Choose timezone
Your profile timezone:
Developer's meeting
→
Europe/Zurich
EHN1:
- WIB connected (i.e., same situation as some months ago)
- some issues with WIB -> FELIX data transfer
- repeated timestamps (not complete frames) (assumed to be WIB issues)
- some data not as expected (e.g. bit flips in timestamp)
- after some msec, data boundaries not correctly identified anymore
- not sure if this could be due to clocking issues on the WIB side
- note: these issues existed already when doing the first tests with WIB (some months ago)
- have started throughput measurements (using srv 17, 18)
- will test on srv 19 later (hopefully it will not crash the machine again)
- Jingbo (monitoring) asked for some ROOT files (emulated data)
- seems to be all working
- suggestions:
- WIB: check to see if issues on both WIB links are correlated
- WIB: check whether RCE is having similar issues
- NIC: try with e.g. a 40 Gb/s interface (if the Mellanox card continues to give problems)
Milo: talked to group wanting to use WRM chip for pattern recognition
- apparently they are interested in getting re-ordered but uncompressed data
- this looks doable (Roland: the QAT compression, from a s/w perspective, is merely a fifo; the original data can also be retained)
- but some concerns about bandwidth usage; it is not clear if they want to use this only once or longer-term
- Milo will invite them to come and discuss their use case with us in some more detail
- suggest next meeting
Next meeting: tentatively planned for December 20 (last occasion before CERN Christmas closure)
There are minutes attached to this event.
Show them.