At Big Switch Networks, we are excited to announce on-schedule delivery of our flagship product – the Big Cloud Fabric (BCF).  It is the world’s first data center networking fabric built with bare metal switches and SDN software.  BCF was unveiled during our July launch while in beta, has secured $1M pre-FCS orders to date and is now shipping to customers.

The BCF product is designed for new pods in existing data centers, typically part of private cloud, big data, and VDI projects.  With the shipping BCF release, customers can build 2-rack to 16-rack pods.  Each pod can have up to 16K VMs or 48K IP end points.

Big Cloud Fabric – Simplicity is Job 1

Big Cloud fabric – a leaf & spine Clos fabric – is architected to function as a “single logical switch”.  Leaf switches behave as logical line cards, Spine switches as logical fabric modules and controller cluster as redundant supervisors. Big Cloud Fabric leverages our Switch Light Operating System on physical leaf and spine bare metal switch hardware and supports the latest Trident II silicon from Broadcom. BCF is designed for 10G and 40G scale and resiliency, featuring headless mode high availability, zero-touch fabric operations, L2/3 forwarding options, application-centric policy, and service chaining. Of course, it supports multi-hypervisor environments (vSphere, Hyper-V, KVM, and Xen).  Management is available via OpenStack, Cloudstack, REST, CLI, or GUI.


BCF’s simplicity comes from its logical chassis operational model.  In a traditional chassis switch, the admin never logs into a line card nor upgrades it; all line cards and fabric modules are completely controlled and managed by the chassis supervisor.  Similarly, BCF (being a logical chassis) operates as a Zero-touch Fabric; its leaf and spine switches are completely controlled and managed (i.e. zero touch) by the BCF controller – this includes OS installation, device provisioning, fabric formation and OS upgrade. 

Additionally, the entire application/tenant provisioning model is logical.  Gone are the days of configuring a switch port and VLAN and SVI and spanning tree and link aggregation across multiple tiers. Instead, BCF uses application/tenant centric configuration model, leveraging logical segments and logical routers constructs; the controller takes care of configuring individual switches and ports/interfaces.  Each BCF device can be an L2 switch or L3 router – hence packet forwarding is completely distributed.  No more complex provisioning across L2/L3 boundaries and managing unwieldy spanning tree protocols. Net-net, there is massive reduction in device consoles required for day-to-day management.  Even troubleshooting is simplified as the BCF controller provides full fabric-wide telemetry and multi-device packet trace. For OpenStack, BCF provides integrated workflows for networking/security/audit teams, with enhancements built into the Horizon GUI.

How Are Customers Purchasing & Deploying BCF?

Similar to the hyperscale-style BCF architecture, we are also bringing simplicity and flexibility in the purchasing of BCF and post-sales support experience.  Customers have a choice of purchasing HW & SW (1) together from a system integrator/reseller or directly from Big Switch or (2) separately across Big Switch (for BCF software) and HW switch suppliers.  Regardless of how the solution is purchased, customers call a single support number for the entire fabric (for hardware and software).  SW pricing is subscription based (per annum) inclusive of license, SW upgrade and technical support.  Additionally, 3- and 5-year subscriptions are available to customers who prefer to purchase up front with higher discounts.  In typical designs, customers realize CapEx savings of over 60% compared to incumbent vendor solutions (based on HW + SW + support costs over a 3-year duration).

In terms of deployment, a BCF leaf/spine pod is built with a row of server racks and connected to existing Layer-3 core.  This is often referred to as Core-and-Pod design where pods from multiple vendors or of different use cases are connected to the existing core network. With this BCF release, customers can build 2-rack to 16-rack pods and up to 16K VMs per pod.  (Note - this is our current QA validated scale; architecture can max out Broadcom Trident II scale.)

BCF Starter Kit

Ready to kick tires on hyperscale-style networking?  With the BCF starter kit, we are making it super easy to implement a redundant 2-rack pod.  At $99K (US list price), you get: 4 leaf and 2 spine HW switches (10G/40G), redundant controller appliances, 3-year BCF software subscription, 3-year hardware and software support, and 40G cables.

Welcome to hyperscale networking!

 --Prashant Gandhi, Big Switch VP of Product Management & Strategy


Additional Resources: