Reply #44: It was bizarre as i recall

[sweetheart]

I was out in california working a contract at Tandem computers at the time, having
just finished a contract doing systems and performance tuning for NASDAQ's SOES
(small order execution system), and the MSI queues for the main order match.
"order match" is exchange parlance for "trade"... when a buy and a sell order "match"
in logical criteria... the actual software system is a small expert system that compares
these criteria in real-time streaming out various data streams as a result, some public,
some private.

Well, i called the computer operations center at nasdaq to find out how the systems
performance was working. At the time, data files on super-mini's were much like mainframe
ones, with certain file-system allocation requirements. One of these was called "maxextents",
and was related to the maximum number of non-contiguous data blocks in a file. When a file
reached max-extents on a running application, it had to be shut down.

WEll, the order-queues in to nasdaq were reaching max-extents, and when they were shut down,
the fault tolerant queue managers lost the order book (means all open orders)... and the
exchange shut down. I had allocated the extents based on how managememt designed the
fault tolerance of the exchange, which was to provide 2 times as much file space as ever
previously used by the exchagne in 1 day... well, that theory failed the exchange that
day.

So i recall that day, standing there in sunny silicon valley a but bummed out that the
performance tuning project failed under duress.