2017-12-14 Network Engineering Team Meeting notes

Review open action items related to demonstration.

Jeff Kantor

Unknown User (rlambert)

Skip this time.

Jeronimo Bezerra

Skip this time.

Jeronimo Bezerra

Reports from leads regarding demonstration:

Albert Astudillo for Sandra Jaque La Serena - Santiago.

Jeronimo Bezerra Santiago - Chicago, overall architecture/plan.

Using Internet2 AL2S. VLAN range assigned. Path from Miami to NCSA is ready.
Miami to SCL is almost completed. Only SCL to SAO is working properly.

Vinicius provided an update on the Pacific side.

Pacific link is not yet stable. Should use the Atlantic side for now.
Troubleshooting the Pacific link.

Unknown User (rlambert) Cerro Pachon - La Serena.

Link certified from Base to Summit. Link to SCL is being worked on.
Loop test between Summit, Base and SCL scheduled for tomorrow.
Link loopback is physical using fiber on summit, maybe internal in SCL
Someone will be creating a physical loop
6 links are planned.
DTNs are on and working. Using servers for testing.
Basic connectivity and network testing to proceed with DTNs when infrastructure becomes available.

Unknown User (kollross) Chicago - Champaign

Servers are racked with correct IP addresses
Trunked vlans from I2 down to NCSA core network.
Complete by end of week
Vlan IDs: SCL 1, NCSA 2, Miami 3.
Testing tools and procedure:
FDT will be used for testing.
Matt prefers to use iperf.
Server mode will be enabled for testing.

Vinicius: software for generating traffic. Issue with iperf when different versions are running. Suggested nuttcp.
After discussion, it was decided that iperf 3 is what should be used.

Jeff Kantor Requirements.

Viewing the demo:
Jeff: during the demo, shall we have an open BJ session for communications and troubleshooting?
How much traffic planning to generate, measure?

Jeronimo: Zabbix will be used for reporting utilization

Grafana is another tool that can be used

Expecting 9.6G per NIC.

Event will be streamed.
JB will provide a URL for people to view demo in RT, and reporting any issues:
Will display only to ourselves for now.

Jeff: what about a start time?
Jeronimo: depends on Sandra and Ron, when infrastructure will be ready.
Can then start process to generate data.

Jeronimo: Scripts will trigger event remotely. If we see poor performance that’s when we will engage everyone.
Sandra: suggested a start time of 11AM Chilean UTC-3. 9am eastern.
Jeronimo: Will create a BJ room for 48 hours.
Goal: 3 servers each with 2 NIC. Expect 9.7G per NIC.
Results will be published internally.

Jeronimo: Since the fiber and the active equipment are all new, expecting there will be some issues. Will claim success if we exceed 40G.

Jeff: Target min of 40G this time, is twice as much as old 2x10, new infrastructure, significant result
Asap, another test(s) with add servers to go up to full 100G.

Jeronimo: Are you planning to have more servers to fill pipe?
Loander servers can stay there for testing.
Jeff: for testing purpose can configure something sooner than would have been purchased on plan.
Plan to go higher, with use of more bandwidth.
SJ: buy extra NICs for servers?
JB: servers are limited to dual NICs. No space for expansion.
JK: confident additional servers can be purchased as pre-purchase of items to be purchased later on plan. Must be careful of finding place to put servers. Limited space, power in LS.
JB: plenty of servers with low rack units can replace current servers to not require more space.

MK: have 100G host that’s not in use. Can potentially be used for certain tests. Currently not connected to LSST network area. Would have to do a peering/routing.

JB: part of the traffic can be sent to Miami and part to NCSA. Goal is to fill pipes in Chile.
MK: biggest issue is the I2 connection that’s potentially the bottleneck.

One amlight link will be used. All vlans will be assigned to one link.
SJ: will demo be on the same link as production traffic?
JB: Yes.
SJ: What if there’s a conflict with start time?
JK: Not a problem if we have to slide the start time.

JB: 10G port at REUNA.
SJ: have a server with 1 NIC that can be used. If needed, it can be made available.

Wrap up and Next meeting date and agenda topics:

Engineering conference call for the demonstration was scheduled for 12 noon Chilean time UTC-3, 10AM eastern.

Next meeting will be TBD after new year 9am PT, 12pm ET, 2pm CLT