2020 OCP Virtual Summit

Case study: Aarki upgrades its data centers in 5 locations with Netris and Mellanox to achieve performance, reliability and agility at a massive scale.

Customer case study - deploying a prebuilt OCP server platform for Nexedi's Rapid.Space

Justin will share Hydro66's experience and lessons learned accepting pre-populated OCP racks into a colocation facility designed for traditional 19" equipment, with an emphasis on the practicalities of painless deployment.

The OCP Foundation developed the requirements and process for a colocation data center facility provider to become a listed Colo Solution Provider on the OCP Marketplace. Being an OCP Ready™ Colo Facility means Enterprise clients have an assured way to deploy OCP servers, storage and networking equipment in an extremely cost-effective and scalable way. Compliance means a proven method for accepting OCP pre-built racks, safe in the knowledge that the data center has been pre-vetted and meets or exceeds all the criteria for hosting OCP gear.

Justin will also describe the company's experiences in applying for and gaining OCP Ready™ Facility status. Hydro66 was the first facility in mainland Europe to achieve OCP Ready™ certification.

The goal of this session is to overview, outline, and explain the technology to deploy & provision a scale-out kubernetes cluster for running & orchestrating containers & VMs on bare metal. Much of the kubernetes ecosystem is focused on cloud & shared infrastructure platforms, but there are certain customer use cases & characteristics that simply require bare-metal nodes - in fact, even cloud platforms continue to run on bare-metal servers one way or another ;)

This presentation will present our experience taking a high-density OCP rack with over 1,000 compute cores and 9TB of memory from bare hardware to running a serverless demo app in 10 minutes. We will outline the choice of OS components, network connectivity, deployment of master & shared service nodes, and parallel deployment of compute nodes, preparing everything to process application traffic.

There are many layers to the kubernetes deployment stack, including infrastructure components built on other, lower level infrastructure components - sometimes four and five and six layers deep. This provides convenient abstractions for different engineering, deployment, and service levels - but also results in a large amount of total deployed software when considered end-to-end. Deploying such clusters on racks of bare metal gear - whether OCP or otherwise - requires preparation and planning, but does not have to be a months-long endeavor. In order to support customers and SPs that require scale-out on-premise hardware solutions, we have had to evaluate and optimize various approaches to bare metal deployment and will share our learnings here.

Minimizing data center energy use requires a 'holistic' cooling approach, implying that both the air handler and server fans must be operated with control algorithms that work together. While some conditions can increase efficiencies of the cooling system, the effect on the server fans and CPU current leakages must also be considered. To obtain the recipe to run the data center at maximum efficiency a 'sweet spot' is identified, yielding conditions that are favorable for both the cooling equipment and servers.

OCP servers are tested in a windtunnel at the ICE data center research facility at RISE Research Institutes of Sweden in LuleÂ. The 10 meter windtunnel consits of a heat exchanger for cooling, a heater, humidifier and a fan. The windtunnel can be used to control the pressure, temperature and humidity at the inlet to the server while measuring parameters such as temperature differential and server airflow.

The workshop presents a custom fan controller that has been implemented on OCP servers to maintain CPU temperatures at a desired setpoint irrespective of the workload, inlet temperature or pressure. The cooling control software is adjusted depending on operating conditions of the servers to minimize the cooling energy.

Adopting this holistic cooling approach with OCP servers and EcoCooling direct fresh air ventilation units has achieved PUE values less than 1.01 as demonstrated in the EU Horizon 2020 funded project, Boden Type DC, where two 150kW OCP data centers were tuned with holistic control tactics.

