ATTENDEES:

Ernest, Terri, Antonio, Debbie, Stephanie, Kukhee, Till, Sheng, Diane 

TOPICS:

Discussed results of previous action items concerning IOC <-> EIOC communications. Attached are slides produced by Diane and Till (I've added Till's slide into my file) 

1) It was determined that using multicast messaging between all units couldwork, including between IOCs and their EIOCs. Need to include RF communications in calcs and also all the current unicast messages. If using the multicast capabilities in the EIOC effects the EIOC performance, then the switches can be statically configured and the EIOCs can continue to send unicast messages EIOC --> IOC.  No additional software effort, but this option requires extra effort in setting up multicast IDs for IOC EIOC communication. It is recommended to upgrade all 2960 switches to 3750s. NOTE: if we can make this a fully multicast network, then we can eliminate the (currently) dedicated switches, provided there are enough ports on the 3750 Field Switches for the EIOCs.

2) changing IOC broadcasts to unicasts requires changes to currently working BPM, BLEN, and RF IOC software, and a scheme for retrieving the second NIC's MAC address must be developed.  1-2 weeks effort for BPMs (Till), additional 1 week each maybe for RF(Kukhee) and BLEN(Sheng). No hardware changes. 

3) adding a third NIC. Requires new NIC for every RF, BLEN, BPM and MGNT IOC. Software effort: ~ 1week for network stack, ~2weeks for driver - if an e1000-based PMC card is selected.

ACTION ITEMS:

* Kukhee & Diane - add RF IOC communications and unicast messages into calcs and slides for all-multicast network traffic. Report by Wednesday next week.

* Ernest - cost estimate for NICs

NOTE: THE DECISION IS TO GO WITH OPTION 1a
 

Tradeoff

Option 1

Option 2

Option 3

Description

Shared Switch. Multicast messaging replaces broadcast.
1a) IOC broadcasts become multicasts, statically configure the EIOC ports on the dedicated switches
1b) all multicast network including EIOCs

Shared Switch. Unicast messaging replaces broadcast

Additional NIC

Technical Risks

* BPM IOCs send broadcast at 360Hz
* 1b) adding multicast to EIOCs could hurt performance

*See latency calc slides - latency is not a problem. 

Requires change in currently working IOC (and possibly EIOC) software

* changes to ip/udp stack can adversely effect the EIOC performance, since the EIOC uses the same code base
* multicast from EIOCs would mean changing code and can effect performance

Maintenance Risks

* 1a) configuring switches makes it a non-standard network and adds effort to SCCS maintenance, but this is a relatively simple configuration setup.

* every IOC / EIOC needs 'hardcoded' configuration that is difficult to maintain

NIC may become obselete
requires particular NIC that will make driver development easier

Additional Software Effort

* NA - already in schedule

* develop configuration scheme
* code configuration changes
* change broadcasts to unicasts
BPMs -  2 weeks
BLEN - 1 weeks
RF - 1 week

* research / test card
* modify ip/udp stack - 1 week
* write driver - 2 weeks

Additional Hardware Cost

* upgrade 2960s to 3750s. but note: if we go all-multicast then we may be able to eliminate the currently dedicated switches.
no additional cost -  already in AIP

NA

a NIC for each Feedback IOC
need a cost estimate

  • No labels