Connected Social Media

In this Tech Barometer podcast segment, Debojyoti “Debo” Dutta, vice president of engineering, AI at Nutanix shares his passion for computational biology and how AI will dramatically change enterprises.

Find more enterprise cloud news, features stories and profiles at The Forecast.

Transcript:

Debo Dutta: If you look at what’s happening today, it’s not just because of AI, but in general because of high throughput data that can be generated out of a biological system. And the amount of data that you can generate was unimaginable even a few years ago.

Jason Lopez: In our series on MedPerf we wanted to introduce you to a computer scientist at the forefront of AI and machine learning development. Debojyoti Dutta, who goes by the nickname Debo. His work is bringing AI to IT leaders across many industries, including healthcare. This is the Tech Barometer podcast, I’m Jason Lopez. 

[Related: Exploring Uses for AI in Healthcare]

Jason Lopez: When you hear about AI and how great it will be for healthcare, you might wonder, well who are the people making it happen? Debo is one of those technologists who’s working to connect the dots of computer science and medicine.

Debo Dutta: When coupled with AI, it’s going to completely change the way we design drugs, how we design antibodies, new therapeutics, I can’t even imagine all the things that we will be able to do using this iteration of AI coupled with the amount of data that we can generate out of living beings.

Jason Lopez: One area Debo thinks AI will have a profound effect is in examining medical images. For example, people with diabetes are at risk for loss of eyesight. With AI assistance, images of the inside of the eye could give doctors far earlier detection of diabetic retinopathy. Other imaging examples include detection of tumors and cancers. But another area where AI could bring a transformation…immunotherapy.

Debo Dutta: Immunotherapy in a nutshell is to program your immune system to fight bad actors in your body. The bad actors could be viruses, it could be cancer cells.

[Related: Building Infrastructure to Unlock AI’s Next Big Leap]

Jason Lopez: Our immune system can detect viruses, but sometimes cancer cells are invisible. Cancer cells can appear normal and the immune system allows them to go about wreaking havoc.  In immunotherapy, the DNA of the immune system’s T-cells are changed to recognize the genetic strands of the tumor cell and destroy them.

Debo Dutta: Let’s see how AI gets into this space. So the programming, the existing T-cell can be done with gene therapy, but how do you know what to program? That’s where AI comes in. Using AI, you can rapidly design sequences. People are today looking at a lot of sequence data as well as literature, feeding them into large language models. And they are actually trying to generate candidate sequences for lab experiments.

Jason Lopez: Advances in large language models as well as computer infrastructure are cutting development times dramatically. Debo says what would have taken years might take a 10th or even a hundredth of the time.

Debo Dutta: I’m not an oncologist. I’m not a gene therapy researcher or an immunotherapy researcher, but I do believe that as a systems person, if I can help the lifecycle of the machine learning go faster, I can help them to make immunotherapy get better and better. We can cure more types of cancers and more diseases.

Jason Lopez: Debo was born in India. His mother is a retired doctor and his father a retired engineer. Those two professions are some of the most sought-after in the country and getting into a good university to pursue those professions is very competitive, and he was able to enroll in one of the top colleges in engineering where he opted to study computer science.

Debo Dutta: But I kept thinking about biology, but I didn’t do much about it. 

Jason Lopez: After graduating, he came to the US to study for a PhD at the University of Southern California, and that’s when he saw another dimension to computer science.

Debo Dutta: It looks like pure science in many ways because it’s a lot of math, applied math and discrete math, and a lot of engineering too. But it can be applied to biology.

Jason Lopez: At USC he encountered a large team doing computational biology. And after his PhD, he had this thought:

Debo Dutta: No matter what I do in life, I need to spend some time in computational biology at USC before I actually get into my career and just be a researcher and learn how computer science can be applied. And that’s when I first picked up what is now known as machine learning.

Jason Lopez: What do you think about the AI boom?

Debo Dutta: Yeah, that’s a very interesting and loaded question. I could spend the entire session just talking about that.

Jason Lopez: Models are at the heart of machine learning. Data is used to train models. Which are deployed to infer from new data. And in the ChatGPT and the generative AI space, models are getting bigger.

Debo Dutta: These models can be as large as 1 trillion parameters. And typically when you see new data, you have to run this data through the model. Big models take longer time, more compute resources. And so measuring the performance will lead to more innovations in the infrastructure space. 

Jason Lopez: That will make models cheaper to develop and cheaper to deploy. But, Debo points out the quality of the model still matters.

Debo Dutta: When you’re generating text, how good is the text? Is it hallucinating? Is this text believable? How was this model developed? Was it developed on open data? Was it developed on somebody’s private data? And if that is the case, what’s the liability of the model if I use that model?

Jason Lopez: When Debo did his postdoc in computational biology at USC, it was called statistical algorithms. He says he was on the right track as he explored signals in biological experiments, especially proteomics: the study of proteins.

Debo Dutta: I got to learn about machine learning as a part of understanding how the human body works.

Jason Lopez: Statistical algorithms were one of his tools at his job, later, at Cisco. He did cloud computing for a while. But then decided to investigate a new area.

Debo Dutta: And what struck me is cancer therapeutics in the US was still not advanced enough by a computer scientist standards.

Jason Lopez: One day, he found himself at a meeting at Stanford with some leading machine learning experts. That group became ML Commons. Subsequently, they began discussing what ML Commons ought to do. It wasn’t long before a medical-focused working group was formed.

Debo Dutta: And that’s when the idea of this med perf paper was born. 

[Related: Alex Karargyris’s Path to Becoming a Pioneer of Medical AI]

Jason Lopez: The MedPerf paper, entitled Federated Benchmarking of Medical Artificial Intelligence with MedPerf, established the vision for the organization.

Debo Dutta: MedPerf was not even a thing. I mean, we just started Med Perf. We got together a bunch of experts from different walks of life, because problems in medicine are hard. They need expertise from all corners. So we consulted with oncologists, computer scientists, data folks, and a lot of people who knew the ins and outs of the pharmaceutical industry.

