Every investor is familiar with corrections in equity valuations, but P/E ratios aren’t the only metric that undergoes mean-reversion on the stock exchanges. Last wednesday, the statistical availability of the exchange’s data center underwent a correction. The event, caused by an improperly executed software upgrade, grabbed headlines and drew criticism on CNBC because it was the longest duration outage due to technical issues on record. To be sure it was a major event, “a bad day” as the NYSE president reported from the exchange floor. However, when taken in context of the cumulative availability over the last 30 or so years, this event should surprise us no more than any of the routine 10% corrections we see in the stock indices.
Availability, in data center parlance is simply the fraction of time that the center is working right. In the data center infrastructure business, the nearly unachievable holy grail of availability scores is called “five nines”. This corresponds to about 5 minutes of unplanned downtime per year. It is easy to do this in some years, but averaging five nines over the long term is a tall order for even world class data centers with unlimited budgets. Note that we even see occasional outages from the likes of google, apple, yahoo, etc. The NYSE hit this elusive target briefly in 1989, then plummeted back to a still respectable four nines after a con-ed transformer exploded and brought down the exchange in 1990. Up until last week’s 4 hour “glitch” the availability had been creeping back up, reaching .99997 before the correction. Viewed in this cumulative manner, one should walk away confident that the exchange is still very reliable and becoming more so over the long term, despite this latest lengthy outage. Traders should be comforted by the similarity of the above graph to the stock charts they are always performing technical analysis on. I’ve yet to see any creative derivatives that allow us to make wagers on the next outage at the NYSE, but I’ll be watching, maybe even wagering.