From: "Robert I. Eachus" <rieachus@comcast.net>
Subject: Re: Use of Ada in the NIF
Date: Thu, 29 Jan 2004 20:51:45 -0500
Date: 2004-01-29T20:51:45-05:00 [thread overview]
Message-ID: <bJWdnRzgDcYuJYTdRVn-uw@comcast.com> (raw)
In-Reply-To: <4b0b52a9.0401291710.69a8f786@posting.google.com>
Gioia Tauro wrote:
> "Since control room operation started, the rate has been at its
> highest, exceeding 300 per month (385 SCRs were filed in October 2002,
> the last month for which data are available). The high rate is
> attributable mostly to the intense scrutiny due to multiple test
> cases, and to the first-ever
> integration of all subsystems in a hardware test that fires
> experimental shots."
>
> Would you have the same situation when a control room staff operations
> start up for a NASA satellite ? Absolutely not. I would suggest
> keeping an eye on whatever can be known about the actual operations
> for NIF to learn whether Ada and the software process it implies have
> actually been applied.
The answer can be found if you read a little further. First you have to
realize that, as implemented, this is a real-time system with extremely
precise timing requirements. The system as a whole is distributed
across more than a hundred CPUs, but when crunch time comes, some timing
requirements are measured in nanoseconds. Second:
The ICCS presently contains some 612 KSLOC�s of delivered
code. 63% of the code inventory is in Ada, 26% is in Java and
the balance is divided between C, HTML and shell scripts. The
measurement protocols exclude significant parts of the software
inventory, which totals approximately 2.0 million SLOCS. The
exclusions arise from redundant materials (for example units in
the inventory of multiple operating systems), code that was
automatically generated by translation tools (such as CORBA
server skeletons), and COTS products.
It is hard to determine what the "real" SLOC count is for error counting
purposes, since apparently CORBA object declarations are not counted as
source code. But the ball park figure is in the 1 to 3 million SLOC.
The defect reports indicate an error rate of between one and two bugs
detected so far per KSLOC.
But there is a final detail you have to understand. This is not a NASA
control room. This is more like building a warship. Some debugging of
software can be done elsewhere, but the real-time nature of the system
as a whole means that system integration has to take place on the actual
system. If you do want to compare to NASA, you should be interested in
comparing the number of open problem reports when the NIF becomes
operational to the number of open problem reports for the support
software for a NASA satallite when launched. (This number is typically
non-zero. For example known bugs in ground data reduction software for
a satellite won't prevent launch. In fact even the on-orbit software
for some planetary missions has been rewritten several times after
launch. Voyager II was the extreme case...)
So given the development plan and the type of facility being developed,
I'd say that this level of software problem reports would probably be
ranked as a moderate risk by the contracting agency at the time the
report was written.
--
Robert I. Eachus
"The war on terror is a different kind of war, waged capture by capture,
cell by cell, and victory by victory. Our security is assured by our
perseverance and by our sure belief in the success of liberty." --
George W. Bush
prev parent reply other threads:[~2004-01-30 1:51 UTC|newest]
Thread overview: 2+ messages / expand[flat|nested] mbox.gz Atom feed top
2004-01-30 1:10 Use of Ada in the NIF Gioia Tauro
2004-01-30 1:51 ` Robert I. Eachus [this message]
replies disabled
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox