Brief Status of the E866 Analysis Upgrade - MJL - 11/1/93
    (I look forward to your comments and suggestions!)

Planning - We have made the decsion to proceed without being
a member of the FNAL DART collaboration since this arrangement
would have provided little benefit to us at considerable manpower
cost. For most of the things outlined below I have real solutions
in mind and the implementation looks staightforward. For the issues
of databases and magnet maps, in particular, things are much less
clear.

Backend Analysis Computers - We need about 150 MIPS of backend UNIX
computer power to analyze 10% or so of the data in real time. No
progress has been made on this front. It appears to be accepted by
the FNAL computing department that they will supply this, and this need
is called out in our MOU. However since the MOU is not completed
and not signed it is difficult to proceed. Once the MOU is done we
will be in a position to begin negotiations with the FNAL computing
department to specify exactly what these systems will be (e.g. computer type,
amount of memory and disk space, etc.), how they will fit onto the
network, who will manage them, and eventually installation.

Hoist Data Path - We plan to use the "hoist" scheme for shipping data
in real time to the backend computers over ethernet. I have a working
prototype for this that consists of several processes implementing
this data path:
1) A hoist server that runs on a generic UNIX machine. It reads data
from a file and forks one or more processes to service remote hoist
consumers. Eventually this process needs to be moved into the VxWorks
environment where it will then obtain its data from the real DAQ system.
2) A hoist consumer that obtains data over the hoist link from the hoist
server and then fills a local event pool in shared memory with this data.
3) A drainer which drains data out of the event pool and then prints portions
of the data out on the screen. Next this guy needs to become the data source
subroutine for our full analysis program.
Some amount of fault tolerance is built into these things already; more will
probably be necessary. I have tested with multiple hoist clients and multiple
drainers sucessfully.

User Interface - I have a working prototype using CERNLIB/KUIP as a command
interface. This needs to be grafted into the big analysis program and fleshed
out.

Shared Memory Histograms - This is supposed to work now with the latest version
of CERNLIB on SGI computers but not on HP's. The HP fortran has a design limitation
that makes shared memory inaccessable from Fortran code so far. But since we
will probably use SGI's in the counting house this should not be a problem.

New CMZ files for Analysis and Monte Carlo - With help from Pat I have
now cleaned up the old E789 programs and made fairly clean CMZ versions
of our main programs which have been tested for some circumstances and
appear to work fine. So everyone should plan on using CMZ to manage the programs;
no more patchy line numbers! In these programs the silicon detector stuff
from E789 has been removed and the E772 absorber configuration has been put
in as the default. A new section in the survey file specifies the absorber
configuration so it is not much easier to change and keep track of. The latter
is not implemented for the analysis yet but it should be soon. Also the
proceedures for using CMZ are fairly simple so far; further work on streamlining
operatations with CMZ is in order. Any CMZ experts out there?

Backend scaler based monitoring - I have not heard of any programing progress
here. Anything from NMSU?

Database - So far we plan to continue the old survey file methods for the
geometrical constants of the spectrometer. However we still need a way of
organizing all of the other data associated with the experiment. Maybe
the study underway at RHIC to put up a database system for experiments there
will help us as well?

Magnet field maps - Considerable discussion at E866 meetings here at Los Alamos
has taken place. I think we agreed that some kind of mapping is required before
we begin taking data. How extensive it will be is still under discussion; the
possibilities of using modeling of the magnet and/or tuneup using real
data tracks to help, are also being discussed. Obviously there will be a substantial
analysis task associated with whatever path is persued here.

Absorber Studies - I understand from Pat that there has been progress at CalTech
in simulating the leakage into station-1 using a GEANT model of the spectrometer.

UNIX farm development - Once the CMZ based analysis has settled we could try to
build it in the farm configuration and test it on an SGI based farm simulator.
It might also be nice to get the simulator running on an HP. Later after we
prove it works again on the simulator we need to investigate where we hit
the IO bottleneck on a real multiprocessor farm and see if we improve it
by reworking some of our code, and by working with the FNAL computing department
to improve the system hardware and software.