Jason Lopez: In our conversation with Debo, we did ask him specifically about Chat GPT. And he relayed what Nutanix customers are saying.

Debo Dutta: I love Chat GPT, it can do amazing things and maybe transform my business, but I have to give away my data to Chat GPT. 

[Related: IT Leaders Get AI-Ready and Go]

Jason Lopez: What they want is a Chat GPT equivalent – kind of a GPT-in-a-box running on their own private infrastructure.

Debo Dutta: Nutanix is very customer-focused. So our reaction to our customer’s request is a turnkey solution, which will help our customers run generative AI models, including large language models that fuel applications like Chat GPT, on their private infrastructure.

I joined Nutanix in 2020 during the pandemic. And when I joined Nutanix, I realized that Nutanix had this amazing platform, which could host a lot of these amazing next-gen AI workloads that would change the world in ways that we don’t even know about. For example, new drugs, new antibodies. So I’m really excited by medicine and AI and infrastructure and what it could do to humankind.

Jason Lopez: Debo Dutta is vice president of engineering at Nutanix and a founding member of ML Commons. This is the Tech Barometer podcast, I’m Jason Lopez. If you like this report and want to know more about MedPerf, we’ve got a couple of other stories on the group, one focusing on the paper and another on MedPerf co-founder and the paper’s author, Alex Karargyris. You can find more at the Forecast, which produces this podcast. It’s at www.theforecastbynutanix.com.

In this Tech Barometer podcast segment, Greg Diamos tells how an early passion for computing led him to a pioneering role Baidu’s Silicon Valley AI lab, where he discovered ways to scale deep learning systems, and went on to co-found MLCommons and AI company Lamini.

Find more enterprise cloud news, features stories and profiles at The Forecast.

Transcript:

Greg Diamos: I think computing is a way to give people magical abilities, like superpowers. I like to live in a world where you can give it to everyone. Like, what if every single person had superpowers? I had basically ten years with nothing to do and just a pile of computers in my garage.

Jason Johnson: Just getting the ball rolling, I was wondering, Greg, could you tell me a little bit about what it is you do?

Greg Diamos: I’m one of the people who builds the software infrastructure that’s needed to run applications like ChatGPT or Copilot. Let’s see, I think it’s helpful to start from the beginning. My last name is Diamos. It’s a Greek name. It’s kind of cut in half. About a hundred years ago, my family immigrated to the U.S. They mostly settled in the West.

Before I was born, my father, he moved to Arizona. And his goal in life was to retire and find someplace relaxing and to go and play golf most of the time. And so he moved to Carefree, Arizona. Most people, when they think of Arizona, they think of Tucson or Sedona or places that normal people would go. You go to the Grand Canyon or something. Carefree is a place where no one ever goes. It’s right in the middle of the desert. There are not a lot of people, but it has a lot of golf courses and it’s very relaxing. It’s like a spa. So it’s a good place for that, but it was a very boring and brutally hot place to be as a 10-year-old boy.

[Related: AI Reorients IT Operations]

So I was bored, bored out of my mind. My mom worked as a database administrator for IBM. IBM at the time, they had a mainframe business. And so they would basically throw away the last generation of stuff. She would just take all the machines that were headed to the dumpster, put them up in the car, drive them out and throw them in our garage because it would be such a waste if such a thing was thrown away. I had basically 10 years with nothing to do and just a pile of computers in my garage. So it took me from when I was 10 years old until when I was about 25 to kind of figure out how the machines worked. It seems, like, magical to me that you could build such a thing. You could do that as a little boy stuck in the desert. I did a PhD in computer engineering. After that, I went to Georgia Tech.

Okay, let me tell the story of this one. I really love telling the story. I actually worked at the Baidu search engine. You could think of it as the Google of China. If you go to the search bar, there’s a little microphone on it. You press the microphone, you can talk to it. It was based on the last generation of machine learning technology. It still used machine learning, but it didn’t use deep learning. This exists right now in all major search engines. Around 2014, 2015, Baidu was in the process of deploying their first deep learning system.

[Related: Alex Karargyris’s Path to Becoming a Pioneer of Medical AI]

One of the projects that we had in the Baidu Silicon Valley AI lab was to apply deep learning, which had just been invented and was just starting to work to improve that particular product. Along the way, we had a number of researchers. One of the researchers was Jesse Engel. He was a material scientist, but he was not a deep learning researcher. Actually, none of us were deep learning researchers because deep learning didn’t exist. One of the things that he did is he performed sweeps over all the parameters in the system.

One of the sweeps produced this plot that seemed to show that there was a relationship between the accuracy of the system or the quality of the system with the amount of data that went into the model and the size of the model. If you think of a neural network as just being a collection of a ton of simulated neurons, not real neurons, but simulated neurons in that system, trained on thousands and thousands of hours of recordings of people talking. It seemed like as you added more neurons or simulated neurons, and you added more recordings of people talking, the system got smarter. It didn’t just happen in an arbitrary way. It happened in a very clear relationship you could fit with a physics equation, E equals mc squared kind of equation, a one parameter equation. It was a very simple, very consistent relationship. We thought that was pretty intriguing. After that project succeeded, the team grew a lot.

[Related: IT Leaders Get AI-Ready and Go]

I inherited a pretty big group of researchers, about a 40-person group of researchers. I had to decide, what should we do? I thought that was the most interesting thing that I’d seen, so let’s see if this actually is a real thing. Let’s try to reproduce this. Let’s try to understand it better. We spent about a year trying to reproduce that experiment, and we absolutely could not break it. It was very repeatable across many different applications.

We tried it for image recognition. We tried it for different types of speech recognition, different speech recognition models. We also tried it for language models. Language models at the time were actually used in speech recognition, basically as a spell check. Essentially what came out of that was this very consistent, very repeatable effect that if you increase the size of the model according to a certain ratio, if you increase the amount of data according to a certain ratio, the quality of the system, like its ability, for example, for a language model to predict the future, like predict the next word, would get better in a very predictable way.

