Wednesday August 30, 2006

I'm sure most people have heard about the recent problems with the ceiling panels in two of Boston's Big Dig tunnels, which came to light when a woman was killed by a falling panel that crushed her car on July 10th.

What you may not realize is that this has ballooned into a much bigger problem, as described in Scott Allen's article on boston.com. Some 3300 brackets in two tunnels are now deemed to be inadequate to support the tunnel ceilings. This in addition to having to reinforce over 10,000 epoxy-secured bolts after a number in each tunnel were found to have come loose. The Massachusetts Turnpike Authority is now expecting the repairs to total $15M rather than the initially estimated $3M.

My concern is how the Turnpike Authority and the State of Massachusetts are going to recover the confidence of the public in this matter. Given the widespread nature of the failures, I don't believe that reinforcing the brackets, replacing the bolts, reinstalling the ceiling panels, and then declaring success will be a very effective argument.

Instead, I would like to see active sensors installed on the ceiling panels. Sensors that can record displacement, or stress, or whatever the engineering teams deem to be the most effective metric to track the health of the tunnel infrastructure. Powering the sensors should not be a problem as I understand the ceiling panels themselves form the bottom side of an air plenum that carries a high-velocity exhaust stream out of the tunnel system. Surely this air stream could be used to power a set of wireless sensors that can be queried and monitored remotely for any future problems. To be totally transparent, the real-time state reported by the sensor array could be available on a public website for all to examine.

Yes, this will add to the cost of the repair. But an active monitoring capability would go a long way towards assuring the public that we will not see a repeat of the tragedy of July 10th.

The Open MPI community has just released version 1.1.1 of the Open MPI library. The changes in this bugfix release are described here, and the source code bits can be found here.

As previously announced, the next version of Sun's supported MPI library for Solaris will use Open MPI rather than our own, proprietary implementation. Our engineers are busy working on both general Open MPI improvements as well as improved support for Solaris in particular.

In the meantime, v1.1.1 is compilable and usable under Solaris 10 on both SPARC and x64, though it currently lacks support for Solaris Infiniband. And this release is supported through Open MPI's community forums rather than directly by Sun.