NPPS Leadership Team

[pinkenburg <pinkenburg AT bnl.gov>]

    Hi Gene,
    
    one needs to read the full report to gauge the priorities and not
    cherry pick. The sPHENIX computing needs are driven by the fixed
    latency reconstruction during data taking. Without this we will not
    be able to establish that sPHENIX is working according to specs.
    There is no point finding out months after the run that you should
    have done things differently.
    
    2.2 Discussion and Recommendations for RHIC Run 23
    The PAC continues to strongly support focusing in Run 23 on sPHENIX
    commissioning using
    200 GeV Au+Au collisions. This is the highest priority and must come
    first. We encourage the
    sPHENIX collaboration to define detector commissioning goals along
    the lines of those that they
    shared with us at this meeting, metrics for what constitutes
    successful detector commissioning.
    Successful detector commissioning should be followed by continued
    running of 200 GeV 
    Au+Au collisions to begin the sPHENIX scientific program. A central
    goal of this part of Run 23
    can be referred to as “physics commissioning”. By this we mean
    collecting, and reconstructing
    and analyzing, the data needed to establish that the sPHENIX
    detector has met its previously
    defined Ultimate Performance Parameters (UPPs). (These were defined
    previously in such a way
    that achieving the UPPs will allow the analysis requirements to be
    met for the data to be taken in
    Runs 24 and 25, so as to achieve the sPHENIX physics goals first
    articulated in the 2015 Long
    Range Plan.) The allocation and deployment of beam time and
      computing resources during Run
    23 should be guided by the perspective that successful
      sPHENIX installation, detector
    commissioning and “physics commissioning” is the prerequisite
      to successful operation of
    sPHENIX in Runs 24 and 25 so as to accomplish its science
      goals, which are the overarching
    priority for RHIC in 2023-25
    
    Chris

On 7/6/2022 11:34 AM, Van Buren, Gene via Phys-npps-mgmt-l wrote:

Hi, all

Here is the final Recommendations document from this year's RHIC PAC:

https://www.bnl.gov/npp/docs/2022-npp-pac-recommendations-final.pdf

I have cut & paste the sections I found relevant to software & computing below for those who don't want to scan the document for these points themselves. The final "Offline Computing" section primarily emphasizes ("most profound concern") the upcoming computing hardware resource needs of both STAR and sPHENIX, not prioritizing either over the other. PHENIX analysis resource needs are also noted.

-Gene


_____________________________

1. Executive Summary
...
The PAC endorses STAR’s emphasis on the importance of maintaining sufficient expert personpower for software, computing and operations at BNL, which is critical for high-quality data collection, and timely calibration and production of data.
...
sPHENIX has met and surpassed its goals for designing, building and testing its software framework for compression, storage and analysis. It has also improved the memory footprint of the reconstruction software to optimize implementation at the SDCC. Software appears to be on schedule for simulation, reconstruction, calibration, compression and storage, for the inaugural run in February. The PAC is pleased to hear that the software is in place and has passed two Mock Data Challenges.

Detailed hardware requirements for off-line computing for sPHENIX have been provided to BNL management, including the use of off-site computing resources. The 15 kHz event-rate of sPHENIX requires an increase in CPU and storage usage of an order of magnitude. Concerns regarding computing resources were expressed in the 2021 S&C reports but have yet to be addressed adequately. sPHENIX estimates an increase by about a factor of two in resources is needed for initial processing of off-line data. It is imperative that sPHENIX and BNL management address these concerns in time for successful data-taking and analysis for Run 23. The PAC urges BNL management to assign highest priority to the procurement of the computational resources needed to analyze RHIC data in a timely manner.
...


_____________________________

3.1 STAR
...
The PAC is pleased to see the progress in the software developments for detector calibrations, alignment, offline data reconstruction, and the initial fast simulations. We look forward to seeing the completion of the full detector response simulation software, the optimization of the tracking software, and eventually – the data analysis and physics results.
...
The PAC endorses STAR’s emphasis on the importance of maintaining sufficient expert personpower for software, computing and operations at BNL, as is critical for high-quality data collection, and timely calibration and production of data.
...


_____________________________

3.2 PHENIX
...
PHENIX made significant progress in data and analysis preservation using international expert systems such as REANA, Zenodo and HEPdata.

The PAC supports these efforts in data preservation. The PAC supports PHENIX’s request to BNL management for resources necessary to complete critical analyses such as b,c→μ decays.


_____________________________

4 sPHEINX
...
4.2 Readiness for Data Taking including Computing & Commissioning Plan

The PAC was pleased to learn that sPHENIX has met and surpassed its goals for designing, building and testing its software framework for compression, storage and analysis. Using the current snapshot, sPHENIX can analyze a Au+Au event in 21 seconds, where the 2019 goal was set at 24 seconds. Further, sPHENIX has improved the memory footprint of their reconstruction software to optimize implementation at the SDCC. Software is on schedule for simulation, reconstruction, calibration, compression and storage, for the inaugural run in February.

Since 2019 sPHENIX has undergone annual S&C reviews. The PAC is pleased to hear that the software is in place and has passed two Mock Data Challenges. Reports from the 2021 S&C reviews were positive with respect to the software, but serious problems in procuring the resources were discussed. An S&C review was undertaken in May of 2022, but the report from that review was not yet available to the PAC.

Detailed hardware requirements for off-line computing have been provided to BNL management. This includes studies of using off-site computing resources, e.g. the OSG. Not surprisingly, the 15 kHz event-rate of sPHENIX (which was the principal motivation in building sPHENIX) requires an increase in CPU and storage usage of an order of magnitude.

The serious concerns about resources for off-line computing expressed in the 2021 S&C report are not yet addressed. As these concerns affect both STAR and PHENIX, they are addressed below.


_____________________________

5. Off-Line Computing for Runs 23-25

The PAC expresses its most profound concern regarding the computing needs for both sPHENIX and STAR for runs 23-25. Despite the significant progress sPHENIX has made in optimizing their computational software, the sPHENIX computing needs remain an order of magnitude greater than previous experiments, and STAR is managing to double their event throughput while implementing several new detector subsystems. The current SDCC planned computing resources are more than 50% short of what is needed to meet the sPHENIX and STAR needs in 2023, and the shortfall is expected to increase in outer years. Without additional computing resources, the ability of sPHENIX to calibrate their analyses and plan for future runs will be severely compromised, and scientific results from both collaborations will be severely delayed. This shortfall needs to be addressed in the coming months to avoid a tragic underuse of the scientific effort being invested to collect this data.

The PAC urges BNL management to assign highest priority to the procurement of the computational resources needed to analyze this data in a timely manner. The PAC requests that BNL, STAR and sPHENIX provide an update on their plans for procuring the needed resources in October of this year.


_____________________________

7. PAC Reommendations
BNL Lab Management
...
The PAC urges BNL management to assign highest priority to the procurement of the computational resources needed to analyze RHIC data in a timely manner.
...


_______________________________________________
Phys-npps-mgmt-l mailing list
Phys-npps-mgmt-l AT lists.bnl.gov
https://lists.bnl.gov/mailman/listinfo/phys-npps-mgmt-l
    
    
-- 
*************************************************************

Christopher H. Pinkenburg	;    pinkenburg AT bnl.gov
				;    http://www.phenix.bnl.gov/~pinkenbu

Brookhaven National Laboratory	;    phone: (631) 344-5692
Physics Department Bldg 510 C	;    fax:   (631) 344-3253
Upton, NY 11973-5000

*************************************************************