Archived Content

The following content is from an older version of this website, and may not display correctly.

I have to admit, I've been impressed with the speed at which Cisco eventually responded to software-defined networking (SDN). I'm not endorsing Cisco's Application-Centric Infrastructure (ACI) or denying that Cisco tried to ignore SDN at first. I'm just saying that, for a big company that's deeply ensconced in one way of thinking about routers and switches, Cisco moved pretty quickly to try to overtake the conversation, doing things like helping create OpenDaylight.

Those moves will culminate on Wednesday, Nov. 6, with the launch of Insieme, the latest Cisco spin-in founded by Mario Mazzola and his team, after their Nuovo and Andiamo ventures. It's the most anticipated news event of the year for SDN.

Varying theories about Insieme have spread during the year. It could be an ovearching SDN-like architecture, or it might be just the next huge Nexus switch with some virtualization features attached.

Here's what we think we know. Your mileage may vary.

It is not overlay-based.
This one's pretty obvious. You can see a Cisco-produced video where Insieme Senior Vice President Soni Jiandani says overlays "don't remove the problems — they just move the problem." You can read a blog by CTO Padmasree Warrior saying a "software-only approach" doesn't scale and robs operators of visiblity.

I don't think anyone is surprised that Cisco doesn't like overlays. The overlay model can mean sucking all the intelligence into the SDN controller, leaving your switches and routers as brainless zombies. Even around Halloween, Cisco doesn't want its franchise equipment dressed up as brainless zombies.

It uses ASICs.
Whether or not they're late, as rumors have it, the ASICs are a central part of the plan. Insieme is also using merchant chips, probably from Broadcom, in some capacity. Non-ASICs have been creeping into the portfolio, as the economics of using them are just too good (and the availability is just too available). But even as merchant chips close the gap with ASICs, Cisco executives are sticking to their guns about ASICs being the guts of any high-end system. They have to.

There's more to Insieme than just one big box.
Here, we're getting into guesswork territory. At Interop early in October, Cisco CEO John Chambers sure made Insieme sound like a far-reaching concept, possibly even touching the access network. (I wouldn't be surprised if that part was just a verbal hiccup.)

... but it's going to be a big box.
There is no way Cisco's home-built silicon photonics get left out of the Insieme announcement. I'm expecting Cisco to play up density big-time, trying to outshout Arista on that count.

There's also the fact that those Broadcom Trident II chips, which is what we all assume Cisco is using for Insieme's more mundane line cards, create big density at 40 Gb/s, as we've already seen in announcements from Juniper and Cumulus. (See here and here.) By definition, anybody can match those numbers; I would guess Cisco will announce bigger numbers, just to be Cisco.

The White-Box Elephant

For me, the biggest question around Insieme is the competition from white-box switches — which Chambers has acknowledged as a threat, rather than a phantom fad. Network World has posited that Cisco's new Nexus switches will be priced to go after Arista, which is a more immediate threat. Generic switches keep strengthening their case, though; it seems only a matter of time before Cisco has to confront the issue head-on.

As for the traditional competitors, Juniper gave itself the opening argument by announcing MetaFabric on Oct. 29, and Arista has scheduled a news event of its own for Monday, Nov. 4. Then Cisco gets its turn on Nov. 6.

(Image: Part of a motivational poster created for Insieme's team. Source: Flying A Design.)