As I spend more time learning about Networking, I begin to realize that there are significant disruptions happening and they are not as obvious as “SDN” or “Open Flow”.
One of the most significant differences between a networking device and a server is that the data path stays entirely inside of a custom ASIC instead of running on a general purpose multi-processor.
Why is this important?
When a customer buys a device, what they are buying is the data path, the rest of the product is a second order purchasing decision. In a networking device, the data path is entirely in the hardware, and therefore – although software is important – it’s significantly less important than the ASIC that processes the packets.
A networking company – for all of the discussion about software – is a hardware company that builds an ASIC that processes packets. They live and die by the success of their hardware and by hardware I mean ASIC.
Historically a networking company built it’s chips and sold a device.
The disruption that is happening in the networking space is that – increasingly – a large chunk of the networking ASIC’s are being built for increasingly important segments by vendors that sell to multiple networking companies.
This is very similar to what happened in the late 90’s when Intel started killing off all of the custom microprocessors in the server market.
A simplistic and silly view is that as a result of this disruption is that everyone is going to go and build white-box switches and routers all of a sudden. For some customers wh have the depth of skill and expertise to build their own networking gear this will happen but for the same reasons I am bullish on storage I am bearish on this happening broadly. The majority of the market will continue to buy gear from networking companies.
The disruption that Intel created in the 1990’s was that companies like SGI that built everything soup-to-nuts had to suddenly go from hardware companies that had software to a systems company.
Customers used to buy SGI hardware to get access to the massive number of processors or the high end graphics engines and they didn’t really care about the software running on them too much. The selling point was the hardware.
As the number of vendors that could deliver the hardware that SGI produced customers went from buying whatever hardware SGI produced running whatever software SGI put on the hardware, to thinking about the whole system and that created more options for the customers.
In effect, the decision criteria went from being what SPECint or SPECfp a MIPS processor had to what is your TPC-C and TPC-D number and whether you were a key partner’s of Oracle.
So what is a systems company then?
A systems company delivers a device that is balanced between hardware, software and packaging. The balance of the three components creates a unique selling proposition that the customer and market is willing to put a premium on.
Systems companies emerge in markets where the hardware is supplied by a small number of vendors and differentiation is created in how you combine the hardware and use the hardware in software and create a package that the customer can buy.
The challenge for systems companies is that sophisticated customers can build the same product from the same components and this creates a pressure to innovate in lots of areas outside of the core components.
The challenge for a hardware company as it makes the shift to a systems company is that the mindset of how you build a product has to change. Whereas in the hardware centric world – the hardware is built and the software guys have to figure out how to make it work, in a systems world there is a fine balancing act between the two. And whereas in a pure hardware world the hardware performance was the be-all of your differentiation, the systems company has a combination of attributes in software, packaging and hardware that create the unique differentiation.
In fact, perversely, how you build software and the easiest way to build software becomes more important over time than any specific hardware platform. And the choice of components becomes more important than any specific component.
The challenge for networking companies is that system design is not where their core focus has been over the last 20 year.s
This change in how you build networking devices from – here’s an ASIC with software to here’s a complete balanced and well integrated system is going to be disruptive to how companies do business.
Now that I said all of that, let me observe, I don’t think my employer, Juniper, or Cisco are in danger of getting disrupted per-se. I just think that the way systems will get built is going to be very different and that that change is going to be very interesting.
And that in my mind is where the real disruption in networking is happening, everything else may just be noise.