Microsoft Ignite 2024 Unpacked | Phil Curran, WEKA

Clips
News
More from Microsoft Ignite 2024 Unpacked

Phil Curran

Cloud Strategist

WEKA

From on-prem to Azure: Optimizing AI workloads with hybrid data solutions

As artificial intelligence and high-performance computing workloads evolve, hybrid cloud data performance is becoming essential for enabling scalability and efficiency in modern infrastructures.This integration is driving seamless data movement, expanded capabilities and improved performance. Key developments include integrating accelerated hardware, such as graphics processing units, with highly performant data pipelines that handle massive datasets, according to Phil Curran (pictured), cloud strategist at WekaIO Inc.WekaIO’s Phil Curran talks with theCUBE about hybrid cloud data performance.“The infrastructure is fundamentally different to support these types of workloads,” he said

play_circle_outline WEKA's high-performance data platform for AI and HPC workloads in Azure.

play_circle_outline Customer success stories in AI and life sciences leveraging WEKA in Azure.

play_circle_outline Integrations planned with Azure CycleCloud and Azure ML for future collaborations.

Info
Transcript

Phil Curran, WEKA

Phil Curran

Cloud Strategist WEKA

Microsoft Ignite 2024 Unpacked showcases how organizations utilize on-premises and Azure with the help of ISV ecosystem, specifically in AI and HPC workloads. WEKA's data platform has shown impressive performance for high-intensity workloads like HPC and AI in Azure, with speeds two times faster than other benchmarks. Customer success stories highlight WEKA's impact on AI and life sciences in hybrid cloud workflows. Data portability and integrations with Azure services drive future developments. Customers can easily get started with WEKA through the Azure Mar... Read more

explore Keep Exploring

What characteristics and requirements do high-performance workloads like HPC and AI typically have in terms of data processing and infrastructure support? add

What is an example of a hybrid use case in life science utilizing WEKA software for capturing medical imaging data on-prem and running analytics workloads in the Azure cloud environment? add

What integration is being demonstrated at the show between WEKA, Azure, and NVIDIA? add

bolt Powered by CUBE AI

Phil Curran, WEKA

search

Rob Strechay

>> Hello, and welcome to Microsoft Ignite 2024 Unpacked. I'm Rob Stretchay, managing director with theCUBE Research. One of the things that we're continuing to unpack today is really the technology ecosystem and how organizations are looking to span on premises and Azure and how really there are a robust ISV ecosystem that is working to make this true, especially in AI and HPC workloads. To help me unpack this portion of it and really dive into how the data moves, because that kind of feeds AI, I'm really excited to have Phil Curran, who's the cloud strategist at WEKA to join me. So welcome on board, Phil.

Phil Curran

>> Awesome. Thanks Rob. It's great to be here. Looking forward to the conversation.

Rob Strechay

>> Yeah, and I appreciate you joining us live from the hotel at Ignite. Again, this is fantastic. Again, it's wonderful that Wi-Fi and everything and the internet works this way.

Phil Curran

>> Yeah, it's awesome to be here. There's such an amazing energy happening here in Chicago this week. The keynote with Satya yesterday was amazing. We've had some amazing conversations around the WEKA environment and talking with Azure and our joint customers, and so looking forward to unpacking all that with you.

Rob Strechay

>> Yeah, so let's dive into what the relationship is between WEKA and Microsoft Azure and really give people kind of a little bit of an idea of what WEKA is so that they kind of get some grounding in how it fits into the ecosystem.

Phil Curran

>> Yeah, absolutely, Rob. So WEKA is a software-based high-performance data platform that enables customers to run their most performance intensive workloads, things like HPC, like AI, in Azure. Now, the thing about these workloads is that they're sort of fundamentally different. You even heard about it yesterday in the keynote. This is accelerated infrastructure. We're starting to see GPUs. We're starting to see offload processing. The infrastructure is fundamentally different to support these types of workloads. And they have varied data performance characteristics. They're usually in some kind of a data pipeline. They require handling of millions of tiny files with sub-millisecond latency, in other words, incredibly low latency, while at the same time delivering massive IO and high bandwidth. And they're usually massive data sets. They're usually petabyte scale, approaching exabyte scale for a lot of these workloads. And so what we're doing with Microsoft and in particular within the Azure ecosystem, is providing that massively performant and highly scalable data solution that can really power these AI and these HPC applications so the customers can do more in Azure with those workloads. And so we're working through product integrations with the team. I'm sure we'll talk again in that. And we're helping customers often bridge from an on-premises HPC or an on-prem AI environment and into Azure. So customers are finding that in addition to the high performance they get in Azure, that we're actually able to bridge from on-prem into Azure with the customer.

Rob Strechay

>> Yeah, I think we will unpack that a little bit more, but first you kind of hit on the performance train and with AI and keeping GPUs in particular fed with data is super important because they're expensive, and data feeds the beast that is AI. How does WEKA's recent spec benchmark demonstrate its performance, especially around EDA workloads?

Phil Curran

>> Yeah, awesome, thanks. So for the folks tuning in, so just ahead of the Microsoft event here, we released a new spec benchmark focused in this case on EDA workloads for WEKA running in Azure, and we were able to demonstrate, running natively in Azure, a level of performance for EDA workloads that is two times faster than any other benchmark that's in spec, including, and that's on-prem or in the cloud. There are multiple submissions across both deployment types, and we're actually two times faster. That benchmark was actually over 7,000 jobs running at sub-millisecond latency, so massively scalable and incredibly highly performant. And again, the ability to deliver that massive performance in the cloud, but deliver line rate on-prem performance is kind of an amazing feat and the main reason customers turn to WEKA for their high performance workloads in Azure.

Rob Strechay

>> Yeah, let's kind of dive into, because I think nothing brings it home like customer win stories. And some of the things that you're highlighting and some of the success that you've been highlighting with Azure, it includes the impact on AI and life sciences from a hybrid cloud workflow perspective. Why don't you dive in and help us understand how are organizations really reaping the benefits, especially from an AI and a life sciences perspective?

Phil Curran

>> Yeah, absolutely. So life science is another amazing use case for us. And again, it's one of these workloads that just has this massive data and performance requirements. So an interesting customer that we're working up actually up in your area in the Boston area is one of the major research hospitals that's focused on cancer research, and they're actually a hybrid use case. So what they're doing is capturing medical imaging data in their labs and then bursting that into the cloud, into Azure, to run their analytics workloads. Now, the way that they're leveraging WEKA is capture the data on-prem, feed that into the Azure analytics environment, and run that high performance data processing in Azure, and then capture the results and then archive it all down. Now, the interesting part about that is because WEKA is software focused, the same software runs on-prem in the cloud, so it's very simple and seamless to manage. And because we leverage things like autoscaling, customers can actually scale up an entire Azure environment to run that environment, and that's exactly what this customer's doing, and then decommission that entire analytics engine, capture it all on-prem, and optimize their use of the Azure infra. So, fast, flexible, and highly cost-effective.

Rob Strechay

>> I think that to me is key. And again, having been at a hyperscaler myself, it's being able to move that data around, because data has gravity, and the right data sets in the right place at the right time for the data scientists. So funny enough, GPUs aren't the only thing that are expensive, in high demands. Data scientists are as well, so keeping them busy to find cures for cancer and stuff is pretty important. Let's kind of roll on a little bit more and kind of look towards the future a little bit with what upcoming integrations are planned between WEKA and Azure and those Azure services. I know there was something around Azure CycleCloud and maybe Azure ML. Why don't you kind of unpack that a little bit for us?

Phil Curran

>> Yeah, absolutely. And so especially with Azure CycleCloud, that's actually a demo we're doing here at the show. It's been incredibly well received by folks. We're actually running a joint demo between WEKA, Azure, and in this case, NVIDIA, because as you know, many of the CycleCloud virtual machines underneath there are actually powered by NVIDIA GPUs in NVIDIA environments. So here at the show, we're demonstrating our integration where WEKA is deployed through the CycleCloud automation, and so it becomes very simple and easy for customers who are building their HPC workloads in Azure to then just use that same seamless tools that they're used to now power that and drive that with high performance data through the WEKA environment. And so we're super excited about the CycleCloud integration, and that builds off of a bunch of the keynote announcements that Satya was talking about today that we're starting to leverage, things like the new VM types, the HBv5's, which have eight times higher memory bandwidth. We're super excited about some of those new infrastructure, high performance infrastructure offerings for HPC and for AI, and we're taking advantage of those. And then on the Azure ML side, that's the next wave that we're seeing, is enterprises start to adopt AI, and so that coming up we're going to be working through the integrations with Azure ML and excited to see how that goes and how customers start using that in the cloud.

