sPHENIX is a new detector at RHIC.

[Martin Purschke <purschke AT bnl.gov>]

Hi Ross,

nice talk.

Just a correction, slide 5 - "Buffer boxes fill during peak lumi, drain 
out at end of fill". Not technically correct. I would say "The buffer 
boxes level the incoming data rates and allow us to send the average, 
rather than the peak rates to HPSS."

Some more background for questions: In principle it would be sufficient 
to operate the system as a "leaky bucket" where a variable rate of water 
flows in and a steady rate flows out. The differences in incoming rate 
is carried by the varying water level in the bucket (buffering).

For technical reasons we write the data to one of two file systems. When 
the written-to one hits a high-water mark, we clear the other (that has 
meanwhile been transferred to HPSS), and switch their roles. In this way 
one is only written to and the other is only read from, which 
dramatically increases the throughput (minimizing the disks' head 
movements).

Since our buffer depth is large enough that we can average over several 
stores (5-10) and their no-data fill time, and also cover breaks for 
accesses, and often no-data MD and APEX days, we can take  advantage of 
the significantly lower average data rate compared to the peak rates. 
The HPSS and SDCC transfers only need to be able to sustain that average 
rate (plus a bit of reserve).

Slide 6
non-RHIC folks will wonder where the 100ns come from (which should 
actually be 110ns == 1 beam crossing). I would try to bring across that 
the acquisition window needs to cover ~118 crossings (those 13 us) to 
get the data just from the triggered-on crossing, and by extending the 
window by n additional crossings we increase the covered crossings by a 
factor of n. This small delta is the "almost for free" aspect.

Hope it helps. Good luck with the talk.

        Martin

On 7/3/22 14:05, Ross Corliss via sPHENIX-l wrote:
> Dear All,
>
> I've posted draft slides for the sPHENIX data reconstruction talk at 
> https://indico.bnl.gov/event/16174/contributions/64794/attachments/41947/70236/sPHENIX.Data.Reco.ICHEP.2022.pdf 
> <https://indico.bnl.gov/event/16174/contributions/64794/attachments/41947/70236/sPHENIX.Data.Reco.ICHEP.2022.pdf>
>
> The time slot is 15 minutes, so I assume 12.5+2.5 for talk+questions. 
>   It takes place in the computing session on Friday, July 8: 
> https://agenda.infn.it/event/28874/contributions/169210/ 
> <https://urldefense.com/v3/__https://agenda.infn.it/event/28874/contributions/169210/__;!!P4SdNyxKAPE!HOjf-fN0HvR_ZMzbk6AZHmhZWHNGkfVDRTWXzeaMUH5cToXrNDht4HVvHfG3hFtfxRIteJAtIzHm0zt3PVyFNHpi1G3TLAwlQzk$> (11:30 
> AM local conference time)
>
> I welcome any comments or suggestions you may have, and I'll second 
> Murad's earlier thanks to all the speakers who have uploaded their 
> slides for the collaboration meeting and other recent reports.
>
> Gratefully,
> -Ross
> ==========
> Dr. Ross Corliss
> Research Assistant Professor
> Center for Frontiers in Nuclear Science
> Stony Brook University
> virtual office: 
> https://stonybrook.zoom.us/my/rossoffice?pwd=ZmZ2SlRIMVFvUUJwbUkyOVNVTmE5QT09 
> <https://urldefense.com/v3/__https://stonybrook.zoom.us/my/rossoffice?pwd=ZmZ2SlRIMVFvUUJwbUkyOVNVTmE5QT09__;!!P4SdNyxKAPE!HOjf-fN0HvR_ZMzbk6AZHmhZWHNGkfVDRTWXzeaMUH5cToXrNDht4HVvHfG3hFtfxRIteJAtIzHm0zt3PVyFNHpi1G3To4rbY1U$>
>
>
> _______________________________________________
> sPHENIX-l mailing list
> sPHENIX-l AT lists.bnl.gov
> https://lists.bnl.gov/mailman/listinfo/sphenix-l


--
Martin L. Purschke, Ph.D.        ;   purschke AT bnl.gov
                                 ;   http://www.phenix.bnl.gov/~purschke
                                 ;
Brookhaven National Laboratory   ;   phone: +1-631-344-5244
Physics Department Bldg 510 C    ;   fax:   +1-631-344-3253
Upton, NY 11973-5000             ;   skype: mpurschke
-----------------------------------------------------------------------