Presents: Edoardo Martellli, Eli Dart, Marian Babik, Trstan Sullivan, Vitaliy Kondratenko, Joe Mambretti, Bill Johnston, Dale Carder, Tom Lehman, Andy Lake, AJ,Richard Yang, Jim Chen, Ofer Rind, Julio Ibarra, Marcos Schwarz
SC23
NOTED demo
- FNAL will participate to the demo. Tom will set up a SENSE circuit to connect CERN and FNAL. Edoardo will send en email too FNAL, Tom, Bill to coordinate the task.
- The SENSE vlans used last year should be used again. They needs to be tested.
- Joe: CANARIE is setting up a 400G path to University of Victoria. A 400G path from Seattle to Starlight is also being set up. It has not been decided yet if it will be part of AutoGOLE. There are two other alternative path possible, if this one doesn't come up.
- Vitaliy: SFU has agreed to purchase the 400G optics, so it's very likely that the link will happen. Clarified that SFU is the main location of TRIUMF Tier1.
- Bill has sent a new version of the map with the resources that will be used for the SC23 demos
- Joe: FNAL should be added to Bill's map
Packet marking demo:
- Marian would need the dashboard that was used last year with Scinet to show the packet marked with the flowlabel fields. Eli thinks it can be done.
- CERN will also try to set up a dashboard for the marked traffic that that transit via CERN.
- RichardY asks if it would be possible to aggregate all the statistics collected in the different places to try to correlate and compare the transfers. It should be possible with the fireflies tags
- The marking of the production FTS traffic wasn't planned for SC23, so it's unlikely to happen. The proposal with concrete plans for packet marking for DC24 was uploaded to the DC24 folder.
DC24
Edoardo and Carmen have written two proposals and added them to the folder https://cernbox.cern.ch/files/link/public/aClTXJenZxpF5qw
- One is about accounting of flowlabel tagged packets and can be merged with the scitags proposal.
- The second proposal is to dry-run a new version of NOTED that is triggered by link saturation alarms on CERN routers. In case of interest, it may be adapted to other sites.
Eli: ESnet now has a large deployment footprint of perfSONAR interfaces in LHCONE that could be used for DC24 monitoring
- https://ps-dashboard.es.net/maddash-webui/index.cgi?dashboard=51%3A%20ESnet-LHCONE
- These are IPv6-only, with separate interfaces for iperf and OWAMP testing.
- We are planning per-experiment pages for ATLAS and CMS - the intent is to have the relevant ESnet perfSONAR hosts on one axis and the appropriate community perfSONAR hosts on the other axis.
- AndyL: we need to discuss the tests with Shawn to not overload servers. Also to be discussed the dashboard to show the fireflies monitoring.
- Marian: there is already a LHCONE perfsonar mesh. They can be part of the mesh of ESnet probes, but we need to decide how and where the full mesh is managed.
- RichardY: there are some problem with the data generated by some perfsonar transfers. How can they be fixed? Richard to get in touch with Marian.
- Tristan: would it be possible to install flowd on the ESnet perfsonar servers to do the packet marking?
- Eli: the flowlabels can be set with iperf. But flowd can also be considered. What label should be used? The activity ID can be perfsonar. The experiment ID can be a special one made up for this activity.
