Tim McConnaughy: 0:01

Welcome to the Cables to Clouds podcast, your one-stop shop for all things hybrid and multi-cloud networking. Now here are your hosts, tim, chris and Alex. Hello and welcome to another episode of the Cables to Clouds podcast. As always, I am your co-host, tim McConaughey, at Carpe DMVPNPN on Blue Sky. I don't use that other site anymore, really.

Chris Miles: 0:29

Got rid of it.

Tim McConnaughy: 0:29

Yeah, if you go to my profile now it says like 404, not found or not. Wow, I didn't delete it. It's still there, but I don't post to it anymore. So yeah, big news, big news, yeah, big news, yeah. So follow me on the other site. It's carpe-dmvpn and of course, with me, as you obviously can tell, is my co-host, chris Miles. I forgot what are you on Blue Sky?

Chris Miles: 0:55

I think it's at the cloud main.

Tim McConnaughy: 0:59

Okay, at the cloud main. So yeah, make sure you follow the cloud main. I'm no longer BGP main.

Chris Miles: 1:06

He left that behind and now he's in the cloud. I need to buy that domain and then maybe I'll switch it. That'd probably be the smarter thing, to do.

Tim McConnaughy: 1:12

Yeah, so this week we actually just got back from AWS reInvent Chris and I did. We both went there to support our employer, yeah, so, yeah, it was a good time. Met some people we hadn't seen in a while, met some new people that we knew but hadn't seen, and I don't know if we met anybody we hadn't already known online, but there was a lot of people there. It was like 57,000 roughly people in Vegas this time. So, yeah, it was pretty nuts people in Vegas this time.

Chris Miles: 1:46

So, yeah, it was pretty nuts. I definitely met some people I had not met in person. Plenty of older guests of the show, like Sharla at Gifted Lane I met her in person. Dejuan finally got to link up with him, and then a few others that just people I've been following on LinkedIn for the longest time. So if, uh, if we, if we ran into each other, if we shook hands, uh, it was great to finally meet you.

Tim McConnaughy: 2:10

Uh, let's see, I'm trying to think, were there any? Yeah, we had Andrew Brown as well. He was, uh, hanging out at the show. Um, I'm trying to remember if there were any other guests that we've had on that we kind of ran into at the show. I'll have to go back through our discography if you will, and figure that out.

Chris Miles: 2:24

But no, it was great, our sworn enemy will. Of course, yeah, of course.

Tim McConnaughy: 2:27

You know, I'm sure he'll laugh when he hears that, uh, yeah, we, we saw will. Um, yeah, no, it was. It was good to finally meet a lot of these people in in person and, uh, spend some time with them, uh, yeah. So yeah, I gave a, I gave a breakout there, which was pretty cool. It's on YouTube now. The sound guy you know put the because because of the polo that I was wearing, he put the mic like really close to my mouth so you can like hear my, you can hear like me breathing and my lips smacking and everything. So please make sure you go listen to this and make sure you go, don't worry about it.

Chris Miles: 3:04

Tim is definitely his biggest critic. I watched the talk and I thought it was great.

Tim McConnaughy: 3:10

I don't think there's anything appalling.

Chris Miles: 3:13

I think the content is good and I think everyone should go check it out. We'll put it in the show notes.

Tim McConnaughy: 3:18

One thing that we both agree on is that, of course, aws reInvent, I swear is like 90% for developers. So finding infrastructure, people there network or otherwise is like really, really hard. So when we did finally get to find some of those people, we talked at length with a lot of them about the challenges of cloud infrastructure. So that part was nice, but definitely it's more of a developer, I think. Focused convention yeah, 100%, I think during Garmin's opening keynote.

Chris Miles: 3:52

Like he said right at the beginning, it was, like you know, this reInvent is all about developers. Like he did not call out infrastructure in any capacity.

Chris Miles: 4:01

So that kind of shows you where things lie. You know it is what it is. I think we've all, we've all known that for some time and that's why, when we have these conversations based around, uh, you know, on-premises infrastructure that has moved to the cloud, um, you know, sometimes you feel so much about it, you start a podcast about it, like, uh, like we've done, but, um, you know, um, it's good to have these conversations it's definitely growing.

Chris Miles: 4:24

I think it's, uh, I think we have probably talked more about this in the last year than than ever before, so it's it's definitely on the uptick. But, yeah, there's no question about who they build this conference for.

Tim McConnaughy: 4:34

Yeah, Agreed. I mean, most of the time, 90% of the talks are are development based, or well, actually 90%, 99% of the talks are.

Tim McConnaughy: 4:41

AI this year, if it's not AI, it's mostly development, and then there's the odd breakout, like mine, with their focus is on infrastructure, but there were a few, so there's definitely some. We'll put a lot of that stuff in the show notes. Speaking of we're gonna cover, for this episode kind of the topic will be a roll-up of kind of the interesting announcements that they made at re, at reinvent. So they were a ton and we'll get those. We'll get. There's good, there's so many more than we're, than they're that we're going to cover. Um, so we'll make sure we get those in the the show notes. But, uh, we, we went through and kind of curated a list of of things we'd be we thought were kind of interesting to talk about. So, um, you want to lead us off with the uh, with the uh, the chip bananas, that, uh, that was thrown out here.

Chris Miles: 5:27

Yeah, I think I think, before we start talking about things that were released at aws reinvent that we found interesting, I think the more important thing ahead of time is to talk about the shit that we do not care about.

Chris Miles: 5:38

Um, which I know I'm I'm speaking I'm speaking for myself, I guess a bit here, like I'm not a chip guy. When they come out and be like, oh, this new chip does this, that I could not give a shit less. Like I understand how it moves the, you know, moves the industry forward. It's always a good thing, but like man, I do not care about, like speeds and feeds, especially in the chip status, right. So it's, like you know, new version of uh tranium 2 instances, which apparently now tranium 2 is used for training and inference.

Chris Miles: 6:08

I thought there was a different chip for entrance, inference or whatever it's called. I can't, but matt garman even says that they don't do, no, don't do well with naming. So they, uh, this tranium 2 chip is supposed to be used for both training and inference. So that runs on a new set of instances. I I also talked about this new Tranium 2 Ultra servers thing which is basically just, it sounds like Tranium 2 instances in a cluster, so I don't know if it's like a cluster of four. Kind of looks like a placement group type thing to me with those type of instances, which is cool. And apparently Tranium 3 will be out, uh, sometime next year. So, um, you know, if you're, if you're into all that, then, uh, that's, that's what you got coming, and I hope you were very excited and uh, you know, and it didn't go in one ear and out the other like it did for me.

Tim McConnaughy: 7:00

Yeah, no, honestly, the coolest thing uh, not coolest thing, sorry. The most interesting thing, should I I say about the Tranium stuff is just that Amazon is leaning very heavily into owning the entire supply chain of generative AI. I remember last year how many times were we joking about how Amazon or AWS wants to be the leader in AI, but they just aren't. They were screaming it from the rooftops last year, but like they, just they aren't like they. They were screaming it from the rooftops last year.

Tim McConnaughy: 7:27

Um, I think this might be what is a a move by abs to try to essentially corner the, you know, the supply, not corner the supply chain, because I mean microsoft, other other chip makers are going to make chips, but just so that they can basically do their own rail, like you know from, from inception to, you know, from, from digging the rocks out, you know from, from digging the rocks out of the earth to producing whatever the output from Gen AI looks like. So that is an interesting thing and, um, is it going to be better than what what other competitors come up with? I don't know, but there's definitely power in owning your supply chain, like you know. So that's that. That part is interesting at least.

Chris Miles: 8:02

Yeah, I mean mean I don't think we need to really talk at length as to what the value is for a business if they can just start deploying EC2-based instances to run things like that, rather than standing up their own NVIDIA infrastructure, whether it be on-prem or anything like that. So that's no question about that and I totally agree. I think they're leaning into making it as easy to consume as possible, because you know they're they're they're the tech giant that they are.

Tim McConnaughy: 8:30

they can do it at, you know, way better speeds than you could ever do it right yeah, in scale, right, we were talking about when we talked about with uh, when we had peter on. We were talking about like, come on, like who, what, what enterprise can truly try to build like an AI data center, you know, beyond the tiniest little thing. And so, yeah, amazon's leaning hard into that and we'll see how that works for them. So, yeah, that again I'm very much in the don't care model of from the technology perspective. I'm not into the addressable infrastructure and all that good stuff, but I think that from a business perspective like that, you kind of appreciate the strategy that they're employing. Ok, so, all right. So Amazon also kind of the big.

