In December 2019, we made a daring announcement about how we’d ceaselessly trade the economics of the web and pressure innovation at speeds like nobody had ever observed sooner than. Those had been bold claims, and now not unusually, many of us took a wait-and-see angle. Since then, we’ve endured to innovate at an an increasing number of speedy tempo, main the {industry} with leading edge answers that meet our consumers’ wishes.

Lately, simply 3 and a part years after launching Cisco Silicon One™, we’re proud to announce our fourth-generation set of gadgets, the Cisco Silicon One G200 and Cisco Silicon One G202, which we’re sampling to consumers now. Generally, new generations are introduced each 18 to 24 months, demonstrating a tempo of innovation that’s two instances sooner than commonplace silicon building.

The Cisco Silicon One G200 gives the advantages of our unified structure and focuses in particular on enhanced Ethernet-based synthetic intelligence/system finding out (AI/ML) and web-scale backbone deployments. The Cisco Silicon One G200 is a 5 nm, 51.2 Tbps, 512 x 112 Gbps serializer-deserializer (SerDes) instrument. This can be a uniquely programmable, deterministic, low-latency instrument with complicated visibility and keep an eye on, making it the perfect selection for web-scale networks.

The Cisco Silicon One G202 brings identical advantages to consumers who nonetheless need to use the 50G SerDes for connecting optics to the transfer. This can be a 5 nm, 25.6 Tbps, 512 x 56 Gbps SerDes instrument with the similar traits because the Cisco Silicon One G200 however with part the functionality.

To reach the imaginative and prescient of Cisco Silicon One, it was once crucial for us to spend money on key applied sciences. Seven years in the past, Cisco started making an investment in our personal high-speed SerDes building and discovered straight away that as speeds build up, the {industry} will have to transfer to analog-to-digital (ADC)-based SerDes. SerDes acts as a basic construction block of networking interconnect for high-performance compute and AI deployments. Lately, we’re happy to announce our next-generation, ultra-high functionality, and low-power 112 Gbps ADC SerDes in a position to ultra-long attain channels supporting 4-meter direct-attach cables (DACs), conventional optics, linear pressure optics (LDO), and co-packaged optics (CPO), whilst minimizing silicon die space and gear.

The Cisco Silicon One G200 and G202 are uniquely situated within the {industry} with complicated options to optimize real-world functionality of AI/ML workloads—whilst concurrently using down the price, chronic, and latency of the community with important inventions.

The Cisco Silicon One G200 is the perfect resolution for Ethernet-based AI/ML networks for a number of causes:

~ With the {industry}’s very best radix transfer, with 512 x 100GE Ethernet ports on one instrument, consumers can construct a 32K 400G GPUs AI/ML cluster with a 2-layer community requiring 50% much less optics, 40% fewer switches, and 33% fewer networking layers—enormously decreasing the environmental footprint of the AI/ML cluster. This protects as much as 9 million kWh according to 12 months, which in step with the U.S. Environmental Coverage Company is identical to greater than 6,000 metric lots of carbon dioxide (CO2e) or burning 7.3 million kilos of coal according to 12 months.

~ Complex congestion-aware load balancing ways permit networks to keep away from conventional congestion occasions.

~ Complex packet-spraying ways reduce introduction of congestion scorching spots within the community.

~ Complex hardware-based link-failure restoration delivers optimum functionality throughout huge web-scale networks, even within the presence of faults.

Right here’s a better take a look at a few of our many Cisco Silicon One–comparable inventions:

Converged structure

~ Cisco Silicon One supplies one structure that may be deployed throughout buyer networks, from routing roles to web-scale front-end networks to web-scale back-end networks, dramatically decreasing deployment timelines, whilst concurrently minimizing ongoing operations prices by way of enabling a converged infrastructure.

~ The use of a not unusual tool building equipment (SDK) and usual Transfer Abstraction Interface (SAI) layers, consumers want simplest port the Cisco Silicon One atmosphere to their community working components (NOS) as soon as and employ that funding throughout numerous community roles.

