Modern data centers demand software-centric infrastructures across compute, network and storage for speed, simplicity, flexibility and economics.  A new class of hyperscale-inspired applications, such as HDFS, Cassandra, Storm, Spark, Jenkins as well as a new application form factor – containers – require rapidly deployable, logically isolated infrastructure that natively supports application elasticity.  DC/OS – an open-source cluster-level OS leveraging Apache Mesos – enables a modern design for computing by constructing one logical (super) computer from standards-based servers. This is particularly appealing because at Big Switch we leverage concepts of software-defined networking (SDN) and CLOS fabrics to construct one logical (chassis) switch from standards-based open networking switches.  The combination of DC/OS and Big Switch’s Big Cloud Fabric (BCF) is optimal for designing modern data center clusters (or pods), while ensuring attractive economics and complete vendor choice for both hardware and software.

DC/OS + Big Cloud Fabric: Architecture

Figure 1 depicts the logical and physical perspective of DC/OS computer and Big Switch network.

DC/OS is ideal for modern applications as it is quite simple to deploy, operate and scale.  Change management is also very simple and quick.  Servers can be added or deleted from the DC/OS cluster at will.  Once a cluster is deployed, all of the cluster’s HW resources are pooled and available to applications. The cluster can be logically partitioned to provide isolated environments for tenants and/or applications.  When deploying an application, DC/OS automatically assigns servers and allocates precise HW resources (CPU, memory, storage) on each server.  An application can be scaled up or down with a single CLI command.  Additionally, DC/OS is a complete solution for deploying hyperscale-style computing environment in normal-scale enterprise and telco data centers.  Everything you need comes ready to run, out of the box. And because it is based on Mesos, it is battle-hardened.

Big Cloud Fabric is a production-grade, high-performance (10G/40G) and resilient leaf/spine switch fabric built with software-defined networking (SDN) principles and open networking switch hardware.  It implements a hierarchical control plane, mostly residing in a centralized controller with residual components placed on switches to ensure scale and resiliency (including controller-less operation).  The BCF Controller acts as a single pane of glass for full operational and change management of the entire fabric.  In principle, BCF constructs a pay-as-you-scale chassis switch, which can grow from 2 slots to 40 slots to support a cluster (or pod) of 1250+ servers (even larger sizes are possible with newer switch hardware).

DC/OS + Big Cloud Fabric: Unique Benefits

DC/OS and BCF are highly synergistic not only in terms of architectural principles but also the benefits they provide to data center customers. 

DC/OS + BCF also offers very unique benefits.  For example:

  • Server detection & link configuration: As new servers (hosts) are added to the DC/OS cluster, no configuration is required on Big Cloud Fabric.  The BCF controller detects new servers and automatically configures server-leaf switch links (single leaf or redundant leaf).  Servers are also visible in the BCF Controller, along with their attributes such as MAC/IP addresses and host-name.
  • App Elasticity: As DC/OS schedules and scales/shrinks applications across various hosts, there is no network configuration required.  BCF’s tenant-centric architecture, logical switch architecture ensures that network automatically supports applicaiton elasticity (no manual provisioning is required).
  • Visibility: DC/OS hosts are fully visible in the BCF controller, and are also visible through built-in fabric analytics based on (hyperscale-inspired) ELK stack (Elasticsearch, Logstach, Kibana).  All network-related events (including link failures) are captured.  Months worth of log and event data is stored on the controller and is searchable through sophisticated filters.
  • Trouble-shooting: DC/OS application traffic across Big Cloud Fabric leaf and spine switches can be traced across source-leaf / spine / destination-leaf, including ingress and egress switch ports – this significantly reduced time to resolve connectivity issues and issue isolation across compute and network

Video demo of DC/OS and BCF provides further details on the joint deployment and its unique benefits.

DC/OS + Containers + BCF

With the emergence of DC/OS and SDN fabrics, we now have a cohesive approach to deploying modern applications as well as container environments (such as Docker and Kubernetes) at scale and with attractive TCO.  Over the next few years, we will see a massive shift away from complex and brittle legacy environments to modern, transformative designs for compute and networking.

To experience how BCF simplifies networking for Docker containers and provides container visibility & trouble-shooting, check out our BCF + Docker video demo.

How to Get Started?

It’s easy and free to get started with both DC/OS and BCF.

  • DC/OS is open-source and can be downloaded from  All you need is few servers.
  • Big Cloud Fabric Community Edition is also free and can be downloaded from  All you need is one or two open networking compliant switches.  To try BCF on-line, sign up for our free Big Switch Labs and test drive variety of use cases.

Prashant Gandhi

VP, Products & Strategy

Additional Resources: