Want better tooling for PRBS mode for transceiver qual

[rmustacc]

Right now we have the ability to set ourselves into a PRBS mode; however, understanding that it's working is rather complicated and limited. The ucli has something like:

bf-sde.pm> prbs-show
prbs-show
Integration time: 10ms
...
-----------------------------------------------------------------------------------------------------------+--------------+----------+ ---------+---------+---------+---------+---------+---------+---------+-..
PORT |MAC |D_P|P/PT|PRBS  |    0     |    1     |    2     |    3     |    4     |    5     |    6     |    7     | Total Errors |   BER    |        1e-05     1e-06     1e-07     1e-08     1e-09     1e-10      1e-11
-----+----+---+----+----------+----------+----------+----------+----------+----------+----------+----------+--------------+----------+ ---------+---------+---------+---------+---------+---------+---------+-..
39/0 |18/0|272|2/16|PRBS7|         0|         0|         0|         0|          |          |          |          |            0 |  0.0e+00 | ---------+---------+---------+---------+---------+---------+---------+-..
40/0 |18/4|276|2/20|PRBS7|         0|         0|         0|         0|          |          |          |          |            0 |  0.0e+00 | ---------+---------+---------+---------+---------+---------+---------+-..
65/0 | 0/0|  2|0/ 2|PRBS7|  51674046|          |          |          |          |          |          |          |     51674046 |  5.0e-01 | 

There are a few things here that are worth calling out:

• Right now I had actually used swadm link set-prop prbs 31 on 39/0 (qsfp6) and 40/0 (qsfp7), but it shows us in PRBS mode 7.
• It'd be good to get PRBS-specific counters out somehow when we're in this mode.

More generally, we're using PRBS modes as a way to try to understand the effectiveness of transceivers. When the gearbox doing FEC is in a dsp we lose the ease of the link counters fec logic that exists. The main things that we want to be able to get out is:

• How much PRBS data is being received?
• What are the observed error rates?

And use this as a proxy of link quality. It's hard to use the normal rmon counters because we're using PRBS rather than actual traffic.

[rmustacc]

Note, I'm increasingly convinced that something is wrong with the dendrite prbs mode enablement right now. Using the dendrite API I set a pair of links to opposite PRBS patterns, but the prbs checker did not fire or notify anything. When I manually enable it through ucli to opposing patterns, I do get errors logged.

[Nieuwejaar]

Note, I'm increasingly convinced that something is wrong with the dendrite prbs mode enablement right now. Using the dendrite API I set a pair of links to opposite PRBS patterns, but the prbs checker did not fire or notify anything. When I manually enable it through ucli to opposing patterns, I do get errors logged.

What do you mean by a pair of links? Is this a special setup where we have two qsfp ports connected to one another? If so, is this on a lab system I can use to debug?

[rmustacc]

Note, I'm increasingly convinced that something is wrong with the dendrite prbs mode enablement right now. Using the dendrite API I set a pair of links to opposite PRBS patterns, but the prbs checker did not fire or notify anything. When I manually enable it through ucli to opposing patterns, I do get errors logged.

What do you mean by a pair of links? Is this a special setup where we have two qsfp ports connected to one another? If so, is this on a lab system I can use to debug?

Yeah, I had a bunch of QSFPs that are connected between the two switches. Some of these are rigged up in both london and dublin.

[Nieuwejaar]

root@oxz_switch1:~# swadm link create -s 100g --fec rs qsfp22
Created link qsfp22/0
root@oxz_switch1:~# swadm link set-prop qsfp22/0 prbs 31
root@oxz_switch1:~# swadm link enable qsfp22/0
root@oxz_switch1:~# swadm link get -o L,M,e,P,l  qsfp22/0
Port/Link   Media    Enabled  PRBS    State
qsfp22/0    Optical  true     Mode31  Unknown
root@oxz_switch1:~# swadm link counters prbs qsfp22/0
Lane  Errors
0     129448048
1     130527952
2     129539888
3     129546256

wait a few seconds for the connection to get established and stable

root@oxz_switch1:~# swadm link counters prbs qsfp22/0
Lane  Errors
0     0
1     0
2     0
3     0
root@oxz_switch1:~#

Without modifying the SDE, we don't have enough information to calculate the error rate and/or total data being transferred. I have some SDE changes in flight and I could export that data if it would be useful.