[Related: Seeing AI’s Impact on Enterprises]

We published this as a paper. It was called Deep Learning Scaling is Predictable Empirically. This is, I think, came out in 2016. I just thought it was so weird that you could actually predict intelligence. We tried even more to break it. We went and consulted all sorts of machine learning theory experts, and I actually finally understood why it was happening. There actually is a theoretical explanation of why this is happening. The result of all of this is basically we can’t break this thing. This is actually a real thing. This is a real relationship that’s repeatable.

It’s absolutely the most amazing thing I’ve ever seen in my life because if I try and explain what that means at an application level or a user level, it means that we have a repeatable recipe for intelligence. We have a simple equation that you can write down, and then I can just apply this recipe, and I can create intelligence. The implications of that, we understood what they were. I think, in particular, Dario Amodi, who is also a researcher in our group, really got it, what the implication of that was. He was a biologist who had done very detailed computational biology experiments. He took that and went on to put it into GPT-2. They just kept scaling it, so it kept getting smarter. GPT-2 became GPT-3. You had to have an enormous amount of computation to run this thing.

[Related: Will AI Workloads Overload IT?]

Finally, it seemed to cross a threshold around ChatGPT and GPT-4. It’s not just spell check anymore. Now it’s abilities that we think of as being very uniquely human abilities, like the ability to read and write English, the ability to reason logically, the ability to plan, the ability to write software correctly.

It’s not like these are just being produced in a research lab. We’re actually seeing them integrated into products that are deployed to all users. ChatGPT right now has more traffic than Netflix. It’s not just AI researchers who are talking about it anymore, who are playing around with these things. It’s actually pervasively available. If we project into the future, as long as we can keep feeding it with data and keep feeding it with computation, it’s going to keep getting smarter, and we’re going to see new abilities emerge out of it, many other human abilities and potentially even beyond human abilities.

Jason Lopez: Greg Diamos is co-founder of ML Commons, as well as co-founder of Lamini. He’s also a founding member of the Silicon Valley AI Lab, where he was on a team that helped develop deep speech and deep voice systems. This is the Tech Barometer podcast. Tech Barometer is produced by The Forecast, where you can find more tech stories about topics from enterprise software to the cloud revolution, AI, and digital transformation. Check us out at theforecastbynutanix.com. This is Jason Lopez. Thanks for listening.

Recent survey data found that after years of declining rates, well-being incentives are on the rise for employers, with 73% offering either incentives or disincentives in 2023. This may be fueled by growth in Lifestyle Spending Accounts (LSAs).

On this episode of the Business Group on Health podcast, we welcome Uri Gneezy, an expert in behavioral economics, to discuss his research on designing thoughtful incentives that can help individuals and organizations make lasting changes.

Guest: Uri Gneezy, PhD | Behavioral Economist, the University of California, San Diego, and author of Mixed Signals

Digital twins are revolutionizing how we model, simulate, and optimize everything—from product designs to real-world systems and processes across virtually every industry.

For our manufacturing process, we developed Intel Operations Recon to model, simulate, and optimize our manufacturing operations, improve yields accelerate problem resolution and training, increase uptime, and more. And now we’re making Intel Operations Recon available to all kinds of industries that can benefit from it. And the software can be customized to model, simulate,

Intel Operations Recon enables you to see your world in three modes:

Intel Operations Recon benefits also extend beyond efficiency. You can also use it to improve safety and customer satisfaction, reduce costs and risks, contribute to your sustainability initiatives, and more.

In this video, we provide an example of how Intel Operations Recon could be applied to the airline industry.

If are you ready to unlock unprecedented levels of visibility, insights, and efficiency in your operations quickly and easily with Intel Recon, reach out to [email protected] to get started.

In this Tech Barometer podcast segment, explore how a partnership between data networking giant Cisco and hybrid multicloud software company Nutanix is changing perspectives and opening opportunities for people who build and manage IT operations.

Find more enterprise cloud news, features stories and profiles at The Forecast.

Transcript:

Lee Caswell: 750 million new applications in the next three years, more than the past 40 years of computing. How do you assimilate that quickly without overrunning your staff?

Jason Burns: You’re able to roll out these applications and not have to worry about how much time it might take to stand up new servers, stand up the network. All of that piece just can be more seamlessly provisioned so that you have more time to deploy applications.

Lee Caswell: Let’s say it a different way. Infrastructure doesn’t get in the way of supporting new applications. The best partnerships are customer-driven.

Jason Lopez: Those are the voices of Lee Caswell, vice president of Product and Solutions Marketing and Jason Burns, Director of Technical Marketing, both of Nutanix. This is the Tech Barometer Podcast Here, they’re in conversation about the evolving nature of network infrastructure and application deployment in the context of Nutanix’s partnership with Cisco. They chat about different perspectives and understandings that network, storage, and compute teams bring to the table, unified communications and Cisco UCS, rapid application deployment and many other facets of the partnership

[Related:Focus Shifts to Migration in Wake of Broadcom’s VMware Acquisition]

Lee Caswell: Today, for example, I was talking to a customer who’s the first Cisco customer we have in the Americas, not the first we have worldwide. And one of the things that really stuck out for them was the relationship that they have with Cisco, you know, expanding from networking into compute, has really meant that they have a level of trust there and confidence. And now that Cisco’s reselling the Nutanix offer, it’s really bringing together the networking and the, call it the infrastructure teams, in a way that we brought in the past, the storage and the compute teams together. So I’m curious on your perspective on that, Jason.

Jason Burns: From my perspective, I’ve seen the same thing. When customers are deploying a solution from Cisco, it is exactly that. Oftentimes you’ll see something with the unified communications on top of Cisco UCS and the Cisco networking piece as well. And so the combination there of being able to bring in the entire application on top of the computing system and then the storage as well is really appealing to the customers. And then having that delivered as a solution rather than something that they’re piecing together part by part.

