Modernizing your infrastructure and making room for AI is MUCH easier than you think with AMD EPYC.
When it comes to virtual machine migration, it can be difficult to discern what is accurate and what is just a myth. AMD's own Robert Hormuth, uncovers the misconceptions around VM migration and why it's increasingly essential for companies. https://bit.ly/4aUDCQW
The new Dell AI/GenAI ready data platfom is out! 🎉 📣 PowerScale’s new NVMe all-flash line up - with GPU direct, non-disruptive scaling, multi-tenant capabilities, universal data access with multi-protocol support, federal-grade security and seamless interoperability with flexible public clouds. All in a 1U node supporting 30TB QLC SED drives! Changing the game for high-speed storage and enabling the most demanding AI & GenAI file workloads. "The true challenge lies not in reaching the peak, but in remaining at the top" #GartnerLeader #DellTechnologies #UnstructuredData #GenAIrevolution
New Blog Post: Next-gen AI Workloads Require Next-gen Storage, Introducing the Dell Technologies PowerScale F210 and F710 platform
Global Chief Architect Leader, Field CTO Organization | Speaker | Corporate Mentor | IBM Quantum Senior Ambassador | Member, Board of Directors, | Leader of Communities, CAF | 19,000+ Connections
So—what about memory? Can you software-define something as core to computing as memory, which requires extremely high-speed communication and needs to be transparent to whatever operating system is accessing it? Sure, storage can be virtualized and software-defined—it works in the realm of milliseconds. But memory reads and writes are linked to tasks within a CPU measured in nanoseconds. "Previously, people thought about virtualizing memory the way they thought about the four-minute mile [in track and field]," explained Kove CEO John Overton. "People thought you just couldn't do it, so they didn't try." With this new class of software-defined infrastructure, it is possible to dynamically allocate #memory to support #physical, #virtual, #clustered, and #containerized computing, all while reducing overall #hardware and #energy costs. I remember having a dialog on this a few years back with Overton. Click below to read more. https://lnkd.in/gwWbH7D3 Kelly Switt
Very interesting to be able to do that. A pool of memory available that can be shared to scale horizontally or vertically to meet just in time computing needs and released afterwards. This is different then compute nodes available… see the article for more details.
Global Chief Architect Leader, Field CTO Organization | Speaker | Corporate Mentor | IBM Quantum Senior Ambassador | Member, Board of Directors, | Leader of Communities, CAF | 19,000+ Connections
So—what about memory? Can you software-define something as core to computing as memory, which requires extremely high-speed communication and needs to be transparent to whatever operating system is accessing it? Sure, storage can be virtualized and software-defined—it works in the realm of milliseconds. But memory reads and writes are linked to tasks within a CPU measured in nanoseconds. "Previously, people thought about virtualizing memory the way they thought about the four-minute mile [in track and field]," explained Kove CEO John Overton. "People thought you just couldn't do it, so they didn't try." With this new class of software-defined infrastructure, it is possible to dynamically allocate #memory to support #physical, #virtual, #clustered, and #containerized computing, all while reducing overall #hardware and #energy costs. I remember having a dialog on this a few years back with Overton. Click below to read more. https://lnkd.in/gwWbH7D3 Kelly Switt
FOLLOW ME for breaking tech news & content • helping usher in tech 2.0 • at AMD for a reason w/ purpose • LinkedIn persona •
Most of the new construction that is going on is already allocated, which is hurting a lot of enterprises as they plan for the future. How do you bring AI into your business if you have no more power, space, cooling? Server consolidation could free up much-needed power for AI workloads. Check out the detail here, with input from Dell'Oro Group's Baron Fung. #innovation #ai #AMD #technology Link: https://lnkd.in/emi9f5uz
At Swift we are happy to collaborate with Kove and Red Hat to power our cutting edge #ai ML platform. Though we’re just at the beginning of this exciting journey, we’re already making significant strides forward. Software Defined Memory (SDM) significantly enhances high-performance computing for machine learning models by offering flexible, scalable memory, and unlimited memory resources. It allows ML models, especially large ones, to efficiently utilize memory on servers across the datacenter, improving computational speed and resource allocation. It can use far less energy too. This adaptability is crucial for training and running complex models, as it optimizes performance and reduces bottlenecks related to memory constraints. Click below to read more. https://lnkd.in/gwWbH7D3 #highperformancecomputing #breakthrough #innovation
