Just yesterday, I was a suit-wearing Musk-eteer, deep in thoughts of “wow, what a world we live in. With artificial intelligence (AI) definitely sluiced into the public consciousness with good intentions, and the continued persistence of cloud computing trading on the thrown scrap of convenience and cost-effectiveness against the dark ‘service-as-a-service’ whims of any company comically evil enough to enter into a consortium.”
And then, I was struck: what if the all new X dot com’s ethernet infrastructure isn’t reliable enough to power this new world? Luckily, a press release from The Linux Foundation announced THE ULTRA ETHERNET CONSORTIUM (UEC) alongside a whole range of big-name tech companies you’ll recognise, like Meta and Microsoft.
It’s all a bit “the enemy of my enemy”, but maybe the Mighty Morphin’ Consortium’s aims will be clear and concise in how they’ll benefit businesses.
[Looks to webcam]
“To deliver on Ethernet-based open, interoperable, high-performance full-communications stack architecture to meet the growing network demands of AI & HPC [high-performance computing] at scale” has, as it goes, been the lofty objective of man for centuries.
My father often speaks fondly of his father’s father predicting that, “come that new century, they’ll deliver on Ethernet-based open, interoperable, high-performance full-communications stack architecture to meet the growing network demands of AI & HPC at scale.”
Many campaigners gave their lives campaigning for Ethernet-based open, interoperable, high-performance full-communications stack architecture, and this ideal, never dying, has filled the world with joy, hope, and limitless ambition.
Alas, now Dr. J. Metz, Chair of the UEC, is ripping the plaster off and setting ‘realistic expectations’, (just when we thought we were done with them experts).
“This isn’t about overhauling Ethernet,” he wrote, in a stark disregard for his lifelong fans. “It’s about tuning Ethernet to improve efficiency for workloads with specific performance requirements.”
Love an ethernet but couldn’t radically reform a whole one x
His incendiary screed, expected to bring progressives to arms in the streets to spit “BIG WIFI” into his doorstep, indicated that ethernet speed is not the concern, but rather establishing and maintaining “specifications, APIs and source code”, targeting four key areas to standardize:
1) protocols and data structures for Ethernet communication
2) link-level and end-to-end network transport protocols,
3) similar mechanisms targeting congestion, telemetry and signalling
4) “Software, storage, management and security constructs”, to suit “a variety of workloads”, presumably AI, machine learning and cloud computing employments
If all of this is giving you brain damage, take solace in another soundbite from the soon-to-be-exiled Metz: “We’re looking at every layer – from the physical all the way through the software layers – to find the best way to improve efficiency and performance at scale.”
All this reliability and standardization stuff sounds modest and achievable, which I suppose is great news for businesses in places with solid connection speeds, those there Londons and Silicon Valleys, but less so for those of us left running our father’s father’s AI startup down’ Tinsley pit with 36/15 megabits a second.
Via Tom’s Hardware
Source link
