Cisco's top secret spin-in Insieme Networks is developing a fabric controller for high-performance, high-density leaf-spine data center networks that will enable orchestration beyond the network to the entire data center infrastructure. Plans about Insieme's work were released to SearchNetworking this month by a source close to the company, who asked to remain anonymous.
The fabric controller will make the network much more responsive to the applications it serves -- a strategy Insieme calls "application-defined networking," the source said.
A call seeking comment was made to a phone number on Insieme's website. Insieme did not respond.
Long-rumored to be working on a software-defined networking (SDN) technology that would eventually be folded into the rest of Cisco's SDN portfolio, Insieme Networks' technology more closely resembles Juniper QFabric. Insieme's fabric controller will run on NXOS, the operating system that powers Cisco's Nexus data center switches, the source said. The product will also rely on custom application-specific integrated circuits (ASICs) designed by the company.
Preliminary product plans indicate that the fabric controller could be available in three form factors: a fixed 36x40 Gigabit Ethernet (GbE) device and two modular devices supporting as many as 288 and 576 40 GbE ports respectively. These fabric controllers would potentially function as the spine devices in a leaf-spine architecture with other data center switches.
The Insieme fabric controller will support twelve spines in its architecture with flexible insertion. It will scale to tens of thousands of ports and support more than one million host routes and more than one million endpoints.
"It's a simple unspoken fact that the Nexus 7000 is fundamentally throughput-limited," said Peter Christy, research director at 451 Research. Christy has not been briefed by Insieme, but he has heard plenty of rumors flying around Silicon Valley. "If you look at what's happened to switches between the time [the Nexus 7000] came out and today, you can get much higher port density in a much smaller physical environment now."
In addition to enabling a data center fabric that is much larger than what Cisco can offer today with its Nexus 7000 and 6000 series switches, the Insieme fabric controller will also orchestrate Layer 4-7 services and tie them to storage and compute resources by integrating with orchestration systems such as OpenStack.
Do you know anything about Insieme Networks or any other stealth networking companies?
Insieme will have an application network profile system that will allow customers to rapidly deploy new applications. When an administrator spins up a new application, the Insieme fabric controller will automatically apply policies and services based on pre-configured profiles. Rather than building a service from scratch every time the enterprise needs to add an Exchange server, for instance, the administrator can simply bring up the new server and the network will spin up all the services automatically.
Insieme Networks will also use hardware learning to eliminate the need for multicast traffic in overlay networks that use tunneling protocols such as Virtual Extensible LAN (VXLAN). This technology will enable a penalty-free overlay network that gives customers software scalability with hardware-grade performance. Insieme believes it can free up 80% of a customer's potential network capacity through this technique.
The last Cisco spin-in that Insieme's founders were involved in, Nuova Systems, built much of the technology that would eventually become Cisco's Unified Computing System (UCS). Christy of 451 Research believes Insieme is trying to take some of the unified management innovations pioneered with UCS and extend them out into a more broadly managed system.
"UCS Manager gives you the ability to slice and dice this big chunk of physical capacity into much smaller subunits, but with the ability to do it flexibly and to make those decisions not be permanent," said Eric Hanselman, another research director with 451 Research.
By extending that notion out of a UCS rack and applying it to an entire data center network, one can imagine a system that dynamically applies compute, storage and network resources throughout the data center to applications with service chains tied to them via application network profiles.
Let us know what you think about the story; email: Shamus McGillicuddy, news director.