Unlocking the Potential of Deep Learning with Virtualized AI

Run:AI | April 22, 2020

  • A virtualized AI infrastructure for DL would run a single workload on multiple shared physical resources.

  • Article by Google’s Francois Chollet discusses the skill acquisition-based approach to gathering intelligence.

  • The conventional computing stack – from processor to firmware to virtualization, abstraction, orchestration and operating layers through to end-user software.

Is the recent progress in deep learning true artificial intelligence? A widely-discussed article by Google’s Francois Chollet discusses the skill acquisition-based approach to gathering intelligence – the one currently in use in modern DL. He argues that with huge data sets available for training models, AI is mastering skill-acquisition but not necessarily the “scope, generalization difficulty, priors, and experience” that true AI should incorporate. Even with our progress in AI, and specifically DL, we are nowhere near the limits of what DL can achieve with bigger, better-trained more accurate models, those that take into account not only skill but experience, and generalization of that experience.

Understandably, this has put intense focus on computing power, particularly the hardware that enables data scientists to run complex training experiments.


Nvidia increasing sees DL as a key market for its GPUs and bought Mellanox to speed communication inside a GPU cluster. With its recent acquisition of Habana, Intel is likely betting that custom AI accelerator hardware is a better match.

Other AI-first hardware includes Cerebras’s massive chip in a custom box that’s designed for the specific types of intensive, long-running workloads that training DL models require. In the cloud, Google’s Tensor Processing Units offer another bespoke option.

For companies running their own DL workloads, more compute is generally better. Whether exotic AI accelerators or tried-and-tested GPUs, quicker model training means more iterations, faster innovation and reduced time-to-market. It may even mean we can achieve “strong” AI (i.e., AI than goes beyond “narrow AI,” which is the capability of doing a single, discrete task) quicker.


In 2020, continuing the trend of recent years, companies will invest in ever-more AI hardware, in an effort to satisfy data scientists’ demands for compute to run bigger models to solve more complex business problems. But hardware isn’t the whole picture.

The conventional computing stack – from processor to firmware to virtualization, abstraction, orchestration and operating layers through to end-user software – was designed for traditional workloads, prioritizing high-availability, short-duration operations.Training a DL model, though, is the opposite of this sort of workload. While running a model, an experiment may need 100 percent of all the computing power of one or multiple processors for hours or even days at a time.

Read more: Virtualization can transform your company’s IT infrastructure


10 most essential performance metrics that businesses should consider when managing their virtual machines.


10 most essential performance metrics that businesses should consider when managing their virtual machines.

Related News


CBC Tech Partners with Zenlayer to Extend its Global eNet SASE Connect to Global Emerging Markets

Businesswire | May 16, 2023

CBC Tech, a leading provider of next-generation Network-as-a-Service (NaaS) in more than 30 markets, today announced a partnership with Zenlayer, a leading global edge cloud services provider, to offer secure, high-speed and global-scale connectivity solutions to its customers worldwide.Leveraging Zenlayer's massively distributed infrastructure, CBC Tech will be able to extend the reach of eNet Secure Access Service Edge (SASE) Connect –- from the current 25 PoPs (Points-of-Presence) to 50 worldwide over its global fabric. Its customers will not only enjoy high-speed connectivity into far-reaching locations across ASEAN, India, Middle East, and South America, but also access a suite of additional edge cloud services powered by Zenlayer, such as bare metal, virtual machine, and IP transit. CBC Tech, at the same time, will deploy security services over the Zenlayer backbone as part of the joint offering. “We look forward to working with Zenlayer and leveraging their global infrastructure to be able to offer SASE services, integrating network and security services with deterministic and high-performance middle-mile optimization to global and China enterprises, powered by CBC’s eNet fabric. At the same time, via this collaboration, we can also enable Zenlayer to extend SASE service to their customers,” said Hoo Shu Yee, VP, Product and Solutions of CBC Tech. Operating over 280 hyperconnected edge nodes in the fastest-growing regions in the world, Zenlayer is a leader in edge cloud and the easy-button partner for SD-WAN/SASE providers to extend their global reach in an elastic and cost-effective manner. “The partnership combines advanced networking capabilities of both companies to drive our business forward and stay competitive in today’s market,” said David Xie, Chief Product Offer of Zenlayer. “It’s exciting to see how this strategic endeavour will enhance eNet SASE Connect’s capabilities to meet the rapidly evolving market needs.” About Zenlayer Zenlayer, a massively distributed edge cloud services provider, offers on-demand connectivity through its vast network of more than 280 PoPs, with expertise in fast-growing emerging markets like Southeast Asia, India, China, and South America. Businesses utilize Zenlayer’s global edge cloud platform to instantly improve digital experiences for their users with ultra-low latency and worldwide connectivity on demand. About CBC Tech CBC Tech is the next-gen Network-as-a-Service (NaaS) provider with our own patented SD-WAN/SASE fabrics across 50 cities in China and internationally. CBC Tech enables cloud-native network and security for enterprises over our SD-WAN fabrics integrated with our congestion-free IP and networks, mobile access, network appliance, multi-cloud connects and cloud-based security to deliver the best application performance with excellent customer experience. For more information about CBC Tech, please visit us at cbctech.com

Read More


ZEDEDA and Avassa Partner to Deliver Secure Edge Application Orchestration for Platform Teams and Developers