Lee Caswell: You know, as I’m watching and talking to some of the Cisco folks, there’s some really interesting things that Nutanix does that they didn’t have access to before and now make it just easier to sell anHCI-based solution. One is, you know, the ability to have mixed nodes, for example, to be able to have nodes with different capabilities, whether it’s disk and flash or different CPU capabilities. That level of flexibility actually makes it a lot easier to sell this. And now integration, you know, support for ACI, for example, really helps from an integrated management solution. So I’m seeing those two things really set the bar as to how we’re now we’re bridging the network stack along with compute and storage.

Jason Burns: Yeah, the integration ofNutanix AHV and Cisco ACI is something that I’m personally excited about. As a former network engineer, I love to see the automation point where we can make networks inside of the network fabric and have that pushed into the compute fabric. What you get from that action is a system where you’re able to really easily deploy applications because the network substrate is already there for them. That’s part of what Cisco has done to integrate with Nutanix AHV. That’s really complemented by what Nutanix is doing on the virtualization side, both with our flow virtual networking and flow network security for virtual networks and micro-segmentation inside the hypervisor itself. So you have this physical network layer that Cisco is providing, and then you have the virtual network layer that Nutanix provides there, and that’s all being tied together.

Lee Caswell: Yeah, it’s kind of fun. I don’t know if you’re familiar with the concept of parallax, which is basically things, lines move as you move. It looks like things are moving because you’re looking at it from a different vantage point. And so in many cases, the network community is looking at things from one vantage point, where you’ve got now the storage and compute teams coming in and looking at it from a different view. For example, things like capacity utilization and performance from a storage standpoint are around IOPS, consistent low latency during failure conditions. Those are the things that a storage person cares about. And at the same time, now you’ve got all this based on the underlying network infrastructure. So the ability to have both of those lenses, I think, is really an interesting way to think about how to deploy this more seamlessly over time.

[Related:Why Hybrid Multicloud Matters]

Jason Burns: The lenses is a great point of it. The network team, rightly so, controls the network because they need to have the best performing network for their applications. They hold some of that responsibility, and therefore they hold the provisioning control as well. Same thing for the compute and storage side. But going back to that comment around lens, you get visibility from the storage side of what the network is doing. And then from the network side, you get visibility into what the compute is doing, thanks to the integration of ACI and AHV. So the solution works a bit better now that these entities are aware of each other.

Lee Caswell: You’ve also got inter-site integration, looking at how you go and view, manage, and monitor the holistic system. And so, I’m particularly excited about this because there’s over 60,000 Cisco resellers, for example, who are familiar with inter-site. And now to be able to say that inter-site can go and monitor Nutanix environments today, manage shortly, is really an exciting way to go and leverage the experience and expertise of that Cisco reseller base.

Jason Burns: The inter-site piece is also very interesting because I came from a world of deploying Cisco UCS with local UCS managers. So for me, I’m learning more about Cisco through this and learning that they have inter-site as a tool. It’s a tool that I wish I had when I was deploying UCS servers years ago. The ability to look at a fleet of servers and manage them all from that point would have been really helpful to me as an administrator back in the day.

[Related:The Spirit and Hybrid Multicloud Mindset of a Healthcare CIO]

Lee Caswell: Yeah, I was looking at some of the roadmap items we’ve been talking about. One of the things that a lot of customers have been looking for and are excited about what is coming with Nutanix and Cisco together is the ability to run on the latest blade servers that are storage-rich to be able to take that new form factor. Because certainly Cisco is known in the UCS world as having performance-oriented systems with a lot of memory access. That was the early design of UCS. And now the ability to bring storage-rich capabilities and Nutanix’s value on our HCI element makes it possible now to go and run the highest performance applications on this new joint offering.

Jason Burns: Yeah, absolutely.  And I’m looking forward to seeing expansion into that blade ecosystem because that’s what I had seen a lot of deployed myself, these high-performance applications running on a blade backed by some sort of storage that might have resided outside the blade. And so looking at what we have as possible future options with Nutanix compute-only accessing the storage remotely, that to me just seems like a win-win for configuration in these blade server deployments.

Lee Caswell: Especially important, by the way, for the highest performance applications because of the licensing model, it turns out. So the ability to have a contained, let’s call it compute-only node, for example, where you can now go and have, let’s say Oracle licensing to pick one running dedicated on that node means that now you’ve got a predictable licensing expense going forward. And so that’s important because, you know, there was a risk in the early days, you could basically have licenses spread across all of the CPUs and the cores. And of course that would be prohibitively expensive. So the idea that you can contain licensing to this actually means now you can take the highest performance applications that are licensed by core and make sure that financial exposure is guarded there.

Jason Burns: Anecdotally, one of our customers who’s a service provider, they deploy a lot of different solutions. I was helping them out with a technical design I said, you know, this is how I would do it. And he said, well, Jason, about 40% of the way I design things is technical and the other 60% is driven by the licensing requirements of all the solutions I’m deploying. One solution, while I might from a technical side think it’s the right answer for licensing or financial reasons, might not be a viable solution for them to deliver that service.

[Related:Ransomware Resilience: Evolving Data Protection]

Lee Caswell: Super interesting. Even as you look at what’s next after databases right now, the number one workload on Nutanix. Analytics now from our own survey data is 49% of our customers are running analytics. And a big part of that, of course, is Splunk. So now acquired by Cisco, I see a pretty interesting opportunity to go bring together these high-performance analytics along with our Cisco partnership. Bring that all together and make that really a solution from a customer standpoint.

Jason  Burns: Yeah. I’m also interested to get hands on that in the lab myself. Part of what we do is we, in the tech marketing team, we develop these lab environments for people and we try to demonstrate our own software and try to pull in realistic use cases where we can. One of those use cases is you have this Nutanix software stack. It’s generating alert data, event data. How can you send that into a tool that’s very similar to what the customer has? A lot of our customers ask us, how do I send all this data into Splunk? If I were able to, in the future, have a system that can send all that data into Splunk and then show our customers how to do that, to me, that seems like a great offering and something that customers would benefit from. So I think that’s how they deploy it in the real world.

