Nvidia’s planned $40 billion takeover of chip-architecture firm Arm Holdings is not your typical merger. Oftentimes in a merger it’s one company taking over a weaker competitor that it has vanquished, something Nvidia knows all too well. Over its history, Nvidia has purchased several competitor GPU makers, most notably 3DFX in 2000.But here, the situation is different. First, the two companies don’t compete. Nvidia was a licensee of Arm chip design with its Tegra processor aimed at smartphones and tablets—and a rare failure for Nvidia as it never really caught on.To read this article in full, please click here
After months of teasing and rumor, GPU and AI vendor Nvidia announced it would purchase Arm Holdings from its parent company SoftBank for $40 billion. The purchase includes $21.5 billion in Nvidia stock and $12 billion in cash, including $2 billion payable at signing. That will break the piggy bank because Nvidia had $10.9 billion in cash on hand as of the most recent quarter.Softbank acquired Arm in 2016 for $31.4 billion in 2016. At the time, SoftBank CEO Masayoshi Son said it was an investment in the Internet of Things. But SoftBank, known for its profligate spending on acquisitions and investments, made some bad investments in WeWork and Uber, among others, and was saddled with $25 billion in debt.To read this article in full, please click here
Lenovo Data Center Group on Thursday introduced four new hyperconverged infrastructure (HCI) products aimed at a variety of workloads, including virtual desktop infrastructure (VDI), SAP HANA, Microsoft Azure, and Kubernetes.HCI products have grown in popularity because they are easily deployed and can get a variety of workloads up and running quickly. HCI is available either in hardware/appliance form or as software. HCI hardware vendors are the usual suspects – HP Enterprise, Dell, Lenovo – while the software vendors include Nutanix and VMware.Lenovo is focused on ready-to-deploy HCI solutions from both software firms. It boasts that its hardware is easy to deploy and manage with simple updates, automatic scalability and a consumption-based use model.To read this article in full, please click here
Arm Ltd. last week announced the Cortex-R82, a chip that is both storage and data processing-capable, which could enable a whole new generation of storage devices that help process the data they store.Storage processor chips, such as those made by Marvell but also storage device makers like EMC, handle the I/O and disk management, but if you want to process the data, that job falls to the CPU. This means data has to be moved in and out of the drive to be processed, a job that falls to two separate devices.But there is an emerging hardware category known as computational storage where the processing is done where the data resides, rather than moving it into memory. Data is processed through various methods, like indexing and schema, eliminating the latency of data movement and freeing up the CPU. Obviously this can only be done on SSDs.To read this article in full, please click here
Secure access service edge (SASE) is a network architecture that rolls software-defined wide area networking (SD-WAN) and security into a cloud service that promises simplified WAN deployment, improved efficiency and security, and to provide appropriate bandwidth per application.Because it’s a cloud service, SASE (pronounced “sassy”) can be readily scaled up and scaled down and billed based on usage. As a result, it can be an attractive option in a time of rapid change.[Get regularly scheduled insights by signing up for Network World newsletters.]
While some vendors in this space offer hardware devices to connect at-home employees and corporate data centers to their SASE networks, most vendors handle the connections through software clients or virtual appliances.To read this article in full, please click here
Marvell Technology Group announced last week that it has decided to cancel its ThunderX3 Arm-based server processor for general-purpose server use in favor of vertical markets and the hyperscaler server market.Marvell was best known for making controllers for storage and networking devices before it bought Cavium, an Arm server developer, in 2018. The company announced the ThunderX3 in March and on paper it looked like a real monster, with 96 cores and four threads per core.To read this article in full, please click here
A German tech site claims to have internal AMD documents that show the next generation of AMD Epyc server processors will boast a significant performance gain. AMD declined to comment on the veracity of the article.Hardwareluxx posted what it said were details from internal AMD slides revealing the performance potential of AMD's next-gen server processors, codenamed “Milan,” otherwise known as Zen 3, due to ship later this year.According to the slides, Zen 3 is in many ways similar to the Zen 2 generation (aka “Rome”) currently on the market. It will be socket-compatible with the first and second generation of Epycs, so current owners can swap out the older chips for newer. It will have a maximum of 64 cores, which is the same as Rome. It will support DDR4 memory and PCI Express 4.0 interconnects, like Rome. One difference is that instead of two 16MB L3 caches, Milan will have one 32MB L3 cache.To read this article in full, please click here
Researchers at University College London claim they’ve obtained a new top internet speed of 178Tbps – a fifth quicker than the prior record and fast enough to download the entire Netflix catalog in under a second, they say.To achieve that, the researchers used different bandwidth ranges than are typically used in commercial optical systems. Traditional fiber infrastructure uses bandwidth of 4.5THz with 9THz becoming more available commercially. In UCL experiments, the scientists used 16.8THz.[Get regularly scheduled insights by signing up for Network World newsletters.]
To do this the researchers used a variety of amplifier technologies, customizing which ones they used for each wavelength to optimize its performance as measured by phase, brightness and polarization, according to a press statement put out by UCL. These customization packages are known as geometric signal constellations.To read this article in full, please click here
All-flash storage arrays are fairly common in the most mission-critical of environments, where response time is of the essence. But if you move a step down to where archival activity takes place, there are still plenty of hard-disk/flash hybrid storage arrays.That’s because hard-disk drives have retained one advantage over flash: capacity. HDDs from Seagate and Western Digital are pushing into the 20TB range for a whole lot less than a comparable flash drive.But Pure Storage thinks the second generation of its FlashArray//C storage platform will be a hybrid flash/disk storage killer because it has the capacity to match a hybrid array with an all-flash storage setup. The pitch is that with FlashArray//C’s capacity, enterprises can do away with hard-drive-based storage, which draws much more power than an SSD and generate far more heat.To read this article in full, please click here
Digital content should be considered a fifth state of matter, along with gas, liquid, plasma and solid, suggests one university scholar.Because of the energy and resources used to create, store and distribute data physically and digitally, data has evolved and should now be considered as mass, according to Melvin Vopson, a senior lecturer at the U.K.'s University of Portsmouth and author of an article, "The information catastrophe," published in the journal AIP Advances.Vopson also claims digital bits are on a course to overwhelm the planet and will eventually outnumber atoms.To read this article in full, please click here
Pluribus Networks has rolled out new software and analytics packages that take aim at customers looking to build and manage software-defined data-center fabrics.The packages include a new release of the company’s core network operating system, Netvisor One, and the accompanying Unum management software as well as a new version of its Insight Analytics platform. They're all designed to simplify the operations of large-scale traditional and distributed edge data centers, the company said.[Get regularly scheduled insights by signing up for Network World newsletters.]
Netvisor ONE is a virtualized NOS that provides Layer 2 and Layer 3 networking, distributed fabric intelligence. It virtualizes switch hardware and implements what the company calls an Adaptive Cloud Fabric. Adaptive Cloud Fabric operates without a controller and can be deployed across a single data center, or targeted to specific racks, pods, server farms or hyper-converged infrastructures, the company said. To read this article in full, please click here
Aiming to help customers support modern data-center networking technologies, Apstra has enriched its Intent-Based Networking software to include better operational features but also adds more connectivity for its third party support of Juniper and open source SONiC environments.The company’s core Apstra Operating System (AOS) was built from the ground up to support IBS in that once running it keeps a real-time repository of configuration, telemetry and validation information to constantly ensure the network is doing what the customer wants it to do.To read this article in full, please click here
There is no sadder moment in the backup world than finding out the file or database you need to restore has never been backed up. Understanding how systems, directories, and databases are included in the backup system is the key to making sure this never happens to you.The first step toward this goal is making sure that servers and services you want backed up are registered with your backup-and-recovery system.[Get regularly scheduled insights by signing up for Network World newsletters.]
