NetNews Usenet Archive 1992 #16/NN_1992_16.iso/spool/sci/physics/11729

Internet Message Format | 1992-07-27 | 3.9 KB

Path: sparky!uunet!dtix!darwin.sura.net!jvnc.net!yale.edu!qt.cs.utexas.edu!cs.utexas.edu!sun-barr!ames!agate!agate!matt
From: [e-mail] (Matt Austern)
Newsgroups: sci.physics
Subject: Re: SSC cost
Message-ID: <[e-mail]>
Date: 27 Jul 92 14:23:07 GMT
References: <[e-mail]>
Reply-To: [e-mail]
Organization: Lawrence Berkeley Laboratory (Theoretical Physics Group)
Lines: 59
NNTP-Posting-Host: physics2.berkeley.edu
In-reply-to: [e-mail]'s message of Mon, 27 Jul 1992 16:24:56 GMT

In article <[e-mail]> [e-mail] writes:

> I may type in a fuller treatment of the data-handling problem later in
> the week. {Mr. Austern, will also want to take note.} I said that the
> figure being "bandied about" is a "Terabyte per nanosecond," and
> submit the following teasers. From the Conference Record of the 1991
> IEEE Nuclear Science Symposium and Medical Imaging Conference, Nov.
> 2-9, 1991, Santa Fe, New Mexico, USA, vol 2 of 3...
>
>   "Data Acquisition Studies fro the Superconducting Super Collider"
>   by Milner, Booth, Botlo, Dorenbosch, and Wang of the Physic Research
>   Division of the SSC, and Idate, Kappor, and Raj of the University of
>   Texas, p. 929:
>
>     The collider will provide a beam crossing every 16 ns, with a mean
>     of 1.6 collisions per crossing (100 mb cross section) occuring at
>     the design luminosity of 10^33 cm^-2 s^-1, giving a 100 MHz inter-
>     action rate. We can expect data for each event to be about 0.4
>     Mbyte, giving a raw data rate of 40 Tbyte/s...
>
> Perhaps Doctor Hallam-Baker would care to translate that, or he might
> wish to refer to another paper from the conference...

No translation is necessary.  Mr. Tice asserts that the SSC will
produce one terrabyte of data per nanosecond, then he quotes a paper
saying that the rate is 40 terrabytes per second---this is eight
orders of magnitude lower than the figure he asserts!  Eight orders of
magnitude is not a trivial difference: it is the difference between a
figure which is high, but managable, and a figure which cannot
possibly be useful.

Further, as the paper he quotes clearly says, this is the "raw" data
rate---that is, this is the data rate if you wrote every single event,
at every single beam crossing, to tape.  The whole point of any
experiment in high-energy physics, though, is that in most
interactions, almost nothing interesting happens, so you don't bother
to look at those events at all.  Any high-energy detector selects out
those events which are likely to be worth studying, triggers on those
events, and writes only those events to tape.  Typically, somewhere
between 1 and 10 events per second (*not* one hundred million per
second) get written to tape.  This is another six or seven orders of
magnitude.

Look: this is not a new problem with the SSC.  The issue of triggering
has been well studied at every high-energy collider, and it is a
well-known issue.  This is not something that appears in newspaper
articles, but it does appear in any of the technical literature of
this field.  (I have been quoting the SDC Letter of Intent, but you
could equally well look in, for example, the technical papers
describing the CDF detector.)  The people who are building these
detectors know perfectly well that they cannot write a terrabyte of
data to tape every nanosecond, and there is no point in claiming that
they are so utterly foolish as to imagine otherwise.
--
Matthew Austern              I dreamt I was being followed by a roving band of
[phone]               of young Republicans, all wearing the same suit,
[e-mail]    taunting me and shouting, "Politically correct
[e-mail]       multiculturist scum!"... They were going to make
[e-mail]         me kiss Jesse Helms's picture when I woke up.