Lee Caswell: Of course, no conversation would be complete without a discussion of generative AI. Thinking for a moment about how we’re taking GPU-enabled Cisco nodes and now being able to do our GPT in a box, for example, together with Cisco UCS. Pretty interesting, right? As a way to go and say, how can I get started? You could be training a model, probably not a large language model, which is happening in the public cloud, say, but you could certainly be modifying or tuning a model with your Cisco-enabled, GPU-enabled nodes. And then being able to take some of the value that we offer on the software side to help customers get started on their AI journey.

Jason Burns: Yeah, Lee, I see people ask that all the time, which is, how do I get started? How do I deploy this? What’s your best practice guide? And working with Cisco, what they have on their side is the Cisco validated designs. And so what I would love to see is a Cisco validated design for something like generative AI, kind of complementary to our GPT in a box solution for, okay, here is how you get from end to end a deployment with Cisco UCS servers, the Nutanix AHV hypervisor, your GPUs of choice, and how do you build that out from end to end? For me, the validated design is a great starting point because it shows customers that we’ve tested this out in the lab. And if you follow this recipe, you also can have success.

[Related:Will AI Workloads Overload IT?]

Lee Caswell: I guess it strips out both the real risk by doing predetermined testing and then also any perceived risk. So you can say, all right, if I can take the infrastructure piece out, which by the way, is what every data scientist wants. It’s like, how do I make sure the infrastructure is there responsive and fast? So we do that. And by the way, reliable and day-to operations and compliance, all the things that IT brings to the party. And then at the same time, be able to deploy new applications really quickly, whether they’re libraries like PyTorch, for example, or bringing in new applications, bringing in our support for the NVIDIA and VAI libraries. I mean, these are just ways to say, hey, we’ve got all the building blocks and can help you with putting them together.

Jason Burns: And having the validated design or whether that’s a Nutanix validated design or a Cisco validated design means that you’re really speeding up the deployment. You’re cutting out a lot of those investigatory steps and just following along a cookbook rather than inventing the wheel on your own.

Lee Caswell: I think for a lot of customers, particularly, they’re looking now at these new server-based architectures as a way to build out an architecture for the next three to five years, where you’d put all of your new applications. It’s rarely just like rip and replace, take something out. That’s why we still have Unix systems in there. It’s additive in the sense that now you’re going to take these server-based architectures, add that in, but put all the new workloads on there. Certainly that’s what happened with x86 because the economics and the simplicity of moving to a x86 architecture based on Ethernet, which is what we’re doing here. And then deprecating over time, fiber channel, SANs, switches, and even the expertise. So you can basically focus on the new stuff is how I’m viewing this market.

Jason Burns: As a practitioner, what I like is that that gives me the time to focus on the new things. If you’d asked me a few years ago if I was going to be learning about generative AI models, I would have said, no, that’s not really something. But now that the infrastructure side is more taken care of, that’s a place where I can look at these new applications and maybe learn more about how Kubernetes works and how to deploy a modern app in Kubernetes or maybe learn more about an AI application rather than spending all my time just getting the compute and storage and networking infrastructure up and running.

Lee Caswell: So this infrastructure as code transforms into everything as code. If you think about over time, how do we make this easier for you to take and basically treat AI as just the next workload? Many customers from Nutanix started off inVDI, then went todatabases, then went to analytics, and now are looking at AI, generative AI, as the next thing. And if you can assimilate that easily with the same team instead of having a separate architecture, boy, it just makes it a lot easier to get some operational leverage, particularly when you’re having trouble finding talent.

Jason Burns: That is something that I see every day, even in some of the new designs that I’m rolling out. You might do something where you have an engineer walk through and manually set something up as a proof of concept. That’s pretty time-consuming to do. Now, if you have something like infrastructure as code or your application deployment defined as code, the subsequent deployment of that can happen really fast. And so you just cut out a lot of time there for deploying that.  That really is moving toward everything as code with the solution we’re showing here. So you have the Cisco ACI portion, where you have these APIs for configuring your network. You have the Nutanix compute and storage part, again, with lots of open APIs to configure. And then these modern applications on top of it, which might be deployed as a series of config files against aKubernetes cluster. You can just keep all of that code for the entire infrastructure and even application configuration in a code repository where someone can just check that in to make changes to it.

Lee Caswell: Yeah, it’s always struck me that ACI was probably one of the best acronyms in the industry, application-centric infrastructure. The idea that you’re thinking, hey, how do I make sure that applications are optimally located? And there’s a really interesting way to think of how we introduce value jointly here, because when you bring these together, these technologies, now all of a sudden you’ve got a full complement of how to make a seamless outcome for optimally locating applications and their associated data over the hybrid multi-cloud universe.

Jason Burns: And as I was seeing adoption of modern applications and containerized applications, that was the one thing that struck me the most, is that all of these apps are going to generate and use incredible amounts of data, and they’re going to need highly available storage that’s also high performance. I didn’t see on the market a dearth of those solutions. I saw that Nutanix really had a great offering in that space, that people were going to be deploying these modern apps and using Nutanix storage to back that app deployment.

Lee Caswell: Both of our companies, I think, have terrific brand loyalty, by the way, in part because of the support experience that customers have gotten. And one of the things that impressed me about the relationship that’s more than just your average partnership is how we’ve integrated the support path so that our support systems and support tickets can be follow-up and make sure that the customer experience is singular in nature. Curious on your thoughts on how you’ve seen that develop, and why is that important?

[Related:How Flash Memory is Reinventing Hybrid Cloud Storage]

Jason Burns: Working at Nutanix these last nine years, I’ve heard consistently again and again that customers value Nutanix support. And in my previous life, I was actually a Cisco support engineer long, long ago, and heard the same thing, that customers valued the support that they got from Cisco. So we have a history of two great support organizations, and being able to support our customers really is what I know that they’re looking for. It keeps them coming back. It keeps them expanding solutions on Nutanix or expanding their solutions on Cisco, because they know that when something goes wrong, they can give us a call and get routed to someone who can help them.

Lee Caswell: I have a theory on why our companies are so well suited together. So my theory is this, that architecturally, from a design point standpoint, network engineers tend to make things fast first and then reliable later, because you can always retransmit. And storage companies come at problems with exactly the opposite lens, which is that they try and make things reliable first, and only then try and make them fast. And you can certainly see that through the enhancements we’ve made over the latest versions of AOS, for example, and moving into newer flash technologies, you know, NVMe and new CPUs and everything. So I’m curious to say, how do you think about it, right, architecturally, isn’t that an interesting way to think about why the engineering teams have so much appreciation for each other?

Jason Burns: That definitely rings true to me. And I’ve seen that growth at Nutanix over time, where all of our subsequent software releases have basically worked on performance over all others, but reliability was always there, because from a storage perspective, you had to keep that data. Data corruption wasn’t really a possibility for consideration, but we really worked hard on increasing performance over time. And it seems like we’ve kind of met in the middle here between Cisco and Nutanix.

Lee Caswell: I think storage people, right, preserving the data is number one and all the data services that we have. And I think that’s a really interesting value also, because for most customers now, this is a really unpredictable environment. Trying to forecast, for example, your mix of blocks, files, objects, data, trying to forecast whether you have VMs or containers three years from now, and start to think about, you know, what you have on-premises versus the cloud. The simplifying design that Nutanix brings is the fact that you’ve got the same data services across all of those endpoints. If you go and buy different unique architectures for each, you’re going to have a different way to go and protect that data, different snapshots, different recovery mechanisms, different disaster recovery. And so it’s a really easy simplifying, you know, way to basically make this easier for customers to get operational leverage and deploy things in an unpredictable environment.

Jason Burns: Nutanix has a lot of that flexibility on the data services side, and I see Cisco having that flexibility on the compute side. The longevity of some of these Cisco UCS chassis, for example, or how long you might have a UCS deployment is pretty incredible. And that’s just something that you can keep as a long-running deployment and take out parts of it, maybe blades, maybe certain rack servers while keeping the whole of the deployment there. You get the ability to upgrade in place and really extend the possible lifetime of a deployment. And for me, that was always the beauty of the Nutanix cluster, because your AOS storage cluster could live for quite a long time, even across multiple generations of Nutanix hardware.  If you are adding newer-generation servers and removing older-generation servers, you have a pretty long-lived deployment without a lift and shift.

Lee Caswell: So one way to think of this is in a highly unpredictable environment where applications are changing really fast. You don’t know necessarily if there are going to be block interfaces or file interfaces or object interfaces. By being able to buy one thing, you can actually get started. And then by having seamless licensing that allows you to go and basically deploy at will over time, you’ve got this flexibility now, along with the fact that we’re the only company that gives you a consistent way to do data protection across all three of those protocols. And by the way, across VMs, traditional VMs, and now newer container systems orchestrated by Kubernetes. So it’s a really fantastic way, I think, to go and give that flexibility for customers who actually don’t want to be storage experts, it turns out.

Jason Burns: That would describe me as well, a former network engineer who has come to the storage world, but Nutanix has made that a lot easier. Storage is not as intimidating as I had thought it was after working with Nutanix. Let me say it this way. You know, what I’m hearing from a lot of CIOs now is that they’re interested in consolidating architectures so that they can address the fact that they don’t, let me say that in a different way. Yeah, one of the things I’m hearing from CIOs recently is about trying to consolidate the number of different architectures that they have. So for example, a simple one would be like separate filers and separate SAMs, for example, bringing those together, and a different object store, bringing that all together. And so for CIOs thinking about how they do it, they’ve already done that with Cisco on the networking front. They have a single vendor on the network front that gives them everything across different architectures. And so now one of the opportunities here is by working with Cisco and Nutanix together, now you’ve got the opportunity to basically address any of the degrees of freedom you’re likely to anticipate in the future.

Jason Lopez: Lee Caswell is vice president of Product and Solutions Marketing at Nutanix. Jason Burns is director of Technical Marketing at Nutanix. This is the Tech Barometer Podcast, I’m Jason Lopez. Tech Barometer is produced by the Forecast. We have a treasure trove of tech stories at the forecast from all across the technology landscape. You can find more at:

In response to growing concerns around hospital safety and the impact on patient experience, The Leapfrog Group was founded to establish core safety standards and grading mechanisms for hospitals and empower employers to make data-informed purchasing decisions.

On this episode of the Business Group on Health podcast, Leah Binder joins us to discuss the work The Leapfrog Group is doing to promote transparency around hospital quality and safety and drive a culture of improvement.  Leah shares how this data can be used by hospitals to improve their processes, and by employers and consumers alike to make informed choices.

Thank you to today’s episode sponsor ComPsych.

Guest: Leah Binder | President and CEO, The Leapfrog Group

Psychedelic-assisted therapy is a growing avenue for addressing mental health conditions such as treatment-resistant depression, and more progress is expected in the coming years. On today’s episode of the Business Group on Health podcast, we’re joined by Andrew Penn, a leader in the study and practice of psychedelic-assisted psychotherapy.

Penn discusses the latest developments in the rapidly evolving world of psychopharmacology, including a new drug application submitted to the Food and Drug Administration in late 2023, designed to treat post-traumatic stress disorder. Penn describes the patient experience during psychedelic-assisted therapy, the integral role of nurses in current and future care models, and projections for future treatment innovation.

Thank you to today’s episode sponsor ComPsych.

Guest: Andrew Penn | Psychiatric Nurse Practitioner and Clinical Professor, University of California, San Francisco

Intel IT is using computer vision and AI for inline inspection in Intel’s assembly and test factories.

Keeping Intel’s global network of factories running at top efficiency is one of Intel IT’s primary roles. Over the last decade, we have introduced numerous automated, artificial intelligence (AI)-based solutions that significantly reduce waste and improve product quality. Recently, we added a new capability using computer vision (CV) and machine learning that enables us to perform inline inspection during assembly and test processes. Inline inspection can catch defects and tool issues much earlier than offline inspection. Additionally, our solution, called Intelligent Wafer Vision Inspection (IWVI), can easily detect micrometer-sized defects that are hard to see even with a microscope, as well as defects that are impossible to detect after the process is complete.