For example, if you start using a new SaaS such as Salesforce, no backup system will automatically notice that addition and start backing it up for you. If you are fully virtualized on VMware, the systems will automatically notice if you add a new node to the configuration. But if you start using Hyper-V or KVM, no backup system will automatically notice there is a new hypervisor in the data center and start backing it up. And of course your backup system will not notice you installed a new physical server.To read this article in full, please click here
Most of the buzz around artificial intelligence (AI) centers on autonomous vehicles, chatbots, digital-twin technology, robotics, and the use of AI-based 'smart' systems to extract business insight out of large data sets. But AI and machine learning (ML) will one day play an important role down among the server racks in the guts of the enterprise data center. AI's potential to boost data-center efficiency – and by extension improve the business – falls into four main categories:To read this article in full, please click here
Microsoft is bringing artificial intelligence to the task of sorting through millions of servers to determine what can be recycled and where.The new initiative calls for the building of so-called Circular Centers at Microsoft data centers around the world, where AI algorithms will be used to sort through parts from decommissioned servers or other hardware and figure out which parts can be reused on the campus.
READ MORE: How to decommission a data center
Microsoft says it has more than three million servers and related hardware in its data centers, and that a server's average lifespan is about five years. Plus, Microsoft is expanding globally, so its server numbers should increase.To read this article in full, please click here
Privacy is one of the big holdups to a world of ubiquitous, seamless data-sharing for artificial intelligence-driven learning. In an ideal world, massive quantities of data, such as medical imaging scans, could be shared openly across the globe so that machine learning algorithms can gain experience from a broad range of data sets. The more data shared, the better the outcomes.That generally doesn't happen now, including in the medical world, where privacy is paramount. For the most part, medical image scans, such as brain MRIs, stay at the institution level for analysis. The result is then shared, but not the original patient scan data.
READ MORE: Cisco challenge winners use AI, IoT to tackle global problemsTo read this article in full, please click here
Another study finds that the data center is far from dying. That's not surprising to learn from the Uptime Institute's annual data center survey. However one trend that did stand out in the research is that power efficiency has "flatlined" in recent years.Uptime says big improvements in energy efficiency were achieved between 2007 and 2013 using mostly inexpensive or easy methods, such as simple air containment. But moving beyond those gains involves more difficult or expensive changes. Since 2013, improvements in power usage effectiveness (PUE) have been marginal, according to the group.To read this article in full, please click here
Visible light communications (VLC) systems are an alternative to radio-based wireless networks and serve a dual purpose: They provide in-building lighting, and they use light waves for data transmission. VLC uses modulated light as a data carrier, while the visible spectrum provides light.Using VLC for data transmission has some advantages. It offers decent bandwidth; it offers security because walls, floors and roofs obstruct the data-carrying wavelengths, which reduces the risk of eavesdropping; and it's inexpensive since it's simply incorporated into light fixtures or, in emerging developments, worked into displays and other surfaces.To read this article in full, please click here
Cisco has issued a number of critical security advisories for its data center manager and SD-WAN offering customers should deal with now.On the data center side, the most critical – with a threat score of 9.8 out of 10 – involves a vulnerability in the REST API of Cisco Data Center Network Manager (DCNM) could let an unauthenticated, remote attacker bypass authentication and execute arbitrary actions with administrative privileges on an affected device.Cisco DCNM lets customers see and control network connectivity through a single web-based management console for the company’s Nexus, Multilayer Director Switch, and Unified Computing System products.To read this article in full, please click here
Data center provider Switch has selected Tesla as the battery supplier for a massive solar project at its northern Nevada data-center facilities.It's a geographically easy alliance as Switch's campus is right near Tesla's Gigafactory Nevada manufacturing facility. While best known for its cars, Tesla has also made quite an entry in the battery space with products such as the Powerwall, Powerpack, and Megapack energy storage products.To read this article in full, please click here