The Boden Type Data Center (BTDC) is an EU funded Horizon 2020 project with the key objective of demonstrating a range of technologies, creating a small scale OCP style data center that operates with extreme efficiency. This 500kW project includes a 150kW pod dedicated to the efficient operation of legacy Windmill OCP equipment. Additionally, a 150kW module is located in the ICE data center research facility at RISE Research Institutes of Sweden in LuleÂ, which houses Winterfell OCP equipment. The data centers are both equipped with EcoCooling direct fresh air ventilation units that contain supplemental evaporative function to provide cooling. This case study involves a live demonstration of the two data centers to show how new holistic process controls can be employed to minimize the energy used in providing an appropriate thermal environment for the servers and maintaining stable microprocessor operating conditions. One data center shows such stability under fluctuating workloads. The other data center highlights similar stability of microprocessor temperatures under varying supply temperatures, which reflects different climates. In both case the effectiveness of the cooling is exemplified by achieving PUE values of less than 1.01.

The workshop explains how such a performance can be achieved using OCP equipment with only software changes and simple inter-connectivity of the microprocessors, a custom server fan speed algorithm and the cooler control system. Consideration is given as to whether the demonstrated integrated approach could give OCP hardware a competitive advantage in emerging markets.

In many areas of Technology, African businesses have jumped the curve, partly due to not having invested in stepping-stone technologies. However, in the Data Centre, Vanity Brands have held a stronghold that has held back content development and distribution at a price that is consumable by over a billion Africans.

Using its own Case Studies based on implementation of Open Compute infrastructure, Atlancis is working with Mobile Network Operators, Banks and Service Providers to reduce the time and cost of distributing valuable content, and in so doing, democratising the internet for Africa.

This presentation will take the audience on Atlancis' OCP journey and look at the future role OCP will play in developing Africa's technology landscape.

Cloud&Heat Technologies first implemented OCP rack and server design preliminary specifications in a customer project in 2019. The result was a 40 ft mobile data center container with a power density of 400 kW and equipped with 125 innovative, in-house developed, water cooled server chassis for server hosting and rendering farm purposes. The compact rack construction was adapted in accordance with the OCP 21ʺ standard. The backplane was designed to provide an interface with the power supply as well as a connection to the cooling system, enabling flexible and snap-in insertion of the chassis. With a design PUE of 1.08 the container saves more than 300,000 kWh electric energy and over 200 t CO2 per year compared to a conventional datacenter. With German energy prices that leads to a saving of more than 60,000 $ per year. This keynote will give an insight into the design challenges met and practical experience gained during planning, installation and operation of the prototype. Focusing on potential contributions to the OCP project, I will present Cloud&Heatís latest developments in the field of water cooling. I will also show direct water-cooled power supply units with a 3 kW and 15 kW power supply. All in all, Cloud&Heat has 4,500 servers now in the field. Our company is focused on sustainable and green IT hardware and software services; having evaluated a variety of cooling solutions, we found hot-liquid cooling to be the best and most efficient way of realizing heat recovery and seamless building integration.

This presentation will talk about emerging trends in campus networks and the benefits of a common operational model for data center and campus networks. It will also address automated network validation and troubleshooting techniques for campus networks and the use of modern CI methods for automated deployment of new software and configurations.

Since 2016, we have adopted more than 4,000 OCP servers in our data centers. We will explain how we approach OCP deals for adoption and how we integrate and operate OCP servers and other related components in our infrastructure. Specifically, we will explain the day to day activities of YJ's server team and some of the OCP initiatives we are working on. We will also explain the differences between Japan and other countries related to the adoption of OCP. We will use a series of metrics to compare OCP servers and EIA19(Standard Servers), including CAPEX and OPEX, to highlight the benefits of OCP adoption. Since adopting OCP servers our the data center has changed more than ever. We will talk about why it has changed, what has changed and how this has improved the mindset and facilities at yahoo japan.

As a Hong Kong based carrier that has built orchestration software to help other Telcos convert their legacy business to Overlay based SDN network, DCConnect has highly demanding networking requirements that necessitated a move from legacy, vendor centric switching to an OCP-based white-box model.  In this presentation, Jonathan will share our experience of our transition to that model, including a brief overview of the company, the requirements that sparked the change, the original state of our network, the transition to our OCP-based current state, and where we will go from there.