High-resolution cameras take multiple images per second while a grinding tool thins the wafer, and a protective polyester film is installed. The images are analyzed by a machine-learning model at the edge. If defects are detected, the solution can sound an alarm or even stop the tool.

The solution, which includes Intel® Core™ i9 processors, Intel® Xeon® Scalable processors and Intel ARC A770 discrete GPUs, is being deployed to all Intel assembly and test factories.

The IWVI solution has been identified as a critical solution to help ensure the quality and reliability of new generations of products in Intel’s factories of the future. Its standardized design results in fast deployment and easy extendibility to new use cases. The PoC demonstrated that the solution could accurately identify many types of defects in the wafer-thinning process, including indentations, scratches of various sizes, grinding marks, stains, cracks, bubbles, wafer shift and mount shift.

The PoC demonstrated that up to 50% of wafer-thinning issues can be detected earlier with inline inspection compared to offline inspection. When the IWVI alarm sounds, the wafer can be reworked, which avoids whole-wafer delamination during downstream processes. More importantly, the solution goes beyond what was possible with offline metrology:

Using the IWVI solution, we achieve several business benefits:

We anticipate that the solution will eventually extend to additional processes, and we are excited to share our success with other manufacturers, who may be able to adapt the solution to additional use cases.

In this Tech Barometer podcast segment, IT industry analyst Steve McDowell and Lee Caswell from Nutanix discuss the challenges and risks IT operations leaders face in the wake of Broadcom’s acquisition of VMware.

Find more enterprise cloud news, features stories and profiles at The Forecast.

Transcript:

Steve McDowell: Broadcom announced quite a while ago that they’re acquiring VMware.

Lee Caswell: It’s not a light switch moment. 

Steve McDowell: Because of regulatory hurdles and all sorts of other things, it stretched out for multiple quarters. That is where I think the initial wave of concern came in, because now not only do I have to wait and see, I have to wait multiple quarters to see. 

Lee Caswell: Very rarely in IT do you just decommission something now instead, what you’re thinking is, all right, how do I make sure that I’m containing the risk, and then taking all the new things and putting that on the new modern infrastructure.

Jason Lopez: In the ever-evolving landscape of IT infrastructure, companies are increasingly finding themselves at a crossroads, particularly in the wake of significant industry acquisitions such as Broadcom’s purchase of VMware. This shift has ushered in concerns for VMWare’s customers. 

[Related: Hedging Into Expected Broadcom Acquisition of VMware]

Lee Caswell: What happens if they’re not one of the top 2,000 customers?

Jason Lopez: Lee Caswell, Senior VP of Product and Solutions Marketing at Nutanix, says VMWare customers are worried about three major things: pricing going forward, and we’ll dive deeper into that in a few minutes, as well as support for products and the risk to personal careers.

Lee Caswell: Those customers are worried from a support standpoint. I think those customers are legitimately worried that they’re not going to get a call. Partners are worried, too, because Broadcom doesn’t have a great relationship with partners. So I think that support piece will come to be the thing that really actually unhinges them. And the third one is about personal career risk. If you can’t get support and you’re running something and you didn’t make a change, then all of a sudden it’s going to be back on you.

Steve McDowell: There’s a lot of uncertainty among the IT community about what Broadcom is going to do with VMware.

Jason Lopez: Steve McDowell is principal analyst at NAND Research.

Steve McDowell: Broadcom historically, they’ve largely acquired companies that have largely commoditized and they maintain them forward, but you don’t see a lot of continuing innovation. Are they just going to take the core VMware business and maintain that moving forward? Because the reality is a technology like VMware, or VM in particular, is very sticky, right? It’s hard to kind of rip and replace. 

[Related: Broadcom’s Acquisition of VMware Stirs Uncertainty]

Lee Caswell: The way that customers are talking to me about this right now is that what they had with VMware for the past, you know, almost 20 years, right, was a very safe offering with known risks. And now what’s happened is what was the safe bet is now the risky bet. And it’s got unpredictable risks.

Steve McDowell: You know, if you have a VMware license, you’re probably not going anywhere for the life of that infrastructure, right? But where the uncertainty lives is on, you know, when I do come up to a replacement cycle or I have a greenfield project, you know, is VMware going to continue to deliver the level of innovation?

Jason Lopez: This has left many customers contemplating a strategic approach where they maintain their existing VMware infrastructure but refrain from expanding. Instead, they’re considering investing in new applications and technologies on alternative platforms based on newer architectures.

Steve McDowell: From Broadcom’s perspective, I think everything they’re doing makes a lot of sense for their investors and for maybe existing VMware customers.  But where there’s, I think, going to continue to be a concern: how does the VMware technology continue to evolve, right? Am I still going to get the level of support I had? And am I still going to be paying the same amount of money? There’s also news out of Broadcom earnings, they’re re-examining kind of their subscription strategy, subscription bundle, and subscription pricing.

[Related: IT Leaders Get AI-Ready and Go]

Lee Caswell: There was a blog published by the leader of all the infrastructure products, saying some very interesting things that people had been expecting, but now they’re in print. So number one was it’s the end of any new perpetual licenses forever. That was interesting because that’s the way that most VMware customers have transacted businesses in the past. Also when those perpetual licenses came up for support and services, you’d have an SNS or support and services element, and there also no new SNS contracts as well. So those two things basically said that, hey, every customer is moving to a subscription model. Now, what you know in your personal life, right, is also true in business. When you move to a subscription model you end up paying more.

Jason Lopez: Caswell cautions that subscriptions, strictly speaking, don’t automatically equate to higher pricing. Subscriptions work when a customer maps out the long-term costs and benefits. But other factors within a subscription matter, such as bundling.

Lee Caswell: Subscription is one way where you would pay more, even if the prices were the same.   But now what we’re looking at is that the bundling of products is starting to look like you may have to go and move up into higher level products where the way it’s being described is you’ll get more value. And that could be true, but the question is, it may be value that you wouldn’t have ordinarily bought. Think of it like when you have a cable TV subscription and you get 800 channels and you’re like, okay, I’m gonna watch like 20 of them. Can you give me the 20? And they say, no. And so now you’re stuck with a whole lot of stuff that you may not have originally wanted. 

