In this article

The amount of data being produced today is staggering, and it's still growing at an exponential rate each year. Data sources are increasingly complex and varied, driving the need for organizations to deploy high-performance computing on a scale that can be challenging to make cost-efficient. 

At the same time, organizations looking to become more digitally agile are transitioning from purpose-built appliances to a more software-driven infrastructure that can deliver specific business outcomes.

To achieve this scale and flexibility, businesses are turning to non-proprietary, bare metal servers that have long been used by the Amazons, Googles and Microsofts of the world to tremendous success.

As the name implies, bare metal means just that — original design manufacturer (ODM) machines with no software, just CPUs, memory and storage. Customers provision these machines and provide all the software from the operating system on up. 

Web scalers like Google or Amazon have mastered this, which is why they're able to spin up services and bring them to market so rapidly. But as technology solutions become more complex, how many organizations are equipped to prepare these servers for seamless network operation? 

Current methods for provisioning servers, storage and network devices are complicated and can be a challenge for many organizations. Put another way by Nathaneal Jean-Francois, a senior network architect at NS1: "A lot of the things you have to build yourself, but it's not impossible to do."

Sounds like a headache. But given today's economy, a necessary evil.

To stay relevant in a rapidly changing market that has a propensity to shift workloads to the public cloud, organizations of all sizes need a more efficient way to mimic the public cloud provisioning models.

Organizations struggle to provision hardware for a variety of reasons.
Organizations struggle to provision hardware for a variety of reasons and new provisioning methods are required.

Organizations can do it by leveraging open source technologies, open standards and creating a layer that abstracts infrastructure into a series of easy-to-navigate graphical user interfaces (GUI) and advanced and open APIs.

World Wide Technology has come up with a solution to tackle each of those components in its Bare Metal Provisioning System (BMPS), a multi-server solution that is engineered as an appliance that can be placed directly in a data center and will start to discover assets that respond to a common management standard called Redfish.

Redfish is an open and agnostic API standard created by the Distributed Management Task Force (DMTF) to allow broad consumption of data center assets using a standard schema that can be readily adopted by any major original equipment manufacturer (OEM). This is a significant change and departure from vendor lock-in products that mandate the use of their tool to manage a server, storage or network device. And this development is important because it allows operators (DevOps teams) to use a single automation routine across multiple OEMs with minimal refactoring of the scripts, etc.

Start at the bottom

Software is important, but hardware should be considered the foundation of your house. If it's not stable or built in a reliable and consistent manner, the entire structure is fundamentally unsound. 

This is where BMPS comes into play, as it creates a standardized and automated process for machines to be provisioned and consumed by the organization — a repeatable process that can be counted  every single time for every single machine.

BMPS enables users to select specific stacks and operating systems of their choice and provision them across a vendor-neutral ecosystem of hardware using the DMTF Redfish standard. And, unlike other provisioning options, WWT's BMPS is completely hardware-agnostic, giving customers their choice of platforms to use. Specifically, BMPS allows for a: 

  • Common platform that DevOps can manage.
  • Uniform API that crosses compute, storage and network.
  • Unified and fully aware telemetry across hardware and software stacks.
  • Provider design that is modular so products can be added without adding an entirely new system.
  • Simple subscription model to manage all components.
  • Solution that is open source and capable of responding to immediate and broad market needs.
An architectural view of WWT's Bare Metal Provisioning Solution.
An architectural view of WWT's Bare Metal Provisioning Solution.

What makes up BMPS? 

BMPS uses several different components to accomplish this.

A single Red Hat CloudForms server to run Red Hat's Cloud Management platform, otherwise known as CloudForms. CloudForms is a modular software suite that allows operators to build "hooks" into various systems throughout a private and public cloud infrastructure. These hooks usually consist of an API connection into the target device through a RESTful API. After the connection is made, the system gathers as much information as it can about what is inside of the asset. This information can be wide ranging from a simple CPU, memory and hard disk footprint all the way up to virtual machines and containers running in a cluster. 

