Author Archives: Jayshree Ullal
Author Archives: Jayshree Ullal
As client users, devices, and IoT continue to proliferate, the need for switching management and workload optimization across domains increases. Many sub-optimal and closed approaches have been designed in the past. Arista was founded to build the best software and hardware, equating to the highest performance and density in cloud/data centers, and now evolving to campus switches. In 2020, we introduced the smallest footprint of Arista CCS 750 and 720 series switches as a fitting example of the highest density and lowest footprint.
In 1984, Sun was famous for declaring, “The Network is the Computer.” Forty years later we are seeing this cycle come true again with the advent of AI. The collective nature of AI training models relies on a lossless, highly-available network to seamlessly connect every GPU in the cluster to one another and enable peak performance. Networks also connect trained AI models to end users and other systems in the data center such as storage, allowing the system to become more than the sum of its parts. As a result, data centers are evolving into new AI Centers where the networks become the epicenter of AI management.
As we all recover from NVIDIA’s exhilarating GTC 2024 in San Jose last week, AI state-of-the-art news seems fast and furious. Nvidia’s latest Blackwell GPU announcement and Meta’s blog validating Ethernet for their pair of clusters with 24,000 GPUs to train on their Llama 3 large language model (LLM) made the headlines. Networking has come a long way, accelerating pervasive compute, storage, and AI workloads for the next era of AI. Our large customers across every market segment, as well as the cloud and AI titans, recognize the rapid improvements in productivity and unprecedented insights and knowledge that AI enables. At the heart of many of these AI clusters is the flagship Arista 7800R AI spine.
Recently I attended the 50th golden anniversary of Ethernet at the Computer History Museum. It was a reminder of how familiar and widely deployed Ethernet is and how it has evolved by orders of magnitude. Since the 1970s, it has progressed from a shared collision network at 2.95 megabits in the file/print/share era to the promise of Terabit Ethernet switching in the AI/ML era. Legacy Ethernot* alternatives such as Token Ring, FDDI, and ATM generally get subsumed by Ethernet. I believe history is going to repeat itself for AI networks.
The perimeter of networks is changing and collapsing. In a zero trust network, no one and no thing is trusted from inside or outside of the enterprise network without verification or network access control (NAC). However, for years, organizations have been saddled with bolt-on NAC technologies that deliver cost complexity while failing to be effective. Instead, security-conscious organizations are shifting to a “microperimeter” enterprise that embeds security into the network infrastructure as the proactive way to defend today’s wider attack surface.
The AI industry has taken us by storm, bringing supercomputers, algorithms, data processing and training methods into the mainstream. The rapid ramp of large language inference models combined with Open AI's ChatGPT has captured the interest and imagination of people worldwide. Generative AI applications promise benefits to just about every industry. New types of AI applications are expected to improve productivity on a wide range of tasks, be it marketing image creation for ads, video games or customer support. These generative large language models with over 100 billion parameters are advancing the power of AI applications and deployments. Furthermore, Moore's law is pushing silicon geometries of TPU/GPU processors that connect 100 to 400 to 800 gigabits of network throughput with parallel processing and bandwidth capacity to match.
A pioneer in cloud networking for the last decade, Arista has become synonymous with elastic scaling and programmable provisioning through a modern data-driven software stack. Legacy networks with manual box-by-box configurations for production and testing have led to cumbersome and complex practices. Arista leads the industry in cloud automation built on an open foundation.
As we enter 2022, there is much discussion on the “post-pandemic” world of campus and how it’s changing. Undoubtedly, the legacy 2000 era campus was mired in complexity, with proprietary features, siloed designs, and fragile software ripe for change. This oversubscribed campus is riddled with challenges, including critical outages causing risk-adverse behaviors and labor-intensive roll-outs hampering improvements. The future of the campus has changed as the lines between corporate headquarters, home, remote and transit workers are blurring and creating distributed workspaces. Before the pandemic, the most common network designs were rigidly hierarchical. They were based upon a manual model developed in the mid-1990s. As the demand for scale increased, the end user experience was degraded and the cost per connected host continued to escalate.
Are we ready to evolve the legacy campus to a new cognitive edge for the new and dispersed class of users, devices and IoT/OT? I think so and the time to recalibrate and redesign the campus is now!
Over the last few years, we have seen an age of edgeless, multi-cloud, multi-device collaboration for hybrid work giving rise to a new network that transcends traditional perimeters. As hybrid work models gain precedence through the new network, organizations must address the cascading attack surface. Reactionary, bolt-on security measures are simply too tactical and expensive.