~ Like any our gadgets, the Cisco Silicon One G200 has a big and entirely unified packet-buffer optimizing burst-absorption and throughput in huge web-scale networks. This minimizes head-of-line blockading by way of soaking up bursts as a substitute of the era of precedence circulation keep an eye on.

Optimization throughout all of the worth chain

~ The Cisco Silicon One G200 has as much as two instances upper radix than different answers with 512 Ethernet MACs, enabling consumers to noticeably scale back the price, chronic, and latency of community deployments by way of doing away with layers in their community.

~ With our personal internally evolved, next-generation, SerDes generation, the Cisco Silicon One G200 instrument is in a position to using 43 dB bump-to-bump channels that permit co-packaged optics (CPO), linear pluggable gadgets (LPO), and using 4-meter 26 AWG copper cables, which is well past IEEE requirements for optimum in-rack connectivity.

~ The Silicon One G200 is over two instances extra chronic environment friendly with two instances decrease latency in comparison to our already optimized Cisco Silicon One G100 instrument.

~ The bodily design and structure of the instrument is constructed with a system-first way, permitting consumers to run components lovers slower, dramatically lowering components chronic draw.

Leading edge load balancing and fault detection

~ Beef up for non-correlated, weighted equal-cost multipath (WECMP) and equal-cost multipath (ECMP) load balancing functions with near-ideal traits lend a hand to keep away from hash polarization, even throughout huge networks.

~ Congestion-aware load balancing for stateful ECMP, circulation, and flowlet permits optimum community throughput with optimum flow-completion time and job-completion time (JCT).

~ Congestion-aware stateless packet spraying permits close to preferrred JCT by way of the usage of all to be had community bandwidth, without reference to circulation traits.

~ Beef up for hardware-based redistribution of packets according to hyperlink disasters permits Cisco Silicon One G200 to optimize real-world throughput of big scale networks.

Complex packet processor

~ The Cisco Silicon One G200 makes use of the {industry}’s first absolutely customized, P4 programmable parallel packet processor in a position to launching greater than 435 billion lookups according to 2d. It helps complicated options like SRv6 Micro-SID (uSID) at complete fee and is extendable with complete run-to-completion processing for much more advanced flows. This distinctive packet processing structure permits flexibility with deterministic low latency and gear.

Deep visibility and analytics

~ Programmable processors permit toughen for usual and rising web-scale in-band telemetry requirements offering industry-leading community visibility.

~ Embedded {hardware} analyzers discover microbursts with pre- and post-event logging of temporal circulation data, giving community operators the facility to investigate community occasions after the reality with {hardware} time visibility.

A brand new era of community functions

Long past are the times when the {industry} operated in silos. With its one unified structure, Cisco Silicon One erases the exhausting dividing strains that experience outlined our {industry} for too lengthy. Shoppers now not wish to concern about architectural variations rooted in previous creativeness and generation obstacles. Lately, consumers can deploy Cisco Silicon One in a mess of the way throughout their networks.

With the Cisco Silicon One G200 and G202 gadgets, we prolong the attain of Cisco Silicon One with optimized high-bandwidth gadgets purpose-built for backbone and AI/ML deployments. Shoppers can lower your expenses by way of deploying fewer and extra environment friendly gadgets, experience new deployment topologies with ultra-long-reach SerDes, strengthen their AI/ML task functionality with leading edge load balancing and fault discovery ways, and strengthen community debuggability with complicated telemetry and {hardware} analyzers.

If you happen to’ve been gazing since we first introduced Cisco Silicon One in December 2019, it’s simple to look that that is just the start. We’re taking a look ahead to proceeding to boost up the worth addition for our consumers.

Keep tuned for extra thrilling Cisco Silicon One traits.

Be told extra about

structure, gadgets, and advantages.

Further Assets

Learn my first weblog on Silicon One: Construction AI/ML Networks with Cisco Silicon One

Proportion: