MEMORANDUM
TO: James Shank and John Huth
FROM: Bob Ball, Myron Campbell, Jay Chapman, Steve Goldfarb, Suen Hou, Shawn McKee, Dan Levin, Homer Neal, Rudi Thun, Bing Zhou
DATE: December 10, 1999
SUBJECT: ATLAS Computing at the University of Michigan:
Status, Plans, and Needs
This note is in response to your request for information about computing sub-system activities and plans in the University of Michigan ATLAS program. Though we have provided much of the information in a previous submission [ University of Michigan ATLAS Computation and Software Development: A Proposal to US-ATLAS Computing Management , July 1999] we are herein updating our status, utilizing the WBS structure that has since been established in U.S. ATLAS. We are also extending our request based on additional tasks being undertaken by our group.
We believe we have one of the strongest software groups within US-ATLAS (and ATLAS) and we look forward to making a significant contribution both within the detector subsystems, where much of our effort is now appropriately directed, and ultimately to many of the core areas of simulation, reconstruction and analysis. Indeed, based on the scope of the proposed computing work at Michigan, and the size of our software group, we are hereby requesting that our site be qualified as a Tier 2 US-ATLAS Computing Center.
As a brief overview of our computing activities over the past year, we can report that we have:
We describe below our software activities and future plans and also present our funding request. Our entries are based on the WBS structure of the US-ATLAS Computing.
Current activities and planning on the ATLAS computing
Establish `test bed' for the study of muon final states:
(a) H0 - > 4 mus
(b) Z'/Z0 -> 2 mus
(c) W' -> 1 mu
These reactions will permit us to test the emerging reconstruction code, and the trigger algorithm and its efficiency and speed. Specifically, we will test the GEANT4 based programs, the trigger database, as well as the muon L2 trigger programs.
Responsible Individuals: Bing Zhou, Shawn McKee, Dan Levin
Status: Earlier version of the same physics simulation for ATLAS muon
system has been developed by Bing Zhou; we will continue
the simulation and tests for the new software codes.
We will develop the ATLAS trigger database, and use CDF as our test-bed.
Responsible Individuals: Myron Campbell, Homer Neal, Jay Chapman, and
Chunhui Han
Status: Myron Campbell, Homer Neal and Chunhui Han have already been actively exploring useful parallels between the trigger database now being established for the CDF experiment and the expected ATLAS structure. Contacts have been made with the ATLAS Trigger Coordinator. A work plan and testbed model is being developed. Homer Neal made an initial report on these plans at the recent meeting of the ATLAS database working group at CERN.
Steve Goldfarb has assumed responsibility for the overall ATLAS muon database coordination role. In this role he is responsible for developing and guiding the development of all aspects of the muon database effort, including detector description, the event data model, calibration database and the muon portion of the trigger database. Recently, he has implemented a generic geometry model based upon XML that will provide the ATLAS muon geometry for simulation, reconstruction and analysis software. This generic model needs to be populated with the data from AMDB for each muon subsystem. Michigan has taken responsibility for implementing the XML version of the MDTs based upon the existing AMDB database. Steve has provided XML examples for each subsystem to reference as they implement their generic XML descriptions.
Future geometry database plans include completion and testing of the XML detector description along with implementation and testing of the simulation, reconstruction and analysis interfaces.
The added responsibilities Michigan has undertaken for the muon database coupled with the pressing need for implementations of muon event models and the muon calibration database clearly require an additional person. We are requesting funding for a muon database programmer to work full time on the most time critical aspects.
Responsible Individuals: Shawn McKee and Dan Levin
Status: Dan has developed the auto-calibration code and test it for barrel
muon MDT geometry, and will carry out more studies for the end-cap
MDT chambers.
Shawn and Dan have setup the GEANT4 software on our local machines
(both on HP workstation and on PC Linux), and already successfully
running the GEANT4 ATLAS muon example programs.
We will continue our effort on the muon level 2 software development and
simulations, and will provide testing software for final online hardware
implementation.
Responsible Individuals: Dan Levin and Bing Zhou
Status: Bing started the ATLAS Level 2 trigger studies using MDT sub-system
in 1997. Recently, Dan has setup the ATLAS simulation code on UM
HP machine based on the new ATLAS software. We expect to start the
L2 trigger software development and test shortly.
We will build on our successful web-based lecture tools (Syncomat-3000) in support of the overall ATLAS and US-ATLAS software training activities.
We will enhance existing video conferencing tools through the development and implementation of Quality-of-Service techniques in collaboration with Internet2/CERN/ and MERIT Network researchers.
We will continue to aid in the expansion of CERN/US networking bandwidth through efforts with UCAID and CERN networking. This continues our role in getting CERN connected into STARTAP and its planned joining of the Ann Arbor based INTERNET2(UCAID).
Responsible individuals: Homer Neal, Bob Ball, and Shawn McKee
Status: Our web-based archiving tool has been released and has been undergoing extensive testing in collaboration with CERN Network services and CERN Academic Training. Many LHC Software training talks given by the acknowledged experts in C++ and OO have been recorded and made available to the ATLAS/CMS communities. The unique features include high quality audio, clear PowerPoint slides, all presented with a synchronized video image of the speaker. The files are kept on a CERN server, purchased by Michigan but maintained by CERN staff. These archives can be accessed by any with access to a (free) browser and RealPlayer. The power of this tool in use as a vehicle for outreach and education has also been demonstrated by the recording of over 30 lectures in the CERN Summer Student Program.
