The cause of the technical difficulties of the stock / convertible bond trading system is outlined below.
Circumstances
First, I would like to once again explain what the circumstances were that day.
When the stock / convertible bond trading system (below "trading system") was activated at 6:30 that morning, an error occurred during the reading of files that stored data relating to trading participants, causing the trading system to stop. When the information read should have been trading participant data but was not, the system detected this aberration and ceased operations.
At that point, TSE and Fujitsu, Ltd. (below "Fujitsu") - who is responsible for development and support of the trading system - immediately began investigating the cause of the problem, and attempted to restart the system. Upon restarting, however, the same type of malfunction occurred, and the system once again locked up.
As a result, start-up of the system before the trading session began at 9:00 a.m. proved difficult, and at 8:30 a.m. TSE decided to suspend trading for the time being, notifying all relevant parties.
At around 12:00 p.m., the problem area was ascertained, the incorrect
information was deleted, and the trading system was restarted. The
correct information was read, and trading resumed at 1:30 p.m.
Cause
We have identified the cause of the system malfunction to be as follows. During the recent augmentation of the order processing ability of the trading system, a latent bug that had previously existed in a program connected with trading participant data files was discovered. As such, after confirming with TSE, Fujitsu made a correction to the relevant areas on October 9th, and on October 13th the corrected program was loaded as normal into the trading system. During this process, part of a necessary step was missing in the instruction for the loading operations .The instruction was sent from Fujitsu to Tosho Computer Systems Co.,Ltd , who was charged with the loading operations . As a result, the structure of the system after normal loading was incorrect, having a combination of both the normal program and the old program, and the system malfunction occurred during the disk compression process on October 31.
The reason the malfunction didn't occur from October 14 to October 31 is because the old program still existed and was operating normally during that period. The reason the malfunction occurred on November 1st is because during the disk compression process, the system automatically detected the existence (attributes) of the old program, and disconnected with the normal program.
We deeply regret the great inconvenience that we have caused investors
and other related parties. As for the fact that the cause of this
system error was an omission in the instruction used during the
normal installation process, we are urging the maker to discover
the cause of the omission. I also believe that we must verify whether
or not we had strictly required that the maker possess a prevention
system for malfunctions of this sort up until this point, and also
whether or not we were able to detect the mistake in the instruction
that the maker produced. I also believe that it is necessary to
implement new countermeasures for sustaining market functions when
there is a system malfunction. All employees and executives of TSE
are committed to exerting their fullest efforts to preventing this
type of situation from reoccurring in the future, and we ask for
your understanding and support in this matter.
Takuo Tsurushima
President & CEO
Tokyo Stock Exchange, Inc.