Businesswire | May 09, 2023

ZEDEDA, the leader in edge infrastructure orchestration, today announced its partnership with Avassa, a provider of a container management platform purpose-built for running applications in edge environments. This partnership will enable end-to-end orchestration through the entire stack, ensuring security and visibility from hardware to applications at the edge. The combination of the two solutions will provide customers with a flexible and robust foundation to support existing workloads with ZEDEDA and extend easily to modern container application stacks with Avassa’s edge platform. “ZEDEDA delivers a trusted and secure edge infrastructure built to support legacy virtual and modern container workloads. Avassa brings simplicity to managing all your containerized applications at the edge,” said Michael Maxey, VP of business development at ZEDEDA. “This partnership with Avassa provides a comprehensive solution that customers deploying technologies at the edge require.” As customers look to enable new use cases at the edge, they are developing modern, containerized apps that must run on the edge alongside existing virtualized applications. Through this partnership, ZEDEDA customers will be enabled with the full visibility and manageability of cloud-native container applications while also solving the challenges of existing infrastructure at the distributed edge. According to recent research, 69% of CIOs surveyed said that edge computing is a digital trend that they have already deployed or plan to deploy by mid-2025. This partnership addresses the growing demand for comprehensive solutions that empower customers to deploy new technologies at the edge. “We’re excited to work with ZEDEDA since they, too, purpose-built their solution for the distributed edge. By leveraging ZEDEDA’s edge-native infrastructure solutions together with Avassa’s edge container application platform, companies benefit from easy edge orchestration without compromising security,” said Avassa CTO Carl Moberg. “Together, users can leverage the very same edge virtualization solution from ZEDEDA to orchestrate their hardware and mixed workloads as well as modern containerized workloads, managed and monitored in the Avassa platform.” The Avassa Edge Enforcer will be available in the ZEDEDA Marketplace, providing fully automated deployment of container applications within customers’ distributed ZEDEDA environments. The integration will enable customers to easily access Avassa’s container management solution through one-click deployment. About ZEDEDA ZEDEDA makes edge computing effortless, open, and intrinsically secure — extending the cloud experience to the edge. ZEDEDA reduces the cost of managing and orchestrating distributed edge infrastructure and applications while increasing visibility, security and control. ZEDEDA delivers a distributed, cloud-native edge management and orchestration solution, simplifying the security and remote management of edge infrastructure and applications at scale. ZEDEDA ensures extensibility and flexibility by utilizing an open partner ecosystem with a robust app marketplace and leveraging an open architecture from the Linux Foundation. ZEDEDA delivers instant time to value, has thousands of nodes under management and is backed by world-class investors with teams in the United States, Germany and India. For more information, visit www.ZEDEDA.com. About Avassa Avassa empowers companies to bridge the gap between modern containerized applications development and operations and distributed edge infrastructure. With their application orchestration platform for the edge, Avassa enables businesses to manage the lifecycle of edge applications efficiently, securely, and at the speed of software. Founded in 2020 and based in Stockholm, Sweden, Avassa aims to unlock the cloud operating model for edge applications everywhere and was named a Gartner® Cool Vendor in Edge Computing 2022. Learn more and request a free trial at avassa.io.

Read More


EnGenius Launches Cloud SD-WAN for Efficient Remote Business Networks

EnGenius | March 10, 2023

On March 9, 2023, EnGenius, a leading networking and cloud solutions provider, released its line of SD-WAN Gateway products to meet the growing demand for cloud-based services that prioritize security and efficiency while connecting enterprises with their branch offices. The EnGenius SD-WAN series offers a secure and dependable solution that seamlessly integrates with EnGenius cloud-managed access points and switches to provide a unified view and enhanced security. Software-defined vast area network (SD-WAN) technology is becoming essential for the security and reliability of networks. This is because cloud-based and private enterprise networks are becoming increasingly data-driven. EnGenius SD-WAN Gateways give businesses a network solution that is highly scalable and reliable. With features like EnGenius Auto-VPN and NAT Traversal, the SD-WAN Gateway series provides a secure and efficient network solution. The EnGenius Cloud-Managed SD-WAN Gateway offers several features, such as: NAT traversal and Auto-VPN Topology of the network and VPN status Dual-Wan load balance and failover Firewall policies and stateful firewalls Unified with EnGenius devices and cloud management Automatic updates and touchless system maintenance Cloud To-Go: Management on the go Passthrough and routing options for multiple modes. EnGenius Cloud-managed SD-WAN Gateways: ESG320(new) - 4-Port Gigabit PoE+ ESG510 - 4-Port Multi-Gigabit PoE+ ESG610(new) - 4-Port Multi-Gigabit PoE+ ESG620(new) - 8-Port Multi-Gigabit PoE+ with 2 SFP+ Slots Antony Lai, Vice President at EnGenius, stated, "We are very excited to be launching this new line of SD-WAN products that provide greater stability and security for enterprises and the branches. "He added, "We have worked hard to provide a high-level solution that is easy to deploy and manage while providing seamless integration with EnGenius Cloud-managed products with a secure, reliable, and effective system that is scalable to their needs for managing their networks." (Source – CISION PR Newswire) About EnGenius Founded in 1999, EnGenius Technologies is a global leader in business networking, with its headquarters in Costa Mesa (California). With over 20 years of experience, the company offers a diverse range of feature-rich network solutions that enhance productivity and help businesses achieve their growth objectives.

Read More