Tim McConnaughy: 9:13

What I would say is like the biggest gen I would call the biggest gen AI announcement is the new foundational models that AWS announced at reInvent, which they're calling Nova. And what's interesting about these FMs is well, first of all, of course, they're completely Amazon. Right, they're Amazon-based FMs. They trained and created them themselves, but they're introducing almost like task-focused LLMs, which you've seen a little bit already with like Claude, like they have the version of Claude that's better at writing and you know reasoning models and whatnot. You see a little bit with OpenAI, but this is like really, really focused. And so, actually, as part of the announcement, what they've said is you know, because we've done it this way, it can be more cost effective and less latency in terms of query and response, because these models are very, very, very fine tuned for the specific task. So they have one called Amazon Nova Micro, which is a text-only model, right. So it says quote-unquote delivers the lowest latency responses in the Nova family at a very low cost. So this is almost like the T2 Micro, ec2 instance type of thing, where it's just focused entirely on delivering just text, right. And so it says context linked to 128K tokens, optimized for speed and cost, translation, content classification, text working with text. Like that's what the micro model is. And I'm not going to go through all of them, but, like, just to give you the idea of what that means when we're talking about a task focused model that you know, compare that with, like what is it?

Tim McConnaughy: 10:43

Nova Canvas, which is a state of the art image generation model. This is just like a Dolly type of thing, I assume. Um, you know, quote unquote, studio quality images with precise control, but rich editing features, which I think is interesting. Like I don't know what they mean by rich editing features. I assume the model's doing the editing, but uh, right, um, amazon Nova is canvas excels on humidity evaluation and key benchmarks such as text to image faithfulness evaluation. So does that mean like hey, I'm going to feed you uh an image and can you tell me if it's real or not? Like I'm, I wonder if that's, but, but you can see where I'm getting at.

Chris Miles: 11:19

Like these, these different models are pushing, putting out are are very, uh, task focused, which is which is interesting yeah, totally, oh, and these are bedrock only, by the way yeah, yeah, yeah, but I just want to add that, yeah, it's a great point and it's funny that they're taking this approach, whereas a lot of the other major foundational models are just using like, hey, we have one, we have one that does all of this. I mean, there's there's differentiate, uh differentiation within you know, open eye, that use sora for like the video and the you know image generation type thing, versus uh, you know what chat gpt is doing. But even then, it's only two, two kind of products, right, um, whereas here is, it's like it's all built within nova and, like you said, it's, I mean, if I have, if I'm, I guess, if I'm building in bedrock, this is a. This is kind of a no brainer assuming that the, the content that I get, is good, right, um, so it's, it's, you know there's. It's almost like the uh, the way I think of it is, it's almost like the Apple thing, right, being in the ecosystem has, its, has its benefits, right, um, uh, just being able to use all the direct integrations and things like that.

Chris Miles: 12:29

So, um, I can definitely see it working. Um, but that, you know, proof's in the pudding right, depends on how good the uh, how good the. Yeah, the generation is obviously. That goes without saying, and I wonder what this looks like from a developer perspective, about when you choose which model right? I mean, they're relatively defined as far as like what's text only, what's looking at imaging and video? What's actually generating images and video versus just kind of digesting them? So it seems pretty clear cut, but I don't know how. I'm not a developer at the end of the day.

Tim McConnaughy: 13:01

It's like I don't know how easy it is to switch back and forth between these things right, but yeah, definitely very interesting.

Tim McConnaughy: 13:08

Well, yeah, and the fact that they put it exclusively behind Bedrock is interesting as well, because, remember, bedrock, of course, is just essentially a front-end API, right, that Amazon built to interact with back-end models, so it's definitely very developer focused. But remember, bedrock also, one of the big capabilities they tout is building complex AI agents. So this kind of actually seems to dovetail, like you said, the ecosystem play right, where you're going to build an AI agent and that agent is going to be very workflow focused and now we can actually have our agent actually be very specific about you know which foundational model you know it's using for the specific tasks that it's working on. So, yeah, very, very interesting. Again, this gets back into the what I was saying about the chip thing. Right, amazon's really trying to own the soup to nuts of of what generative value looks like, and I'm also curious to see like, will the developers bite at the end of the day, if it's got a front end API and developers don't have to know too much about you know how the sausage gets made, I think that they're more likely to embrace it, right? So, yeah, I think I'm very curious to see where this goes as well. All right, I'll do one more and which I thought was interesting.

Tim McConnaughy: 14:23

So I did my breakout specifically about RAG at reInvent and I specifically was talking about Bedrock, but also using a disaggregated RAG solution. The idea was that you can pull apart the elements of your RAG, put them wherever you want, and you just need a highly performant network to connect them all and that's how you could do it. So this new capability they just announced for Bedrock I didn't see it before I built my talk, obviously, because they announced it is this idea of RAG evaluation. So this is interesting. So with a RAG, you use a data source, of course, and your data source is it's a file storage, whatever that looks like, right, you're just dumping data into the data source and then you know RAG is doing its thing and it's creating the contextual vectors and all of that, and then vectors could be used for the workflow, for the RAG pipeline. What the capability that they've added is interesting is now that you can use a model. They have a model available that will actually go through your RAG and ingest your RAG, essentially, and look through your data source and all your stuff and just kind of give you a score on how useful the data you have in your RAG is to what you're doing. So it's pretty interesting.

Tim McConnaughy: 15:40

The article goes into basically what they're calling LLM as a judge capability, meaning again that it's going to go through and figure out. So they have check boxes that look at my RAG for correctness, for bias, for whatever, and they even have some of the safety ones like look at it for make sure that I'm not putting out proprietary data. They have a check mark for that and the model actually goes through all of your data store and all your sources and your vectors and everything, and it spits out a report basically that says, hey, your RAG is this useful to what you're doing? And then, of course, as you add new content or hydrate it, refresh it, replace it, you can keep doing that and so you can decide basically am I doing the right thing? Am I putting the right data in my RAG for what I'm doing?

Tim McConnaughy: 16:29

So it's a pretty interesting capability. It's something we haven't really seen. That I haven't. I haven't seen before this idea of of of not just, you know, using your own data to get generative output, but also using kind of reversing. Take, take in an LLM and then you know looking at the data you're putting in and saying how useful is it to what you're doing? So pretty interesting.

Chris Miles: 16:49

Yeah.

Chris Miles: 16:59

I yeah, I wonder how soon they implement a Nova judge model that is used just specifically for evaluating this. I mean, albeit, I'm not far from a rag expert, so I'm assuming this is kind of a preliminary thing, right? Like you get it all set up and you evaluate it using this LLM as a judge thing and, like you said, over time you periodically run this against what you're doing with some kind of I think you feed it, you give it like sample prompts to do this right, Like you tell it to do something and it evaluates how much it's using the RAG data versus the kind of the base LLM that it's using for the content and how useful that data is.

Tim McConnaughy: 17:33

That's right. It's exactly what you do you. You put in this part. There's a little weird to me. I looked in it. I'm like, okay, I guess, I guess this is how it would have to work, but you actually have to create a set of sample questions like prompts and sample responses to give to the model, to give it so it understands. So it understands what it's like. You know how to curate your rag data like. So that part's pretty interesting, so you could still there's still plenty of space to shoot yourself in the foot, blow, you know, and like not do it right.

Tim McConnaughy: 18:01

You basically have to give it the answer key before you generate this right yeah, you have to say, like, this is the kind of prompts we're expecting this rag to care about and this is the response type of response data, that that we expect to get out of it. And then it'll go through all of your data sources and say how close, essentially, how closely that matches your. You know, and, and, and, on what vector and what axes, like I said, the correctness, uh, brevity, like, oh, there's a whole, there's a whole bunch of them, um, so it's an interesting idea. Okay, yeah, that's cool.

Chris Miles: 18:38

We're. We're swiftly approaching this kind of paradigm where AI is doing every little thing in the stack and evaluating its own performance. I don't know what that says about where we're going, but that's you know. That's something.

Tim McConnaughy: 18:51

Well, you know, we met Skynet. Yeah, who cares? It's fine.

Chris Miles: 18:57

Well, I'll be dead by then, anyway. Uh, exactly, all right, one one, one one uh quick one.

Chris Miles: 19:08

I'll touch on here that I thought was kind of interesting as well as, um, albeit, this is, this is even funnier because, like I said at the beginning of the episode, I'm not a chip guy. I don't care about, you know, compute speeds and fees. One other thing I really don't care about is storage. Could not give a shit less about storage. Yeah, um, but I did see an announcement here that I thought was relatively interesting. Um, about s3. Um, obviously, s3 is kind of, I think, amazon's most success or aws's most successful service.

Chris Miles: 19:37

Right, it was the first thing they launched and you know, it was like even even the original yeah, even even um verner vogels's talk about this kind of talks about you know, the ideal building these complex systems and the more you make them simplistic for the customer, like it grows in terms of complexity. Um, pretty good, talk it always. It started out with, uh, you know, a theatrical thing that they built where they were, like he was. He was talking about it was. It was quite funny. I'll touch on it real quick. Uh, he was talking about how like they wanted him to get they were doing a documentary about the early days or something like that. And, uh, he was telling them to talk to interview this guy, al or Alex, and they accidentally interviewed the pizza guy and he was talking about they were like giving pizza guy the credit for all of the things that they did within.

Tim McConnaughy: 20:29

I remember this video.

Chris Miles: 20:30

What's NS3. He's like he was. He was responsible for the shardinging and all this complex stuff that they've done in the system, which was pretty funny but nonetheless. So an announcement that they had was this new bucket type that they've launched within S3, called S3 Tables. Basically it's built upon Apache Iceberg.

Chris Miles: 20:56

If you're a storage database person, if that means something to you, then that's great. It doesn't mean much to me. All I know is it's kind of like seemingly more like a typical database that you're using. But then they also mentioned that they're releasing this thing, a queryable object metadata for your Amazon S3 buckets. So typically, at least my understanding is you know, obviously there's metadata about your S3 buckets, right? There's not a lot of metadata about the objects within the bucket. So you know you're limited to those objects having tags and things like that, and sometimes querying that data can be relatively difficult and hard to maintain.

Chris Miles: 21:39

Um, so there, with this new bucket type, this is basically going to be an index of all the metadata that's attached to your objects and about your objects.

Chris Miles: 21:48

The example that Matt Garman gave was you know, if he goes on his phone and he wants to find a photo from Las Vegas, you know, five years ago with so-and-so.

Chris Miles: 21:57

He just kind of types that in and it and it searches and it finds the photo and he kind of goes through the process of how that works. Um, and you know, tim and I were talking about this before the, before the show, and we really ultimately just seems like like like an indexing of data about the objects in your buckets and points to them. It's kind of just like a Dewey decimal system for your S3 objects is the way I think about it, which is at the end of the day, it's like you're like oh okay, that sounds cool and everything, but how is it used? I mean, obviously, the main use case for this I think is going to be AI, just like going in and querying these things that are in S3. It's in preview right now, so it's not GA as of yet, but it does sound relatively interesting and obviously can be applied to much more than just AI. But I think that's what everyone's going to lean on right now. But I don't know anything to add, tim.

Tim McConnaughy: 22:47

No, I think you got it. It's interesting to me the idea of tabular, essentially indices, like an S3 bucket, that's, tabular indices, because we're really blurring the line between what a database is and, like you know just what object storage is, Because you know, like, that's how databases work. Right, they're indexes and then the indexes point at the actual data and that makes the look up very quick. So and I agree with you that AI probably will be the main use there I mean we were just talking about RAG, for example right At the end of the day, if you're going to very, very quickly find data in your storage, it's this kind of stuff that's going to make it extremely fast to do so. Yeah, completely agree on that, and I'm not a big storage buff either I never have been but I mean anything you can do to make it faster, better, stronger. I'm curious from a pricing perspective, what that looks like Because, for example, your S3 buckets generally are charged by the transaction.

Tim McConnaughy: 23:45

So now, we're adding a third bucket, we're putting another bucket in between our application and the data specifically. So from a charging perspective, are they going to do like a .001 charge every time we look up something in the table? You know?

Chris Miles: 24:01

that is a great point. I did not consider, Well, I mean maybe that maybe this bucket type does have severely lower charges for these querying options.

Tim McConnaughy: 24:11

That's the only thing that makes sense. I mean, I would hope so.

Chris Miles: 24:14

Then, like you know, you're querying the shit out of this one bucket but you never actually go. You know, look at the objects unless you actually find them to be. Talk to the object. Yeah, but that's a good question.

Tim McConnaughy: 24:26

Maybe you save money, maybe you save yeah, yeah, I don't know if they even talk about pricing in here at all.

Chris Miles: 24:32

Uh, oh, no, it says for more information, yeah, go to the s3 pricing page. So, um, we'll have to look into that, but I'm not going to sit here and look through that page on the pod. Yeah, we'll do that later.

Tim McConnaughy: 24:43

Um, I mean, I can't imagine they would do it to make it cheaper. So we'll see.

Chris Miles: 24:47

It's true it is funny because it feels like we are approaching this. We're approaching this like thing where we're gonna have so much data about other data, like you like we're going like three rungs back at this point where it's like um, you know a photo of a cat or whatever the fuck it is, is gonna have you know 10 different pointers to it. Um, for, for reasons unknown, exactly yeah, for for.

Tim McConnaughy: 25:12

Yeah, for for other application purposes. Right For for making it faster to access, faster to use it. Um, and we've actually. It's funny cause, then you know I we've been dealing with this for a while with from, like, streaming telemetry perspective. Right, there's just so much data about data you know, like if you have a day you need your own data, like just to to log all the data about what your data is doing.

Tim McConnaughy: 25:35

Yeah totally, yeah. Anyway, it's pretty interesting. All right, let's move on to the next one. The next one is about Kubernetes, and I actually think this is really cool because not because of the specific thing that it does, but because it's kind of an implementation of something we've talked about for a long time, which is the idea of what hybrid cloud infrastructure is really going to look like in the future. So this one came out of reInvent. It's using your own on-premise infrastructure in Amazon EKS with what they're calling EKS hybrid nodes.

Tim McConnaughy: 26:07

Now, remember, eks is the Amazon managed Kubernetes solution, right? You've always been able to run Kubernetes wherever the hell you want. That's always been the case, right. You've always been able to run Kubernetes wherever the hell you want. That's always been the case, right. This is EKS, this is an elastic Kubernetes service. So what they've announced basically is that, as long as you have, you know you meet the requirements, which is essentially and I looked into it more or less it's just routing between, and reachability between, your on-premise infrastructure and Amazon EKS. You could actually bring into EKS management your own on-premise Kubernetes clusters, right, so you can make it part of the EKS control plane, data plane, all of that good stuff, and to me that's kind of like the hybrid cloud story.

Tim McConnaughy: 26:50

Right, we always have talked about how we need the cloud experience. On-prem is really the golden ticket, and this is kind of I mean, it's not that right. Like we always have talked about how we need, you know, the cloud experience. On-prem is really the golden ticket, and this is kind of I mean, it's not that right, but it's a, it's a step in the right direction and it's and it's showing how, as we move more, as we, as we kind of as, the pendulum swings back towards the middle. You know, after, as it always will, it'll swing to the side, swing to both sides, and then eventually it'll end up in the middle. You know these, this type of stuff I think we're going to see more and more.

Tim McConnaughy: 27:17

Right, hybrid infrastructure is how the cloud providers are going to stay relevant. Like they'd love you to put every, every, every bite of your data in the CSP, but I think that they are understanding that enterprises and businesses are are not doing that for cost reasons anymore. So they're like, okay, well then, we're going to make it easier for you to do this stuff so that you'll stick with us. So I think this is great and I'm glad that we predicted something like this. I don't know a year ago or however long it was that we started talking about what this was going to look like. Anything to add there?

Chris Miles: 27:52

No, I agree with you. I think it is pretty cool. I wonder now, they do have quite a bit listed in here as far as prerequisites for something like this. But one thing I do wonder is because the cloud service providers are usually very, very good whenever they manage the entire system right.

Chris Miles: 28:11

Whenever they own everything, oh good call Whenever they own everything, the you know kind of this. Things are relatively seamless, the integration is great. I do wonder what the caveats are here. With your own on-premises running of Kubernetes, right, I wonder what limitations they run into. It's like you know, maybe I can you know. If I'm just doing node management or something like that, then that makes sense. But like, what do you have in terms of scaling and things like that? I just wonder what the limitations are. And then there's pages and pages about these hybrid nodes and their prerequisites. So there's probably some strong caveats in there. But I'd be interested to actually dive into those.

Tim McConnaughy: 28:52

Definitely a V1, I mean, as it needs to be. Another thing of course is not only you're right. When they manage the whole path, then of course they can kind of do it their way and it makes everything smoother for them, right? Um, of course also, this also makes them much, far more sticky right. They get their, they get their, their claws, indie on-prem right right from a management perspective. So it's.

Chris Miles: 29:15

It's funny because this is like one of the first. Uh, it's definitely not the first, but it's one of the first times we see it going in the opposite direction, where a managed kind of a management system that runs in AWS is now extending to control something outside of AWS.

Tim McConnaughy: 29:32

It's usually the other way around.

Chris Miles: 29:34

Right, maybe you had something on-prem, maybe you're running VMware on-prem and you want to be able to extend that into the cloud. So it's more like taking that construct and moving it to AWS. This is the other way around. Um, so that kind of I think speaks to where people are starting, uh, with a lot of this stuff, um, um, or at least where they're trying to evolve to get to Um, so very, very interesting for sure?

Tim McConnaughy: 29:57

Yeah, I mean, outposts is kind of what they were pushing, and of course, outposts aren't going anywhere. I'm sure that outposts are still a thing. This is just like. This is an outpost, though, right, this isn't like hey, we're bringing in a lot of management infrastructure, but even that outposts is exactly that right.

Chris Miles: 30:14

Whenever, they can own the entire thing if they exactly how this is built, exactly how it runs.

Tim McConnaughy: 30:22

We can troubleshoot it, we can manage it. Yeah, exactly, so very interesting. I am curious to see. I also want to see the adoption. I really wish that we had some way to tell who's adopting, who's doing the V1 of this, to like, hey, yeah, wow, we would love to have one EKS, you know, cluster type of management system, you know, yeah, so I think that's really the question is, who's going to? Who was waiting for this to run Kubernetes on-prem that wasn't already doing it? Like that's, you know what I mean. Like that's interesting. So time will tell on that one.

Tim McConnaughy: 31:00

Yeah for sure, let's see. You want will tell on that one. Yeah, for sure, let's see.

Chris Miles: 31:02

You want to cover the last one.

Tim McConnaughy: 31:04

No, no, I've been talking enough, man. This one's yours.

Chris Miles: 31:07

Yeah, I'll take it. So, yeah, we'll close out with this one, because I just thought this was quite strange. There's obviously, you know, kind of we see vendors do this all the time where they release some kind of feature or something like that, where you're like man, who which big customer asked for this? And this is kind of puts me in that, in that mode where I'm wondering who, which one of their Fortune you know 50 customers asked for this type of thing. But so they announced the option for an AWS data transfer terminal. So they announced the option for an AWS data transfer terminal.

Chris Miles: 31:42

So this is a physical location where you are able to go to. Basically, you're led into a room that has access to a fiber optic cable. I think it even says what the room is equipped with. Let me see what it says. I think you basically go into the portal, into the AWS console, create a transfer team where I think people you're delegating, like where the data goes, who has access to the data, et cetera, and then you basically reserve a time to visit that location and then it says, once you go in, it's kind of an unmarked location.

Chris Miles: 32:19

So there's no AWS branding on the building, the room, anything like that, just for security purposes, they say. It says the room is equipped with a patch panel, fiber optic cable and a personal computer. The patch panel is installed inside a wall mount rack or a small floor rack to allow for additional space on the desk of table. With your personal computer you can see how to remote access to the server during your data transfer process. So I don't know if you're like I don't know what you're hauling into this thing, like they give the example of like bringing in your snowball devices which you know the AWS compute that runs at the remote site.

Chris Miles: 32:56

You need to come in and bring it in to push the data up to AWS. But yeah, it says right now I think they have locations in Los Angeles and New York and Jeff Barr even kind of outlines what his experience was like going into one and uploading some data which I believe was used in South Korea or something like that. But yeah, man, very, very interesting to see this.

Tim McConnaughy: 33:23

I want to back my snowmobile into this thing. I want to back my 18-wheeler into this thing and plug my fiber optic cable into the truck.

Chris Miles: 33:33

Just whip open the doors. 20 billion petabytes yeah right, Just run it inside. Like an undersea cable, like one the uh undersea optical cables and just plug it in it sounds so dystopian man like that sounds like that sounds like some tron shit or something like it sounds so weird, like it's like you're gonna plug your brain in, or something like yeah, like uh, very neuromancer style almost right like I'm waiting for the, the armed guards that are, you know, wearing briefcase.

Tim McConnaughy: 34:04

You know, with briefcases and inside the briefcases are guns, or like, you know, if you, uh, you need a retina scan and if your retina scan doesn't go through, the, you know flamethrowers drop from the ceiling or something like it's crazy stuff. Man, the way they've described this just sounds so dystopian. Yeah, I don't know. It just sounds like some movie, some movie shit like tron or something. I yeah who was asking for it? Like who was like man. You know what I really need? I just need a physical location that I can go to and just hand you all my data and just, and I need a little pc sitting on a desk so I can monitor my data transfer the whole time. It's weird.

Chris Miles: 34:43

What I want to know is I think it's probably somebody using a lot of these snowball type things. Because we talked about these industrial verticals where customers are using this a lot.

Tim McConnaughy: 34:54

Yeah, it's IoT.

Chris Miles: 34:56

So you probably like, but then the people on site are typically not very technical, so they probably needed to make this as simple as possible, like somebody's just like look man, I'll have a, I'll come in, I have a, I have a bucket of fucking 50 of these things. I just want to plug them in and then go home like that's right yeah, I imagine that's smart hand style.

Tim McConnaughy: 35:15

Yeah, exactly, um, that's what I'm thinking. Yeah, that might be, I don't know the case. What I want to do is find one of these locations and then open up a similar location right next door to it, and then I can just be like the man in the middle. They'll bring me their data. I'll just snatch all their data, because everything's unmarked right.

Tim McConnaughy: 35:34

I'll just be like yeah, I'm the Amazon guy, give me your data and I'll just plug it right here. It's weird, I don't know what to say. It's besides, it's weird they do mention that, like you know, the, the use case is very much like uh, you gather up a bunch of ot style data, like the driving, like smart car data and stuff like that. Who is but who's doing the like? Now I got this feeling like we just have, like, like you said, like 50 hard drives inside you know, harvested from a fleet of cars or something that are being plugged into this thing. Um, what do they do? What were they? What were they already doing? I thought that a lot of this stuff was already going over the air back to you know some collection location. Right, what were they doing before that?

Tim McConnaughy: 36:16

Were they doing this themselves, Were they, you know, like. So anyway, it's an interesting choice, interesting product announcement. I'd love to know who's using this. If you're using one of these, man, just like, let us know. Drop us a line, I'd love to know about it.

Chris Miles: 36:32

Let Tim be your personal escort to go to one of these things with you. Yeah, yeah, yeah, give me a badge. Yeah, give me a badge, yeah.

Tim McConnaughy: 36:39

Give me a badge and I'll go in and do the whole Johnny Mnemonic thing or whatever the hell this is going to be there you go there you go Anyway, all right.

Tim McConnaughy: 36:48

Well, that just about wraps us up. Obviously, there were tons more. There's tons more to talk about, but keep it nice and succinct and on task here. Definitely had a good time in Vegas. I don't know what else to say. I was one of zero. I was one of zero members of my team, other members of my teams that didn't get sick. I was the only one on my team that didn't get sick. Uh, everybody else got sick and I did not get with covid or flu or something. Uh, one got covid, one got flu. Uh, one is sick and I don't know that he determined what he was sick with.

Chris Miles: 37:20

But yeah, everybody else got sick so I got my COVID booster a month and a half yeah, he was like oh my god, I'm so tired of it.

Tim McConnaughy: 37:31

I got my COVID booster and my flu shot like a month and a half ago in anticipation. I actually took this whole week off of work because I expected to get sick and just be laid out and now I'm not. So maybe I'll just, you know, maybe I'll get some of that. It's unlimited pto. Maybe I'll get some of that pto back and work. Who knows?

Chris Miles: 37:47

it's either. It's either, uh, the utmost productivity or the least amount of productivity. However, you want to spend it, bro you got it.

Tim McConnaughy: 37:53

You know, do it exactly.

Chris Miles: 37:55

Um, but yeah, just just to kind of tack on to what tim said. This was. I think this, think this was this was your first re-invent as well, right, Tim? It is.

Tim McConnaughy: 38:02

Yeah, it was, so that was my first one.

Chris Miles: 38:04

Obviously, tim and I were both on booth duty most of the time, and Tim even had a talk. I did a few theater presentations, so we were kind of completely locked in the whole time. We didn't really get the full conference experience, I would say, but overall, man, man, this conference was like I really got to give it to AWS. Man, they thought of pretty much everything. There's so many small touches that I've never seen at a big conference like this, which I thought was really cool and it's just so fucking big man. It is, oh, man, because we've been going to cisco live for years and that's always, you know, somewhere in the ballpark of like 25, 30, 000 people, right. So it felt like a lot, but this was like, yeah, double, um, apparently some years nearly triple that right so it's like this is so huge because it's the cloud, is just an extension of the it stack right now.

Chris Miles: 38:58

So it's now just a big it conference and a lot of them are developers. So if you're an infrastructure person, you will feel a little underrepresented, but in the I will say, in the expo there's a lot of really cool vendors, a lot of people that we actually got to talk about and will potentially be having come on the show and talk about some cool problems and ways to solve them and things like that. So, yeah, really exciting it was. It was good. I definitely hope I get to go again next year. Yeah, absolutely.

Tim McConnaughy: 39:26

All right, everybody, uh, we'll go ahead and wrap up here. Um, make sure that you subscribe to us. Uh, you know, uh, hit us up on your favorite dating app, especially especially Chris he's. Um, you know, like I said, buy the breakfast cereal, play the at-home game, follow us on Blue Sky I think we're just at cablestocloudscom on Blue Sky and, of course, all the other socials where you're hopefully going to follow us. So, thanks for joining us and we'll see you next time.

Chris Miles: 39:59

Hi everyone. It's Chris and this has been the Cables to Clouds podcast. Thanks for tuning in today. If you enjoyed our show, please subscribe to us in your favorite podcatcher, as well as subscribe and turn on notifications for our YouTube channel to be notified of all our new episodes. Follow us on socials at Cables to Clouds. You can also visit our website for all of the show notes at Cables to Clouds. You can also visit our website for all of the show notes at CablesToCloudscom. Thanks again for listening and see you next time.