Jason Lopez: The immediate aftermath of the acquisition saw restructuring within VMware, in areas such as partner teams, engineering, and customer relations.

Steve McDowell: Five days after the acquisition closed, the week after Thanksgiving, they laid off 2,500, 3,500 people, I mean, a significant number of VMware employees. And these were in the partner teams, these were in engineering teams, these were in kind of customer relation teams. Not a good signal if you’re trying to build credibility with your customer base.

Lee Caswell: Publicly they just announced they laid off over 300 people in Cork, Ireland, which is the support center. They’re also closing countries that don’t have a sufficient number of customers.

Steve McDowell: First week of December, we saw them announce that they’re spinning out their desktop compute pieces.  So they’re already starting to shed some of the things that are not core VMware capabilities.

[Related: How to Secure Modern Apps and Databases for Hybrid Multicloud Operations]

 Jason Lopez: Today, one of the driving forces behind the way companies use and pay for enterprise software is an anticipation of a massive influx of new applications over the next few years, far exceeding the growth seen in the past four decades.

Lee Caswell: 750 million new applications coming online over the next three years, more than the past 40 years of computing. So you’re going to have this new wave of applications. And guess what? It’s all unpredictable. If you ask people what they’re going to do three years from now, they can’t tell you which applications they are. They don’t know whether they’re going to be interfacing with blocks, files, or objects, VMs, containers, on-prem, in the public cloud. And so risk right now with VMware is that you’re stalled like a deer in the headlights and you don’t know how to move forward because you don’t know what the risks are going to be over time.

Steve McDowell: Those are the kinds of things that when they’re in the air are going to lead to a certain amount of uncertainty. At the end of the day, being in IT is about de-risking your operations, right? Every business today is a digital business. And if I’m putting a piece of software that’s critical to my infrastructure it just makes sense that I’m going to look for alternatives. Not saying I’m not going to choose VMware, but I’m not going to default to VMware. So I’m going to look at Nutanix. If I’m in more in a cloud native mode, I’m going to look at the various Kubernetes things like OpenShift, Portworx, and things like that. 

Steve McDowell: Broadcom’s business model is to take a commodified technology and continue to monetize that over time.

Jason Lopez: Innovation is a cornerstone of success and resilience in the tech industry. McDowell reminds us that it drives competitiveness, customer loyalty, adaptation to market changes, long-term growth, and it influences strategic business decisions.

Steve McDowell: You know, and when you’re talking about a commoditized technology, you’re not talking about a lot of continuing innovation. Now, a lot of the excitement around VMware in recent years was things that were kind of adjacent to their core infrastructure management path, like Carbon Black for endpoint security, for example. Broadcom, two weeks after their acquisition, said “We’re spinning that out.” So they’re kind of validating, I think, some of the initial concerns. That’s not to say that Broadcom’s not going to take care of existing VMware customers. I think they will. They want that revenue stream moving forward. The concern is really around new projects and replacement projects. Is the VMware technology the right technology? And I’m not saying it’s not, I’m saying it’s early to know. So we’re just going to have to wait and see. But, you know, a lot of IT projects can’t wait and see, which is going to force you to de-risk the decision by looking at alternatives.

[Related: Why Hybrid Multicloud Matters]

Jason Lopez: The market has been trying to understand the implications of this acquisition for about a year. And that limited Broadcom’s ability to articulate a strategy. This, sort of, limbo benefited competitors such as Nutanix.

Steve McDowell: Nutanix is not alone. I think everybody who competes directly with VMware in whatever space began benefiting well before the close just because of all of the uncertainty that just continued to linger. Because of where they were in the process,   Broadcom, fair or unfair to them, they couldn’t respond to that, right? I can’t talk about a company I don’t own. So they would talk in generalities and that doesn’t do anything to ease concerns. So I think, you know, we’re a year into trying to grasp what this acquisition means.

Jason Lopez: Steve McDowell is principal analyst and co-founder at NAND Research. In his tech experience, he’s been at IBM, Fujitsu, and AMD among other firms. And he’s been a prominent industry analyst since 2017. Lee Caswell is Senior Vice President of Product and Solutions Marketing at Nutanix. He served two stints at VMWare and in his career worked in many roles such as at NetApp and Fusion-io as well as co-founding Pivot3 in Silicon Valley. This is the Tech Barometer podcast I’m Jason Lopez. Tech Barometer is a production of The Forecast. You can find more tech stories at theforecastbynutanix.com.

Learn the latest in Intel IT’s approach to fault detection for factory equipment.

For many years, Intel IT has utilized a standard infrastructure with Industrial Internet of Things (IIoT) sensors and Intel IoT Gateways for various applications on the factory floor. Simultaneously, we have employed a separate and expensive vendor-based system for fault detection and classification (FDC) on factory equipment. This system has contributed to reduced factory downtime and enhanced predictive maintenance on certain subfab equipment, such as pumps, gas abatement systems, chillers, and more.

To lower costs, increase quality and accelerate necessary changes in the factory, we explored displacing the vendor’s system with our standardized IoT infrastructure. We also added new, innovative command-and-control capabilities. As we connected tools to our IIoT infrastructure, our system proved to be 10x less expensive than the vendor’s solution. It is also more flexible and can communicate with open standards and proprietary protocols.

The “brains” of the system are a collection of Python drivers—written by software developers in any Intel group who wants to contribute to the code—that control both the data collection and the command-and-control capabilities. This open-source solution has zero license costs and democratizes the solution; if another machine needs closed-loop control, an engineer can simply add a new driver to the library. The solution consists of the data collection plane and the control plane:

Having demonstrated lower costs and improved product quality, we are now scaling our IIoT-based FDC solution to over a thousand subfab tools. We estimate the solution’s value as high as USD 16 million for just a single factory that contains hundreds of subfab entities that support the wafer-etching process.