NNSquad - Network Neutrality Squad

NNSquad Home Page

NNSquad Mailing List Information


[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[ NNSquad ] "Cisco Eyes Software Network Revolution"

Ah, the great circle of software/hardware tradeoffs.

Cisco got its start by building specialized boxes to do routing.  Before that, hosts did their own routing.  (Before *that*, specialized boxes - the IMPs of the original ARPANET - did all the routing.)  What drove the great changeover was that doing routing for the networks then appearing was getting too CPU- and memory-intensive for the machines of the day.  Economically, it just made sense to build speciality hardware that could do the job better and faster.

Today, stock hardware is plenty fast enough, and the necessary memory is cheap enough, that there's little reason to build specialty hardware.  The only exception is for the actual bit-movers at the high end, where hardware switching still wins.  But the decision making doesn't have to be out in that hardware.  So we re-invent the intelligent networking peripherals that existed back when hosts were doing the routing - but couldn't keep up with 10Mb/second Ethernet without some help.

And this design has other precedents.  Old-timers will remember AIN, the Advanced Intelligent Network - the effort by the Telco's to put more smarts into their network starting back in the late 1980's or thereabouts.  (Wags at the time said that it wasn't Advanced, it wasn't particularly Intelligent - and it certainly wasn't a Network.)  AIN kept the old switches as a "data plane" but added an external system - often nothing more than a Sun workstation was plentgy - as the "control plane" to do things like re-route 800 numbers based on time of day.

When you can sell specialized hardware with proprietary software, it's really tough for competitors to get into the market and you can make large profits.  When the hardware is off-the-shelf from any number of suppliers, and all you're selling is integration with software working to published specs - you can make money, but nothing like what you made before.  The real difficulty for companies built to work the old way is that they've generally created business approaches with cost structures that no longer make any sense.  Cisco is right to try to get in on this hardware shift - it's better to cannibalize your own products than to let someone else cannibalize them - but it's going to be very, very tough for them.

                                                        -- Jerry

nnsquad mailing list