The rapid arrival of real-time gaming, virtual reality and metaverse applications is changing the way network, compute memory and interconnect I/O interact for the next decade. As the future of metaverse applications evolve, the network needs to adapt for 10 times the growth in traffic connecting 100s of processors with trillions of transactions and gigabits of throughput. AI is becoming more meaningful as distributed applications push the envelope of predictable scale and performance of the network. A common characteristic of these AI workloads is that they are both data and compute-intensive. A typical AI workload involves a large sparse matrix computation, distributed across 10s or 100s of processors (CPU, GPU, TPU, etc.) with intense computations for a period of time. Once the data from all peers is received, it can be reduced or merged with the local data and then another cycle of processing begins.
Arista’s EOS (Extensible Operating System) has been nurtured over the past decade, taking the best principles of extensible, open and scalable networks. While SDN evangelists insisted that the right way to build networks started with the decoupling of hardware and software in the network, manipulated by a centralized, shared controller, many companies failed to provide the core customer requisite in a clean software architecture and implementation coupled with key technical differentiation. This has been the essence of Arista EOS.
The power and potential of the next generation cognitive campus are transformative as the industry undergoes a massive transition to hybrid work in the post-pandemic era. A key underpinning to successful campus networking deployments has been our very first acquisition of Mojo Networks for cognitive Wi-Fi. Arista’s entry into wireless is only in its third year, yet the advances in this space will be profound over the next decade.
In the past decade, the emergence of cloud networks has blurred the line between switching and routing versus traditional routers. Today the industry is at an inflection point, where the adoption of cloud principles for routing intersects the rapidly expanding capabilities of the merchant silicon feature set and scale, creating a disruption of legacy routing architectures.
The rise of cloud migration for enterprises with mission critical applications is redefining the data center. The reality for any enterprise: a systematic approach balancing workloads in the cloud and premises while securing data. Data and applications must be managed as critical assets in the 21st century.
The rise of cloud migration for enterprises with mission critical applications is redefining the data center. The reality for any enterprise: a systematic approach balancing workloads in the cloud and premises while securing data. Data and applications must be managed as critical assets in the 21st century.
Every CIO needs to adopt a cloud strategy typically moving some e-commerce workloads to the public cloud. Yet, the migration path for the modern enterprise can be constrained by legacy barriers. With mission-critical applications that run in a diverse suite of legacy mainframe to helpdesk to IoT devices, how does one get started and what does this entail?
The reality for any enterprise whose core business is driven by a reliance on corporate-owned technology structure with strict ownership of critical assets is that it operates with many constraints. The cloudification and multi-cloud strategy requires a more pragmatic and systematic approach balancing workloads in the cloud and on-premise enterprise networks.
Every CIO needs to adopt a cloud strategy typically moving some e-commerce workloads to the public cloud. Yet, the migration path for the modern enterprise can be constrained by legacy barriers. With mission-critical applications that run in a diverse suite of legacy mainframe to helpdesk to IoT devices, how does one get started and what does this entail?
The reality for any enterprise whose core business is driven by a reliance on corporate-owned technology structure with strict ownership of critical assets is that it operates with many constraints. The cloudification and multi-cloud strategy requires a more pragmatic and systematic approach balancing workloads in the cloud and on-premise enterprise networks.
Arista is trusted and powers the world’s largest data centers and cloud providers based on the quality, support and performance of its products. The experience gained from working with over 7000 customers has helped redefine software defined networking and many of our customers have asked us how we plan to address security. To us, security must be a holistic and inherent part of the network. Our customers have been subjected to the fatigue of point products, reactive solutions, proprietary vendor lock-ins and most of all, operational silos created between CloudOps, NetOps, DevOps and SecOps. By leveraging cloud principles, Arista’s cloud network architectures bring disparate operations together to secure all digital assets across client to IoT, campus, data center and cloud protecting them from threats, thefts and compromises.
Arista is trusted and powers the world’s largest data centers and cloud providers based on the quality, support and performance of its products. The experience gained from working with over 7000 customers has helped redefine software defined networking and many of our customers have asked us how we plan to address security. To us, security must be a holistic and inherent part of the network. Our customers have been subjected to the fatigue of point products, reactive solutions, proprietary vendor lock-ins and most of all, operational silos created between CloudOps, NetOps, DevOps and SecOps. By leveraging cloud principles, Arista’s cloud network architectures bring disparate operations together to secure all digital assets across client to IoT, campus, data center and cloud protecting them from threats, thefts and compromises.