Big Data Needs Cloud Networking

Over the past three years computing architectures have bifurcated, resulting in markedly different network architectures for legacy enterprise and next generation cloud network data centers. Arista has been a leader in championing cloud networks that respond to two key trends.

The first driver has been server virtualization, which literally rewrote the rules on server architecture, driving increased utilization of multi-core processors and increased memory density to facilitate the hosting of the largest number of virtual machines possible on a per server basis. Technologies like virtual machine mobility (enabled by Arista’s VM Tracer and partnership with industry leader VMware) have also driven the need for 10 Gigabit Ethernet on servers, as the default settings for vMotion can sustain multiple gigabits of throughput on a per VM basis.

The second important driver is “Big Data” – resulting from the requirements of Web 2.0 environments, large scale databases, Hadoop clusters, business intelligence analytics and compute intensive applications. Big Data is comprised of datasets that grow so large that they become cumbersome to work with using traditional database management tools. Difficulties in these environments include data capture, storage, search, sharing, analytics, and visualizing. With large numbers of parallel transactions, massive memory demands, and high concurrent transactional volume, Big Data and data analytics have given birth to the need for cloud network architectures designed for low-latency IO, high memory throughput, and power-efficient compute operations.

Standards-Based - Not Proprietary Marketectures

What is clear is that new application drivers are creating a new class of cloud networks as a disruption from legacy enterprise over-subscribed and congested switches. These days, the popular marketing approach from large vendors is proprietary architectures touting “fabric” as their new paradigm. While this hype is good for creating sensational news, it is not a practical real-world view of customer’s need for a standards-based L2/L3 Ethernet-based approach that is multivendor interoperable. This Arista value supports a cloud network architecture that scales to 20,000+ nodes. We build our product family on open standards with purpose-built self-healing software which is now deployed in 700+ customers. Designed from the ground up for the new decade, Arista’s Extensible Operating System, EOS, is a dramatic departure from the legacy and fault-prone monolithic network OS model of a Cisco IOS or even Juniper JUNOS from the 1990-2000 era.

The use of distributed databases and analytics packages demand massively parallel software running on hundreds, or even thousands of servers. This notion directly ties in to the role of cloud networking which supports fast and ubiquitous storage access for these applications, and highlights the essential advantages of a solid software foundation such as EOS with its intrinsic self-healing properties.

New Asset Center for Pervasive Big Data

Data is everywhere, whether it originates from users, applications, or machines and it’s growing exponentially with no vertical or industry being spared. Big Data is applicable in many areas such as meteorology, genomics, biological research, Internet search, finance, and business informatics. Arista is a clear thought leader and market player in this arena.

Our customers are reshaping the traditional networking industry, burdened as it is today by monolithic software and performance bottlenecks. They truly expect their network to be a productivity asset which they can leverage to generate profitable business models by utilizing best of breed standards-based products – often turning to Arista for real products to solve real world problems.

Modern “Big Data” environments are a disruptive force to traditional network architectures in the data center. Welcome to Arista’s standards-based cloud network architecture for Big Data applications. As always, I welcome your views at feedback@arista.com.

References:

• Big Data Cloud Storage - White Paper
• Networking in the Hadoop Cluster - Technical Bulletin