Friday, December 1st 2023
Ethernet Switch Chips are Now Infected with AI: Broadcom Announces Trident 5-X12
Artificial intelligence has been a hot topic this year, and everything is now an AI processor, from CPUs to GPUs, NPUs, and many others. However, it was only a matter of time before we saw an integration of AI processing elements into the networking chips. Today, Broadcom announced its new Ethernet switching silicon called Trident 5-X12. The Trident 5-X12 delivers 16 Tb/s of bandwidth, double that of the previous Trident generation while adding support for fast 800G ports for connection to Tomahawk 5 spine switch chips. The 5-X12 is software-upgradable and optimized for dense 1RU top-of-rack designs, enabling configurations with up to 48x200G downstream server ports and 8x800G upstream fabric ports. The 800G support is added using 100G-PAM4 SerDes, which enables up to 4 m DAC and linear optics.
However, this is not only a switch chip on its own. Broadcom has added AI processing elements in an inference engine called NetGNT (Networking General-purpose Neural-network Traffic-analyzer). It can detect common traffic patterns and optimize data movement across the chip. Specifically, the company has listed an example of the system doing AI/ML workloads. In that case, NetGNT performs intelligent traffic analysis to avoid network congestion in these workloads. For example, it can detect the so-called "incast" patterns in real-time, where many flows converge simultaneously on the same port. By recognizing the start of incast early, NetGNT can invoke hardware-based congestion control techniques to prevent performance degradation without added latency.
Source:
Broadcom
However, this is not only a switch chip on its own. Broadcom has added AI processing elements in an inference engine called NetGNT (Networking General-purpose Neural-network Traffic-analyzer). It can detect common traffic patterns and optimize data movement across the chip. Specifically, the company has listed an example of the system doing AI/ML workloads. In that case, NetGNT performs intelligent traffic analysis to avoid network congestion in these workloads. For example, it can detect the so-called "incast" patterns in real-time, where many flows converge simultaneously on the same port. By recognizing the start of incast early, NetGNT can invoke hardware-based congestion control techniques to prevent performance degradation without added latency.
37 Comments on Ethernet Switch Chips are Now Infected with AI: Broadcom Announces Trident 5-X12
I, for one, look forward to the robot uprising - it can't be any worse than the current reality which seems dead set of meeting the predictions of 2006's Idiocracy.
the dream of every networking guy ever: a obscure proprietary black-box doing non-deterministic and random shit to the data packets!
¡Best troubleshooting ever!
www.broadcom.com/company/news/product-releases/61571"NetGNT works in parallel to augment the standard packet-processing pipeline. The standard pipeline is one-packet/one-path, meaning that it looks at one packet as it takes a specific path through the chip’s ports and buffers. NetGNT, in contrast, is an ML inference engine and can be trained to look for different types of traffic patterns that span the entire chip."According to broadcom the AI can be trained to recognize different traffic patterns. So long as network engineers are allowed to enabled, disable, and train the AI themselves I'd say that's a pretty good level of control. If you have control over the training you can trace back outcomes to the responsible inputs and how they interacted with the responsible neurons and weights which in turn you can use to improve the AI. Theoretically the use of AI should be able to improve traffic control as for many enterprise networks an algorithm based approach cannot optimize for every single network traffic scenario whereas an AI can. It depends on the network though, specialized traffic control algorithms will still be superior for networks with very specific traffic patterns.
This is much faster then rule base mitigation (though most use it in conjunction) there are still inline filtering technologies like wan guard, or corero, maybe you have something from fortinet or you have your APIs tied into akami or cloudflare or noction so you can re route traffic.
Sounds like a lot of "AI bad and I dont understand" in here.
Where are you located? You telling me your network engineers are logged into a terminal watching the L3 or centurylink connection over seabone to south america? We hvent done it that way since like the early 80s. You should talk to your NOC and network engineers and get some budget to up your tooling.
Unlike neural network processors, which were previously called AI, neural network "thinks" and tries to come to a logical conclusion (like a standard living brain [lmao Jellyfish]).
AI is now being included into everything, but all it does it collect massive amount of information and picks the best response out of a hat. Shit supercomputers were built for. There is no logic or thinking involved.
It is artificial, in every sense of the word.