Here Comes Trouble: MIPS For Broadband

In order to get its competitive juices flowing again, infotech needs a metric quantifying the price performance of broadband. Before the Internet crashed the party in the 1990s, the infotech industry spent a decade obsessing about the metric known as MIPS — Millions of Instructions Per Second (or, according to some, Meaningless Indicators of Performance). Digital Equipment’s $100,000 VAX platform, which sported a single MIPS, dominated the landscape circa 1980. MIPS comparisons, albeit flawed, served as the basis for a competition that DEC lost to Intel and AMD.

Given that the average talking Hallmark greeting card probably consumes more than a single MIPS and the $600 PC used by a typical 16-year-old to post videos on Facebook sports an Intel Core Duo chip capable of 27,000 MIPS, the tracking of MIPS no longer generates much excitement.

While the availability and performance of connectivity represents the primary input gating progress today, the hunt for broadband still lacks an equivalent metric.

The telco-influenced FCC (laughably) defines broadband as anything capable of 200Kbs in one direction. By way of comparison, Cisco-backed TechNet proposes a goal of 100Mbps by 2010. The engineers at DEC likely viewed 100 MIPS as ambitious in 1980, but static goals (whether anemic or ambitious) will not produce continuous industry growth. Rather than trying to define the destination, MIPS-like metrics provide a measure of progress. Growth in the infotech industry follows continuous progress in processing power, memory, storage and various other inputs, not an arbitrary goal.

The telco bandwidth mavens will strenuously resist attempts to frame connectivity as a commodity like processing power. Telco CEOs subscribe to the Carl Sagan school of counting in which deploying broadband involves “billions and billions” of dollars. However, compare the $20 billion Verizon proposes to spend deploying fiber to 20 million homes to the $200 billion SBC CEO Whitacre spent cornering the market for copper infrastructure via acquisitions of Pacific Telesis, Ameritech, SNET, and BellSouth. CEO Whitacre could have accomplished far more using his billions to connect every home in the U.S. to fiber.

The continuous improvement challenge depends on the 20 percent of costs associated with termination equipment. The labor-intensive deployment of outside plant infrastructure represents a fixed cost with an upgrade cycle of a decade or more. Equipment vendors already deliver Moore’s Law-like annual cost performance improvements, but there exist insufficient competitive forces to motivate upgrades.

The success of Intel over the years provides a counterexample to the argument that commodity status destroys the prospects for profit. In fact, the present telco vs. cableco connectivity duopoly might operate in the same way as the Intel vs. AMD duopoly that arose out of the hunt for MIPS dominance.

Verizon’s relatively weak traction with FiOS video illustrates the risks of trying to fund infrastructure by selling services; it should take a cue from Intel, which pushes the risk of finding the killer app onto others. But a change in sentiment among telecom CEOs will prove slow in arriving. Abundant broadband enables VoIP apps and the next YouTube that compete with the legacy voice and video services generating a majority of telco and cableco revenue.

Change will require the infotech industry to get directly involved in the delivery of broadband. Creating the connectivity version of MIPS can provide a start. Maybe infocom can repurpose the MIPS acronym to make it relevant for broadband – M______ I_______ P_______ S_______? Any suggestions? Please send them in.

loading

Comments have been disabled for this post