Rob Strechay

>> Yeah, no, I think that that made a lot of sense. Satya did talk about a lot of silicon that's driving that performance and those new types of VMs as well as some of the storage performance that they're looking to gain out of that as well, which, again, for what you guys do and everything like that, it makes a lot of sense how they're going to unpack that for everybody. But one of the things, just to give the folks at home a little bit, and actually probably the people who are actually potentially sitting there in Chicago as well that watch us a lot, how does WEKA really enable AI and HPC workflows across the different deployments? Because as we were talking about, it may not just be Azure. It may be on-premise. It might be multi-cloud because they're using different services in different clouds. They may be using GPU clouds or other specialty clouds. And what are the advantages of that data portability for customers too?

Phil Curran

>> Yeah, that's an amazing question. It's sort of probably the number one or number two question that comes up, right? Because WEKA is software-based and the same software runs on any infrastructure type... And we have extensive partnerships not only with all of the hyperscaler cloud providers, but also with all of the major OEMs who are building data centers. And I mentioned briefly Nvidia. We have a deep and amazing partnership with Nvidia around things like SuperPOD and DGX and all the goodness that those folks are doing. So as customers are choosing where to deploy their AI workloads, the first piece that customers usually turn to us is, "Gosh, I got to get a hold of my data pipeline problem and deliver. My GPUs are being starved. How do I feed the beast with data?" And we're usually able to come in and deliver meaningful, multiple orders of magnitude improvement in the performance. And that's across any infrastructure type. Now, once customers start to get going with their AI project in whatever deployment environment they have, there's usually a moment where they say, "Hey, how do I start to optimize?" or "How do I avoid a potential concern around a lock-in?" or something like that. And because of the data portability that WEKA provides and the ability to deliver that same level of performance across any infrastructure type, the customers are then starting to be able to start to shift around. And we've absolutely seen that where a customer has started to build, say in a data center, but then they want the flexibility, and so, as we mentioned before, burst up into the cloud or shift into a GPU cloud provider and some of the new specialty clouds that are coming up for AI. We've got deep partnerships with most of the leading GPU cloud providers. And so that ultimate flexibility to sort of burst, shift a workload around, or optimize for cost or optimize for performance, that's the conversation we're having with pretty much every AI customer right now.

Rob Strechay

>> Yeah, definitely. And so folks at home know, how does somebody get started with WECA? You're on the floor today, so they can come by and see you still and see the demos and all of that, but help people understand. I would assume Marketplace is a good place to go as well?

Phil Curran

>> Go to the Azure Marketplace. We are available on the Azure Marketplace. We actually are deployable within a customer tenant in Azure, so you can immediately go and get started. Also, check out our website, weca.io slash cloud, and you can get all the details around our cloud integrations, the different services that we're supporting, and of course reach out to the entire WECA team, and we're excited to get started with y'all.

Rob Strechay

>> Yeah, well, I hope they stop by. There's still a few hours left. And again, enjoy the indoors in Chicago. I mean, again, the weather hasn't been horrible there, so that's-

Phil Curran

>> It's actually been okay for Chicago. It's not bad.

Rob Strechay

>> You never know in November, so it's always fun. But hey, Phil, thanks for coming on board and really helping us drive into why the ISVs are some of the hottest things going with Azure. Really appreciate it.

Phil Curran

>> Yeah, thanks, Rob, for your time. Really enjoyed the conversation, and best of luck through the rest of the show.

Rob Strechay

>> Thank you. And thank you for watching Microsoft Ignite 2024 Unpacked. And stay tuned for more on theCUBE, the leader in tech news and analysis.