Global Chief Architect Leader, Field CTO Organization | Speaker | Corporate Mentor | IBM Quantum Senior Ambassador | Member, Board of Directors, | Leader of Communities, CAF | 19,000+ Connections
So—what about memory? Can you software-define something as core to computing as memory, which requires extremely high-speed communication and needs to be transparent to whatever operating system is accessing it? Sure, storage can be virtualized and software-defined—it works in the realm of milliseconds. But memory reads and writes are linked to tasks within a CPU measured in nanoseconds. "Previously, people thought about virtualizing memory the way they thought about the four-minute mile [in track and field]," explained Kove CEO John Overton. "People thought you just couldn't do it, so they didn't try." With this new class of software-defined infrastructure, it is possible to dynamically allocate #memory to support #physical, #virtual, #clustered, and #containerized computing, all while reducing overall #hardware and #energy costs. I remember having a dialog on this a few years back with Overton. Click below to read more. https://lnkd.in/gwWbH7D3 Kelly Switt
Next Generation Storage : Powerscale
Chairman and CEO @ Dell Technologies | Hardware, Software, Strategy
A GPU from NVIDIA will often sit idle if the storage can’t feed it the data fast enough. This is why we created Dell Technologies PowerScale, the fastest AI storage in the world. 🚀
Never a better time to learn about the importance of game changing scale out storage for AI workloads, keeping GPUs busy and delivering innovation. Dell Technologies #PowerScale is a game changer for AI and GenAI workloads. Exciting times ahead. #PowerScale #AI #GenAI #Innovation
Chairman and CEO @ Dell Technologies | Hardware, Software, Strategy
A GPU from NVIDIA will often sit idle if the storage can’t feed it the data fast enough. This is why we created Dell Technologies PowerScale, the fastest AI storage in the world. 🚀
Introducing #Lenovo scalable infrastructure solutions. From servers running the latest Intel® Xeon® and AMD EPYC processors to Ethernet networking solutions, Lenovo offers the tech needed to empower the modern data center. Learn more in this document, courtesy of Integral & Open Systems,Inc.
As AI in the Enterprise continue to evolve, other High-Performance architectures are coming into the spotlight to support those workloads. Dedicated networks like InfiniBand have been around for a while to interconnect compute and storage systems to provide direct access to each other's hardware to overcome the latency challenges of the traditional network stack via RDMA. Now with 400/800GbE in the market, RoCEv2, or RDMA over Converged Ethernet v2, delivers the best of both worlds in a single fabric with minimal compromise. World Wide Technology is testing these architectures to their limits for our customers in our AI Proving Ground to validate practical applications for our customers. Are you ready?
So honored and proud to be on the Supermicro team as it launches its new X13 BigTwin platform. To me, this product delivery shows: INNOVATION : Supermicro continues to deliver better and better products for data center infrastructure * X13 Big Twin platform is a new VMware vSAN solution optimized for enterprise-class hyperconverged virtualized workloads (it is powered by the latest 4th gen Intel Xeon Scalable processors and supports the latest PCle 5.0 NVMe drives) * Comparing with its previous generation product, this new version delivers up to 4.7 X performance and 3 x lower cost VISION: Supermicro pioneers green technology * X13 Big Twin platform servers are super energy efficient * Free air and liquid cooling make a big difference to lower the heat and energy intake of data centers, critical in running AI amidst rising energy cost CUSTOMER SUPPORT: * Building block architect allows for mass customization * This new vSAN solution meets the growing demand for hyperconverged virtualized workloads * Unprecedented performance matters for a wide range of workloads, including virtual desk top infrastructure (VDI), big data analytics, and artificial intelligence (AI) * With modular design, GPUs can be added at PCle Slots for AI WINNING IN AI AND SUPERMICRO WOULD CONTINUE TO GROW * As one analyst (Duberstein) put it, Supermicro “is winning in AI because they have super energy efficient servers and also their building block architecture allows for mass customization and lower costs which is different from other servers and extended models and they are very quick with their time to market” * I think a bigger win is on its way with the new X13 Big Twin platform. Link to Press Release on product announcement https://lnkd.in/ggtHT3_f
Interesting point! How did you prioritize your infrastructure needs before implementing AMD EPYC for AI?