2017-08-24 Network Engineering Team Meeting notes

Review open action items.

Update:

Jeff Kantor created confluence page with LSST Network Traffic Types. These definitions can be used for network design, QoS/Prioritization planning, etc. and will be incorporated into LSE-78 in a future LCR

Jeff Kantor produced new version of Bandwidth Allocation table (need to discuss minimum allocation only for Science Data Transfer, some on Tiger Team suggest we would be wise to reserve a minimum for Control as well)

Tim Jenness Imported verification plans for DM-provided networks into the official Magic Draw System Engineering project. Jeff Kantor Working on TS-provided (Summit Network) with System Engineering.

Jeff Kantor Clarified Virtual NOC requirements (per Network O&M Plan) and timing needed is NLT start of ComCam observing in early 2020. Jeronimo Bezerra says we can get started on integrating AURA/REUNA, FIU NOCs now.

Jeff Kantor clarified 5 seconds requirement for LS - NCSA and distinguished. Unknown User (kollross) reported he and Jim Parsons have been simulating with 160ms delay, 10 G - 10 G connection, a single Camera image, and can already get to 7 - 8 seconds time frame.

Jeronimo Bezerra

Unknown User (rlambert)

Update:

We postponed last meeting to give time to work this. Jeronimo Bezerra reported on draft plan (link) and plan for completing. Reviewing vendor solutions/capabilities (Brocade, Cisco, ...) for QoS by early September. Then will run tests on controlled environment in FL by Sept 21, then test with legacy devices by Sept 28. Will record video of tests and use for training vehicle later. Then discuss with NET, and make recommendations available to broader project. Also will look into over-provisioning.

Jeronimo Bezerra

Reports from leads:

Unknown User (rlambert) Cerro Pachon - La Serena

Sandra Jaque La Serena - Santiago, connectivity to ADASS hotel, presentation design, INRIA involvement, fall-back Chile only demonstration

Jeronimo Bezerra Santiago - Chicago, overall architecture/plan

Unknown User (kollross) Chicago - Champaign

Jeff Kantor Sent clarifying note to NCSA on scope, charter of demonstration, repeated here:

Management perspective

• The Cerro Pachon end of this will use the AURA DWDM installed in the caseta on CP, which is being installed in September
• The LS end of will use the AURA DWDM installed in the existing LS data center, which is also being installed in September
• REUNA will have the LS - Santiago DWDM installed in October and that will be used
• FIU will use the 100 Gbps ring from Santiago - Miami, and REUNA and FIU will need to be connected in Santiago
• We will use i2 existing connections for Miami to NCSA
• None of the connections need to be routed/switched, they can be direct patches solely for demonstration purposes
• The method of the demonstration does not require disk to disk transfers, memory to memory is acceptable
• Memory to memory MAY still allow for simulated image data to be staged (in memory) and transferred (to memory)
• I asked Matt K to determine if this is feasible and if NCSA DAQ can be used to create the simulated image data (we do have other ways to create the data if useful)
• Memory to memory should still permit us to show, via some output to a web portal, the statistics of the data transfer and network performance during the transfer
• IF POSSIBLE, a display of a simulated image on either end before and after transfer would be compelling and visual, and the time to extract and render the display does NOT count against network performance
• Our goal should be at least 4 x 10 Gbps in the transfer, up to a max of 10 x 10 Gbps

Discussion:

Jeronimo Bezerra For ADASS have plan A - La Serena to NCSA, plan B - Santiago to NCSA. Could move perfSonar in Santiago or use other machines, working on 6x10 G. Have to provision both servers and transceivers, but we have identified what we need and where to get them. Miami to NCSA is not a large challenge.

Albert Astudillo reports vendor has delayed DWDM delivery again, won't be in time for ADASS, should still be possible for end of year. Jeronimo Bezerra emphasized will still want for Supercomputing 2017. FIU will also want to send 10 G from Chile to Denver and if possible, on to Korea.

Next meeting date and agenda topics:

September 21