Collaborations are being established with researchers in INTERNET2 and the MERIT NETWORK (the original operators of NSFNET), both here in Ann Arbor, to investigate and deploy the use of Quality-of-Service to attain better web-based video conferencing capability for uses in collaborations like ATLAS.
Our plans are to extend Syncomat to a new release, with specific features designed to meet the needs of US-ATLAS Software Training, to help all US-ATLAS sites to optimize their ability to both record Syncomat presentations and to play back already recorded presentations. We will integrate our LHC software training work with that desired by US-ATLAS. We will see that advances that occur in other related work with CERN/INTERNET2/MERIT are immediately made available to US-ATLAS.
Budget needed for carrying out the computing work outlined above
A professional software developer is needed to develop and disseminate a new release of the Syncomat-3000 software package that will be optimized for use in software training applications. This will include replacing RealPlayer with QuickTime as the basic a/v platform. This individual will also explore improved compression technologies and will play a role in developing final QoS implementations for ATLAS software training and videoconferencing.
The collaboratory technician/installer will assist with the implementation of the packages described above, but will also be available to visit other US-ATLAS sites to train local experts on the use of the packages, setting up recording systems and optimizing the playback techniques for the platforms in use at each given site. This person would, in addition, help facilitate the sharing of training software lectures given at CERN and in US-ATLAS, given our involvement in both efforts.
Equipment requested for support of these activities consists of a single PC system optimized for A/V software development and evaluation. In addition will require a high-end switch with QOS capability for testing and development. Recently CISCO release the 6000 series with support for QOS and this is what our budget is based upon. The supplies indicated in the budget are needed to acquire the related recording media, cablings, microphones, etc., to record, evaluate and archive both final and testing sessions. The travel item is primarily for the technician/installer to visit other US-ATLAS sites and CERN, and for trips for the author of Syncomat to confer with networking experts at CERN on how to best optimize the use of the increasing US/CERN networking bandwidth and the QoS protocol for the collaboratory tools being developed.
As discussed above, the additional ATLAS responsibility Michigan has assumed for the muon database effort will require an additional person to help design, develop, implement and test muon database software. This is critical for the timely delivery of software models and components required for the muon database effort. In addition, we are requesting travel funds for this person, to support 3 trips/year to/from CERN.
To support our simulation and trigger work will require much more CPU power
than we currently have available. Our groups’ experience with running
GEANT and level 2 trigger simulations, coupled with estimates of the required
number of simulated events for reasonable statistics, lead us to conclude we
will require approximately 10 “fast” CPUs (~700 MHz PIII) dedicated
to simulation production. We intend to assemble a LINUX CPU farm, by purchasing
5 dual processor computers, each with 36 gigabytes (GB) of high speed (SCSI
160M) disk and 1 gigabyte (GB) of fast RDRAM (400 MHz). Since these will be
compute servers, we will not normally require monitors, keyboards or mice for
each machine, but instead will use KVM (keyboard, video and mouse) switches to
connect to each machines console when necessary. GEANT simulations require a
good graphics environment, so we will purchase three 21”, multi-synch
monitors, thereby creating three “simulation” stations for software
development. These five compute servers will be tied into our local network
via a 12 port, 10/100 Mbps manageable switch. The last critical piece of
hardware is a backup device capable of quickly storing all the code and data we
generate as well as backing up our development environment. Current DDS-3 4mm
tape loaders can backup 192 GB of data on one 8 tape cartridge overnight,
which is sufficient for our anticipated near term data load.
It is essential to have a computer manager within the group full time to set
up, maintain and upgrade our computers, as well as install and upgrade the
global ATLAS computing software. Currently this work is done by physicists
within our group, whose time would be much more productively spent on the tasks
outlined above. We request support for such a person whose primary
responsibilities will include:
The last part of the budget requests funds to purchase any required software
and supplies. While our chosen operating system (LINUX) is free, there are many
components we will need to purchase: compilers, graphics software, backup
software, CASE tools, etc. In addition, many of the possible database engines
for our production databases may need to be purchased.
Closing Comments
We note the need for support of extraordinary software efforts directed to
detector construction, where little or no support is currently provided from
the computing or detector budgets. An example of such an activity for us is the
implementation of components of our production database for the entire ATLAS
global muon production database in Rome. Another is the temptation that some
will have for relegating our Trigger Database work solely to the DAQ arena,
when in fact there are large components deeply rooted in the ATLAS core
software. It is critical that such key activities in ATLAS not die just because
of present inadequate wbs structures. We urge that this matter be reviewed.
Moreover, we wish to stress the importance of ensuring that strong groups like
Michigan have the necessary local resources to contribute effectively to the
success of ATLAS. We appreciate the need for certain services to be provided at
larger, centralized entities. But we also note that many types of ATLAS
activities are best conducted in university environments of the kind we have at
the University of Michigan.
Please let us know if more information is needed.
PC system optimized for A/V R&D
Supplies (Cables, media, microphones, etc.)
QOS capable switch (Cisco 6000 series)
Computers (2-PIII/733, 36 GB SCSI 160M, 1 GB rdram)
Backup drive with autoloader (Sony TSL-9000, 192 GB)
Monitor, 21" Multisync --- for software development
Network switch 10/100, 12 ports, manageable
Commercial software (compilers, backup, utilities, etc.)
Backup media and supplies (40 DDS3 tapes)
Computing hardware and supplies total
Travel for database software development
Travel for collaborative tools installation, etc.
Total travel
Database software developer (100%)
Overhead/benefits
Collaborative tools programmer (100%)
Overhead/benefits
Collaborative tools technician/installer (50%)
Overhead/benefits
Computing system manager (100%)
Overhead/benefits
Total salaries and benefits
Total Proposal