CloudForms was originally targeted at automating private and public clouds (namely AWS, Google Cloud, Azure, OpenStack and OpenShift). These clouds are considered infrastructure-as-a-service (IaaS) or platform-as-a-service (PaaS) offerings that can be connected into CloudForms via a standard RESTful API connection. WWT devised a method to extend the capabilities in CloudForms to connect directly to hardware in the data center, thus extending the capabilities of the product to manage assets at a much lower level. 

This is critically important to WWT as the consumption model for provisioning hardware in the data center has changed to more of an open commodity model. By adding our own custom provider hook into CloudForms, we can manage common Dell, HPE and Supermicro servers through Redfish. 

The work done between WWT, Red Hat and XLAB — an IT solution provider focused on cloud computing, among other areas — to build this agnostic API facility into CloudForms is referred to as a Redfish Provider. This Redfish Provider is 100 percent open source and is freely available to the public via Red Hat's free version of CloudForms called ManageIQ. 

An aggregator server built from the ground up with the sole purpose of being the central "collector" for all of the Redfish-enabled data center assets. The server sits below the CloudForms server and makes a permanent connection into the CloudForms software via a series of RestFUL APIs and Ansible playbooks. Ansible can automate IT environments whether they are hosted on traditional bare metal servers, virtualization platforms or in the cloud, and can automate the configuration of a wide range of systems and devices such as databases, storage devices, networks, firewalls and many others. 

The responsibility of the aggregator is as follows:

  • Broker connections from CloudForms and Ansible requests (via the Ansible modules built into the aggregator) to the actual hardware.
  • Store all of the server, storage and network node information (CPUs, memory, thermals, power stats, BIOS/firmware versions, etc.) in a centralized database to allow CloudForms to easily grab or tag information as it deems fit and make an accurate selection of the device for the operator when requested.
  • Maintain a system state, which is essentially the aggregator listening and recording all activities on the remote devices. Think of it like a pulse monitor in a hospital room — essential to all components (especially CloudForms) to show how each device it is connected to is performing.
  • The aggregator is the core of the BMPS product and allows you to scale to thousands of nodes and provides an enhanced visibility of an organization's data center assets. While you can provision an asset through CloudForms directly, this is not scalable for large enterprises.

Two network switches bundled into the offering to eliminate the need for organizations to setup their own switching infrastructure for BMPS. By adding these network switches, we can offer customers a complete turnkey solution that requires minimal configuration to bring up to an operational state in the customer's data center. 

The first switch is an Out of Band Management (OOB) switch specifically placed to offer management of the various devices in the BMPS stack as well as OOB management for connected devices. This is considered an isolated control plane, which separated administrative functions from provisioning functions. Having this control plane separated is going to be a requirement for many customers. This OOB switch also has terminal server connections, which will allow operators to make "emergency" connections into devices in the BMPS stack should a major network failure/change occur. 

The second network switch is in place to perform high-speed provisioning tasks as well as server as a control plane for CloudForms and aggregator traffic (databases, back-end API calls, image transfer, etc.). This is completely separate from the OOB switch which is in place for administrative functions only. 

 

Easier said than done

As mentioned, BMPS is completely hardware-agnostic, meaning customers have their choice of platforms to use. This is especially pertinent as more organizations turn to more cost-efficient white box solutions to achieve speed and scale.

But white box can be complicated and requires an integrator to piece it all together

Disaggregating software from hardware allows organizations to realize cost savings by deploying ODM equipment while leveraging the power of software to become nimbler while providing best-of-breed solutions tailored for industry verticals. 

As simple as it may sound, it is not. Organizations desire the value white box can deliver, but typically cannot commit to the labor-intensive processes needed to validate and deploy these solutions effectively. 

That's because building white box solutions is building something new and unknown. Therefore, it's critical organizations skillfully and strategically integrate such solutions to ensure they work as intended when deployed. 

WWT has decades of network architecture experience from end point and customer premise equipment (CPE) to the core network. This allows customers to create effective solutions faster and deploy them at scale so they can more quickly monetize technology investments and more rapidly deliver services to their own customers.

Learn more about WWT's ability to fundamentally transform the service provider industry by helping network operators and other service providers deploy and exploit next generation technology platforms