New Spin-In Called Insieme: A Framing Exercise

22/03/2012

The following are my thoughts on the NYT report that Cisco was creating or supporting a new startup called Insieme.  I have decided to use a framing format for this posting, because I really do not know what Cisco or the Insieme team is thinking.  It is not as if they asked me to review their plan.  All I have done in this posting is look at a number of data points and draw two conclusions.  The conclusions are at the bottom of the posting if you have spent too much time using Twitter and have a diminished ability to focus on reading.

Frame 1: What is a Spin In? Back to the 90s and Ardent

Ardent Communications was founded in 1996 as a planned spin-in startup focused on building a specific product designed for Cisco Systems.  The spin-in model was a method to create a startup that could attract top talent that, in turn, would build a product defined by Cisco in short period of time.  To capitalize the new company, Cisco invited Sequoia Capital to participate in the funding of the new venture.  To foster the entrepreneurial atmosphere, the employee and management pool was sized at 55% [see Cisco Systems: A Novel Approach to Structuring Entrepreneurial Ventures, Stanford University Graduate School of Business, Case Number EC-15, February 2000].  The trigger mechanism on the deal was a put/call feature that gave Cisco the option to purchase the company at a specific time or upon a product deliverable at a specified price of $232.5 million in cash or stock.  The first round of funding for Ardent closed on August 11, 1996.  Less than a year later, on June 11 1997 Cisco announced the acquisition of Ardent Communications for $156 million, which does not includes the 32% of the company already owned by Cisco bringing the total transaction value to $232.5 million.  Ardent was a made to order M&A transaction.  It combined venture capital and a publicly traded stock to build a product outside of the corporate structure and process of Cisco Systems.  The real beauty of transaction was the minimal risk incurred by Cisco and Sequoia Capital.  Unless the team failed to develop the product, it was a guaranteed exit strategy using M&A.

Frame 2: Nuova Systems and Post from October 23, 2011

I wrote a post about Arista, Cisco and Nuova back in October 2011.  Here is the relevant portion of that post.  “To understand what I am thinking about, we need to go back to Nuova Systems.  Cisco acquired the company in April 2008.  From this company emerged the Nexus 5000 class set of products.  What was unique about the company was the funding and the impact it has had on others in the Cisco ecosystem.  Think about where Cisco stood in 2008.  For the FY2008, Cisco sold $13.46B in switching products.  (Note that Cisco’s fiscal year is August-July).  FY2008 was the peak year in Cisco switching revenue before the onset of the global credit crisis in September 2008.  I have detailed the 2009 catch-up spend event here and here, but consider the world before the credit crisis induced spending collapse and subsequent catch-up phase.  Cisco was the dominate market share owner in switching.  There was no one close and they had spent their acquisition capital from the start on this market segment.  This is a just a partial list of Cisco’s switching centric deals starting the most important being Crescendo in 1992.  This list alone totals $6.4B:

24-Sep-93	Crescendo Communications	LAN switching	$94,500,000
24-Oct-94	Kalpana	LAN switching	$204,000,000
8-Dec-94	LightStream	LAN switching	$120,000,000
6-Nov-95	Grand Junction Networks	LAN switching	$220,000,000
6-Aug-96	Nashoba Networks	LAN switching	$100,000,000
3-Sep-96	Granite Systems	LAN switching	$220,000,000
28-Jul-98	Summa Four	LAN switching	$116,000,000
31-Aug-99	IBM Networking HW Division	Computer networking	$2,000,000,000
11-Apr-00	PentaCom	LAN switching	$118,000,000
20-Aug-02	Andiamo Systems	Datacenter Switching	$2,500,000,000
29-Jun-04	Actona Technologies	Data storage	$82,000,000
8-Apr-08	Nuova Systems, Inc.	Datacenter Switching	$678,000,000
			$6,452,500,000.00

Nuova was started in mid-2005.  It was self-funded by the Cisco executives who left to form the company. This list included: Mario Mazzola, Luca Cafiero, Prem Jain, Soni Jiandani as well as outsiders Ed Bugnion and Tom Lyon.  In August 2006, Cisco increased their ownership to 80%.  In April 2007, Nuova had ~200 employees and Cisco agreed to a cap of the potential buyout of remaining 20% at $678M.  A year later, April 2008 Cisco closed the deal to acquire the remaining 20% for $678M.  No venture capital firms were involved to my knowledge.  Jayshree Ullal left Cisco in May 2008 and joined Arista in 2009.”

Frame 3: Google Search Terms

After I wrote the Arista post in October 2011 and mentioned the Nuova team, my blog started getting hits from people searching for the team.  Here is a the summary of those search terms and views:

Search Term	Views
mario mazzola luca cafiero	22
luca cafiero	21
luca cafiero new startup	15
mario mazzola new startup	9
prem jain startup	7
mario luca prem	7
mario mazzola	5
luca cafiero cisco	4
mario mazzola startup	4
luca cafiero startup	3
mario mazzola 2012	3
prem jain OpenFlow	3
mario prem luca	3
	106

Frame 4: Twilight in the Valley of the Nerds

I am a firm believer in providing credit where credit is due and citing sources.  The blog Twilight in the Valley of the Nerds was one of the first sites, if not the first site to post about a new start-up brewing.  You can find the posts here on July 28 2011 and the most recent posting here.

Frame 5: Why a Spin In?

