00:08
Thank you very much, Charlie. It is great to be with you this morning and we're excited to be here. I'm gonna open with a little picture, and I'm gonna ask you how many of you can guess where this is. This is the happiest place on Earth, but it began not quite so happy.
00:29
In 1955, it was announced with great acclaim and excitement, and everybody in the world knew this was going to be the greatest event of the year. Everyone showed up, movie stars, celebrities, the California governor, and so did 17,000 uninvited guests. With counterfeit tickets.
00:52
This is a problem. This is a problem. The infrastructure did not keep up. The Mark Twain River boat was so overloaded that it started to sink. There wasn't enough food. The asphalt was so hot that high heels began to sink into it and people got trapped.
01:10
And there wasn't enough bathrooms. So this was so chaotic that the staff nicknamed this Black Sunday. Of course we love Disneyland today, but on that day the hype far outpaced the infrastructure. We would like to speak with you today about your AI Disneyland. And as you open the gates, how you can be ready for the demand.
01:37
As you know, the pace of AI innovation is unprecedented. What has begun with a super smart chatbot in your, in your pocket is now rapidly evolving towards your job. And your AI strategy depends on the underlying data infrastructure, how data is governed, how data is ready, whether it can be prepped and ready to be ingested, and your underlying AI infrastructure.
02:02
And so what we'd like to do today with you is get a little bit introspective. Across all of our customers, we surveyed them and we found 3 major categories of companies that consume, extend, and build AI. We are introducing the Enterprise AI Capacity index beginning with companies that consume as a service and self host their own AI models and then extend it and enrich with your own data to
02:29
fine tune it and then build and maybe even at cloud scale. And as we do this, we would like for you to imagine yourself in this mix and identify yourself and which group that you're in. And when you walk away today, the number one question we'd like to ask you is. What is the right next step for me?
02:49
Let's begin with as a service AI. These are companies that integrate cloud AI services with enterprise data and are just starting. We typically find that data is highly siloed without many audits or or caveats controls. If this is you, the next right step might be to unify your enterprise data so you can add governance and
03:12
get ready for ingestion. And then it's self-hosted AI where customers want to deploy AI models on their own premise with enterprise private clouds because AI has become important to you now, it's mission critical. You may need to start adding mission critical availability.
03:30
And if so, the next right step might be to simplify and to automate. No more fingers on keyboards. The next category of of companies are those that optimize AI models by enriching with their own data, and they typically find that there are many versions of your data. So you need to add accountability and add validation for for what lineage your data is
03:52
coming from. The next right step for you might be to add that accountability. And then our most sophisticated, complex customers are looking at training or serving their own models or agents, and this is the wild wild west. You don't know what you don't know, and so we need to add elasticity and optimize your costs.
04:11
And the next right step might be to think about how to add that ability to scale up and scale down. And last cloud scale AI where you have massive scale at exabyte level capacities, and you need to go and understand how to grow to these levels. But maybe the next right step for you is how do I turn on my mainstream IT so we're no longer managing these islands of boutique AI
04:33
infrastructure. OK, so we painted these 5 different levels, and we'd like to ask you, what did you imagine? What was the next right step for you? Maybe it was simplify and automate. And if so, we would like to imagine something a little bit cool.
04:50
What if you could take at your own AI agent that could work through this intelligent control plane that would help you simplify and automate this virtualized pool of storage? What if you could do this out of the box? Well, we're gonna demonstrate that here for you today. MCP is the new cool technology.
05:11
For allowing AI agents to be able to interface, interact with new with your infrastructure. Um, this stands for Model context Protocol. And what we're gonna demonstrate today is Claude, our friend Claude from Anthropic. Interacting through MCP interfaces from for monitoring tool named Gravana, a MCP interface that we built on top of Fusion, and a secret vault.
05:35
We're going to ask it a very simple prompt provision of 5 terabyte machine learning training volume set up monitoring for it and run a training simulation over it. Now this is just simple automation. What's magic here is that Claude knows nothing about storage. It's completely oblivious to storage, and we're going to train it in a matter of seconds using
05:56
MCP and pure fusion. Everything that it needs to know to be able to automate this. We're going to take a 6 to 8 hour job of a very advanced administrator, and we're gonna automate that in 60 seconds. You can imagine the power of this, but the magic here is that your business logic is
06:16
encapsulated in this policy in this preset, and Claude is gonna go ahead and and decide on its own how to name the volume, the right preset associated with this workload, how to train, how to monitor, how to run the simulation, and it'll do it completely on its own, and all that pure is gonna offer is a little bit of context, this prompt, and the basic credentials.
06:40
OK, are you ready? It's gonna go by fast. 60 seconds of automation, buckle up. Here we go. Welcome, Sean. How can I help you? Let's provision a 5 terabyte machine learning training volume. And let's make this work.
06:54
OK, I'm going to help you set this up. Let's first communicate with fusion. What's the language of storage? What's your preset? I see a ML analytics preset sitting here. Let's run this and figure out how to run this. OK, we're gonna provision the storage. Now I can see a naming pattern.
07:09
We're gonna name this ML training data set 4 and provision the volume. That happened quickly. It's now provisioned. Let's figure out how to set up monitoring user in Gufana. You're gonna wanna know how this is actually running. So we're going to add IOPs, read bandwidth,
07:26
right bandwidth latencies. We've now set up the The dashboard and let's run the ML training simulation. I'm gonna use a heavy mix workload that works in ML situations with heavy reads and writes, and it's now running. Let me check the current status.
07:45
Let me look at the right metrics and check the grafino dashboard. Everything is set up successfully. It's all automated. OK, volume provisioning is complete. You've got the right set of performance tier, the right snapshot policy, the right QOS, the right level of encryption, and all the dashboards are set up.
08:04
Let's go through and just click on that dashboard and see what it shows us. There's our panels for latency. IOPs, bandwidth, and more importantly, all the CPU memory application usage that you would expect. There you go, all complete, 60 seconds. OK. So let me interpret for you what you just saw.
08:32
This is science fiction. This is not just automating simple tasks and writing down the task to run. This is the fact that the AI knew nothing about storage. Everything was pre-encapsulated. All your logic about the right things in that preset for ML were already pre-written down and it existed independently of the underlying
08:52
storage. And so can you imagine the power of that of writing down your business language in your head and putting it down on paper and then an AI comes in and lives within those guidelines and makes sure that you're totally governed and you're compliant. What you just saw is the power of the enterprise data platform with an intelligent control plane working in an AI ready
09:14
infrastructure. And so today you may have seen multiple things that you thought I could do that thing. I need to unify my data plan so I could add governance and controls. I need to add an intelligent control plane so I can have no fingers on keyboards and I can automate. I need to add the ability for my mainstream IT
09:32
team to take over for this. All of these things are available to you as a result of the intelligent control plane, and if this seems interesting to you, we would like for you to take the next step and get started with us, invite our AI experts to come meet with you, and we want to collaborate with you.
09:48
To help make sure that you have enough food and you have enough plumbing. And there's no counterfeit tickets coming in and overwhelming you. Thank you very much.
