top of page

A Revolution in Server Design

Legacy server architecture is struggling to keep up with modern software design.



We had the opportunity to meet with Tony Craythorne, CEO, John Goodacre, CSciO, and Vicki Grey, CMO at Bamboo Systems Group during the IT Press Tour #35.


According to Tony, legacy servers are not keeping up with the changes in technology. Traditional server architecture is not evolving quickly enough for modern, agile, highly-parallel workloads. The cost of operating data centers exceeds the cost to build them. Legacy server architecture struggles to keep up with modern software design. New workloads like edge, containers, and AI/ML are transforming requirements. Energy has also become a major economic and political concern.


Datacenter power consumption is now a global issue. Data centers consume 3.5% of the world’s energy, produce 2% of the world’s greenhouse emissions, and are overtaking the aviation industry as the world's biggest energy consumer.


A new wave of compute is being seen everywhere except the data center. Arm servers are becoming mainstream. Bamboo is betting on Arm servers to take a significant market share over the next five years. In 1Q2020, X86 server revenue declined by 9.1% while non-X86 increased 38.2%. AWS’ Graviton shows Arm servers can run anything X86 servers can. Apple is expected to announce they are moving away from Intel to Arm across its entire product line.


Building servers with Arm processors. Other Arm server vendors have fallen into the trap of following the traditional server architecture model, creating more powerful processors simply plugged into an x86 socket. To meet the needs of the next generation of data centers, new highly parallel, low energy architectures are required. Bamboo is being designed for next-generation data centers, where power is measured in kilowatt-hours not GHz or Xeon cores. Using embedded systems methodology, Bamboo’s server architecture uses COTS components in a highly dense configuration. This results in using a fraction of the energy for a given workload and a less expensive server.


Bamboo’s server architecture provides energy efficiency and compactness while meeting the scalability needs of modern high-performance workloads as well as processing at the edge. The Parallel Arm Node Designed Architecture (PANDA) provides:

  • More memory bandwidth enables faster data processing

  • Increases security by separating IO and Management from application

  • Delivers more performance in the same space using less energy than large power-hungry chips

  • Delivers the TCO benefits of converged platforms without impacting application performance

  • More efficient use of the available performance to run applications

  • No massive changes to operations/users as it runs the same Linux applications

  • Energy consumption is directly related to use, unlike traditional chipsets.


Use Cases

  • Enterprise DevOps with Kubernetes and Apache with compute that supports parallel microservices for modern design applications.

  • Industrial edge with IoT, data processing at the edge with space limitations, and limited power options for high-density compute with storage and networking with flexible power options.

  • Financial, healthcare, automotive, and smart cities with AI/ML and simulation workloads with price sensitivity in CPU power needed to process huge volumes of data.

  • Collocated clients offering PaaS with space and power constraints and the need for scalability.


For equivalent workloads, Bamboo saves 50% of the acquisition cost, 75% of energy consumption, and 80% of rack space.

コメント


bottom of page