillion has a bunch of connectors and things that they're building in that framework. So I think they're reaching out to partners there. They also introduce their advancement on streaming, Snowpark streaming, and Delta Tables. So these are things, their own building blocks for faster integration of data. Dynamic tables, yes. Delta Tables is tomorrow. Sorry, exactly, yeah. Dynamic tables. I took a, yeah. I was going to say I missed that. Oh my God, Datakoo, wow. We're not going there. Okay, so coming in to the show versus you guys are on a plane this afternoon going to San Francisco. Has your impression changed at all? Notwithstanding, a lot of the stuff's not shipping. Yeah. Has your impression changed? Positive, more positive, more negative, neutral? I am positive. I think Snowflake is really thinking wide and beyond just the data. There's thinking data applications and everything around it. So I'm very gunkal. Yeah, very impressed. You know, I think the container services are very promising. Hopefully we'll see the follow up next year where it's getting to public or is in public and GA is palpable. You know, Unistore and some other things that they developed last year, still waiting. So I'd like to see, make sure these progress from stage to stage is happening quickly. Yeah, at the risk of sound like a broken record. It's, again, it's container services, container services, container services. And the reason for that is it really addressed a key, I guess, sort of shortcoming in the whole Snowpark environment, which is that it really, basically, it was Snowpark last year, up until now, it's like you had to go through UDFs or stored procedures. And that clearly does not appeal to the crowd that they want to reach out to. And I heard that on the floor when I was here last year. I think container services just hits that right on the nose. Yeah, the North Star is they want to be the app store for enterprise data apps. The iPhone analogy. You've got to simplify that. Yeah, and Snowpark is great. Sorry, container services for all the reasons you said, but it's more than, it's like even machine learning to train a model. It was really hard to do it in Snowflake last year. Now they're model registry. So they're introducing the whole MLOps capability natively. And it's also a potentially strong partner play. Yeah, yeah, absolutely. All right, Tata Gang, we got to go. Love to have you guys back. Awesome. Thank you. Good to see you again. Have a good trip. Safe travels. Thank you. Definitely look for John Furrier out there. Yeah, it's going to be in San Francisco. He is at the Intercontinental and also inside of Moscone. So double, double pop-ups. So check it out. Okay. All right. And you too. Check it out on cube.net. We'll be right back right after this short break from Snowflake Summit from Caesar Forum in Las Vegas.