Before I present what I think Insieme is building, a few short thoughts on the idea of a spin-in.  My first reaction to the NYT article was: why is this not being done inside the traditional constructs of the corporation?  In the post-Madoff age of private capital ventures, the accounting and valuation process is much more scrutinized.  I know auditors of PE-VC funds are increasingly demanding support and documentation for valuations.  It has been reported just last month that the SEC is looking into how valuations are marked and the conflicts within investment pools.  If we assume that the Insieme product can be built for $50M and that it would take 24 months, what is the value of the corporate entity and why is it better for Cisco to build a product outside of the company?  I think I can answer my own question.  JNPR started QFabric in mid 2008 and it took a lot longer to bring the solution to market than planned, sometimes doing an engineering project outside the company is a better plan.  When I worked at CrossComm in the late 80s early 90s, we took a core team out of the company to build the company’s next-generation product.  We put them in a building a few exits up the highway where they would not be distracted by the day-to-day events in the legacy product lines.  I can understand how internal resources can be distracted by present day activities; that is something I have blogged about over the past several months.  The question that John Chambers has to ask himself is what is the appropriate price to pay for the asset.  I am going to assume that there will be pressure on the valuation of Insiemi compared to the valuations used for Andiamo and Nuova and therefore I think 10X on an assumed $50M investment means a $500M exit for the team if successful.  These numbers are assumptions on my part; feel free to insert any number for the capital to build a product.  Maybe it is $35M or maybe it is $100M.

Conclusion: What I Really Think they are Building

I received a good comment over the weekend (much of which I agree with) on what CSCO has accomplished in the switching space.  I think there are two paths that the Insiemi team could be going down.

Path 1: Assume there is unfinished business in the storage space, which is why I linked above to the comments over the weekend.  I think it is possible that the MLP team would go build what would become a new Nexus family member with storage network hooks.  This would be Cisco’s strategy to abandon FCoE as a convergence strategy.  DG at Arista already opened this discussion last year.  The new Insiemi box would use a similar approach to Xsigo’s product strategy, but instead of using Infiniband, the Insiemi box would be all Ethernet, but have two separate Ethernet channels (i.e. forwarding planes).  One Ethernet channel would logically allocate bandwidth vertically and be intended for traditional IP traffic deployed in hierarchical switched data networks.  Basically what the world has been building for the last fifteen years.  Nothing new in the first Ethernet channel, it would be the second Ethernet channel that would be disruptive.  This second channel would be for SAN networks and instead of being hierarchical in nature – they would scale horizontally across the compute element.  This second channel would run Ethernet down to the host (N2K?) and split out storage traffic from data traffic.  Maybe this Insiemi box acts like a storage bypass in front of the N7K.  Add in some fibre channel ports and you have a Director killer.  Coincidently you would also have a product designed to force Arista to live in their low latency and cloud market applications and force Big Switch and Nicira to address a scaling challenge that in turn becomes a capital spending challenge.  Path 1 is a strategy to box the competition into a small addressable market space.  Attacking the storage space provides a number of benefits for Cisco.  Once data is placed on a drive (does not matter what kind of drive) it becomes physical in nature.  Facilitating the connections between storage and the compute element would accelerate the demise of FC.  FC will be around for sometime, unless someone helps speed the decline.  Building a transition switch and using SDN like features to facilitate SAN networks (i.e. NAS) while keeping the traditional Ethernet flows separate is a strategy to address some part of the big data market.  Path 1 is really an extension of the current Nexus strategy, which was switching, then added compute, and therefore the next step is to add storage integration.

Path 2: This is the SDN/OpenFlow strategy path.  From this Network World article, we know that the Insieme team has licensed or has access to NOS (Nexus OS).  Here are several links to various postings about Insieme and Cisco’s OpenFlow direction:

– Twilight in the Valley of the Nerds 10.22.11

– Twilight in the Valley of the Nerds 03.16.12 (read comments)

– David Meyer presentation at ONF October 2011 (Is slide #14 a tell?)

– Video of David Meyer presentation at ONF October 2011

– Read “no where fast” comment about Insieme on Cisco blog

Taken in totality, I can envision a new platform that acts as a NOS to OpenFlow arbitration box.  The platform would connect to the Nexus OS (hence the license) and it would have a set of interfaces for connections to the OpenFlow network.  Cisco would use SDN like concepts to control the translation to and from the NOS network from the OpenFlow interface.  Think of it as a session border control (SBC) for your network between the NOS network and the OpenFlow network.  Cisco keeps their installed base closed to direct OpenFlow control, yet provides an OpenFlow portal in which they control the translation and arbitration function between the two networks.  The plan in Path 2 is to keep Cisco market share locked in, but tell customers that Cisco has provided an OpenFlow access point and will plan to build SDN products using NOS and OpenFlow.  I suspect that this would be a complex, software intensive platform, but would be valuable to Cisco as it would provide them a story, a migration strategy to convince customers to keep the status quo, upgrade their N7Ks and wait till the next upgrade cycle for OpenFlow as Cisco will have a SDN/OpenFlow like platform built to Cisco standards.  Does anyone remember Ipsilon Networks?

I really have little idea of what I am writhing about and it quite likely that the Insiemi team is building a completely different product than the two possibilities I have describe in this post.  I just thought it was interesting exercise to consider some development strategies if I was in their position.

/wrk

* It is all about the network stupid, because it is all about compute. *

** Comments are always welcome in the comments section or in private. **