Add Performance monitoring update to CMIS_and_C-CMIS_support_for_ZR.md #1258

jaganbal-a · 2023-02-09T22:37:17Z

Adding performance monitoring HLD .
Please refer to below code PR for the implementation
sonic-platform-common: sonic-net/sonic-platform-common#402
sonic-platform-daemons :sonic-net/sonic-platform-daemons#390
sonic-utilities : sonic-net/sonic-utilities#2927

linux-foundation-easycla · 2023-02-09T22:37:21Z

The committers listed above are authorized under a signed CLA.

✅ login: jaganbal-a (ee69977, 0604af7, 64c949f, e5e1e63, 4578e77, 4a1663e, 09b6e3e, 6c31d76, d8e372c, f04b9b1, 5955831, d455c0e, f8c389c, b7ed93d, 8f53d55, 491b4f2, d036420, 5e59008, 213872f)

prgeor · 2023-03-01T19:03:34Z

@keboliu @mihirpat1 @andywongarista please review

prgeor · 2023-03-01T19:09:10Z

@qinchuanares Please review

prgeor · 2023-03-01T19:15:40Z

@jaganbal-a you need to sign the EASY CLA

doc/platform_api/CMIS_and_C-CMIS_support_for_ZR.md

zhangyanzhao · 2023-04-17T06:29:30Z

The committers listed above are authorized under a signed CLA.

✅ login: jaganbal-a (ee69977, 0604af7, 64c949f, e5e1e63, 4578e77, 4a1663e, 09b6e3e)

doc/platform_api/CMIS_and_C-CMIS_support_for_ZR.md

zhangyanzhao · 2023-04-17T23:06:43Z

The code PRs need be added to this HLD PR

zhangyanzhao · 2023-04-18T15:23:36Z

community review recording https://zoom.us/rec/share/p-EqFKsSHinsWM0mPs0sPlFURq7mCmDVw1URjKLbpcwv3Ib0FVei-9LpTvFubNo.M3RwumppZ9d6y0Uc

jaganbal-a · 2023-04-18T16:25:17Z

Review comments from the community review.

Give platform with an option of 30,60 and 120 sec of PM time interval for the statistics collection.
Jagan >Added 120sec for PM interval in the design with a caveat (60sec PM time window option in the CLI will not provide any output for the platform which chooses PM interval as 120sec).
Add PM window slot schema for the state-DB
Jagan> Added.
Add an example of how the data is sampled by xcvrd for 15min/24hr window slots.
Jagan> Added
With fixed window, at 24hr completion the sampled data in the 24hr window will reset, so at any given point the 24hrs data will not be maintained. So move the 24hr window to a moving window or maintain 3hr window of 8slots (this will allow atleast 21hrs of data at any point of time)
Jagan> 1 more slot added for 24 Hrs PM time window. So at any given time 24Hrs statistics will be available.
Add yang model for the global config CLI to enable/disable the PM statistics collection.
Jagan> It is pending, will add it.
Are these sampling windows sliding or not? >> it seems the windows are fixed, is this possible to extend it for sliding (ex: last 23 hrs)
Jagan> The time windows are fixed. the statistics are reset for every fixed window and the purpose of having fixed window is to understand the failure to a specific time window.
Can the HLD allow multiple sampling windows to capture the PM statistics ?
Jagan> The statistics can be collected only for a interval of time from the transceiver. So here the interval is defined by the host and updated to 60sec/15min and 24Hr time window slots. Hope this answer the question.
What are the use cases trying to solve this problem?
Jagan> Added to the overview section.
What is the SONiC default behaviour if customers don't use this feature?
Jagan> There will be no change.
What is the CPU penalty to requirements to support this sampling feature with SONiC?
Jagan> The sampling of data from the collected statistics will not add any overhead to the CPU.
How does a SONiC user know whether the platform driver has this capability?Sampling - is the driver dropping any data points or sending everything?
Jagan> User need to enable the global CLI to start the functionality and it start kicks in only when there is a 400G-ZR module present in the system/box. Driver not supposed to drop any data as per the design and update the statistics to the DB.
Where is this sampling implemented?
Jagan> It is in xcvrd process.
What must be supported by each ASIC driver?
Jagan>There is no involvement of ASIC driver.

Addressing the OCP review comments

zhangyanzhao

@jaganbal-a can you please help to add the code PRs? Thanks.

jaganbal-a · 2023-05-10T16:37:00Z

@jaganbal-a can you please help to add the code PRs? Thanks.

@zhangyanzhao , I am working on the code changes and I will post once I am complete with unit testing. But this PR is for the high level design. so please approve if you do not have any further question on the HLD.

7.Performance Monitoring - 400-G ZR module

bmridul · 2023-05-24T22:41:42Z

@lguohan , Pls review the document.

zhangyanzhao · 2023-06-14T04:51:50Z

@jaganbal-a code PRs need be included in the HLD before HLD can be merged, can you please help to add the code PRs? Thanks.

Improvement

Update to the flow diagram

Refer to the HLD for problem description and design. HLD: sonic-net/SONiC#1258

prgeor · 2023-09-12T22:12:20Z

doc/platform_api/CMIS_and_C-CMIS_support_for_ZR.md

+
+#show int trans pm history 15min window
+commands:
+1 - 11 PM window number


@jaganbal-a what would the output be for window number history for which sample is yet not collected.? Say after boot, 15 minutes has passed and user wants to see window# 11 for 15min window

root@sonic:/home/cisco# show int trans pm history 15min window 11 -n asic2 Ethernet232
Tue Oct 24 17:33:41 UTC 2023
PM window: 15min
Ethernet232: Transceiver performance monitoring data not available for the requested window

prgeor · 2023-09-12T22:13:24Z

doc/platform_api/CMIS_and_C-CMIS_support_for_ZR.md

+
+#show int trans pm history 60sec window
+commands:
+1 - 14 PM window number


@jaganbal-a can you show the calculation for the window range. Why 1-14? Why not 1-20?

For 60sec window size, by design 15 windows are allocated.
In that, 14 windows are used for history -numbering from 1 to 14 and 1 window is used for viewing current statistics.

For 15min window size, 12 windows are allocated.
In that, 11 windows are used for history and 1 window is used for current.

For 24Hrs window size, 2 windows are allocated.
In that, 1 window used for history and another window used for current.

Mentioned here : https://github.com/sonic-net/SONiC/blob/d0364205764aa49decfca1eb669c76d7f0c8f4e0/doc/platform_api/CMIS_and_C-CMIS_support_for_ZR.md#pm-time-window-or-pm-window-

prgeor · 2023-09-12T22:17:20Z

doc/platform_api/CMIS_and_C-CMIS_support_for_ZR.md

+
+##### 7.5.1 Configurations
+1. performance monitor enable - Global CLI to enable PM on all ports. 
+Before this configuration get implemented, PM will be enabled by default on all ports


@jaganbal-a is this future implementation? i.e per port configuration?

@prgeor , Yes, this will future implementation. Global CLI to enable/disable PM on all ports.
Raised a github issue :- sonic-net/sonic-platform-daemons#402

prgeor · 2023-09-12T22:21:17Z

doc/platform_api/CMIS_and_C-CMIS_support_for_ZR.md

+- 60sec and
+- 120sec
+
+It is recommended to choose 30sec, platforms that have high CPU load can choose 120sec as PM interval. By default 60sec is the PM interval when no input provided by platform, please refer '7.5.4' for platform input.


@jaganbal-a can you provide a hyper link to section 7.5.4 https://stackoverflow.com/questions/2822089/how-to-link-to-part-of-the-same-document-in-markdown

changed to hyperlink

prgeor · 2023-09-12T22:35:25Z

doc/platform_api/CMIS_and_C-CMIS_support_for_ZR.md

+The 60sec time window are filled as it is read from module if the PM interval is 60sec, the 60sec sample collected from module is then sampled for 15mins and 24Hr window every 60seconds.
+
+
+<img src ="https://user-images.githubusercontent.com/97986478/236910604-bbb52e77-5c62-4a7e-b64d-ff883d9b57d3.png" width=60% height=50%>


@jaganbal-a why is the avg rx power remain constant at -10.61

corrected the avg in the example and added pm window end time.

prgeor · 2023-09-12T22:37:03Z

doc/platform_api/CMIS_and_C-CMIS_support_for_ZR.md

+        sampled_pm_dict['prefec_ber_min'] = min(float(pm_data_dict1['prefec_ber_min']), float(pm_data_dict2['prefec_ber_min']))
+        sampled_pm_dict['prefec_ber_max'] = max(float(pm_data_dict1['prefec_ber_max']), float(pm_data_dict2['prefec_ber_max']))
+
+        sampled_pm_dict['uncorr_frames_avg'] = self.average_of_two_val(float(pm_data_dict1['uncorr_frames_avg']), float(pm_data_dict2['uncorr_frames_avg']))


@jaganbal-a one example for min, max, avg is sufficient. We dont have to show for all PM parameters

prgeor · 2023-09-19T17:23:03Z

doc/platform_api/CMIS_and_C-CMIS_support_for_ZR.md

+Each window field(Key) in TRANSCEIVER_PM_WINDOW_STATS have a value string comprises of following fields.
+
+	pm_win_start_time           = 1*255VCHAR                       ; PM statistics start time for the window.
+	pm_win_end_time             = 1*255VCHAR                       ; PM statistics end time for the window. 
+        pm_win_current               = 1*255VCHAR                       ; PM statistics collection is Progressing on this window. (True/False)
+        prefec_ber_avg               = FLOAT                            ; prefec ber avg
+        prefec_ber_min               = FLOAT                            ; prefec ber min
+        prefec_ber_max               = FLOAT                            ; prefec ber max
+        cd_avg                       = FLOAT                            ; chromatic dispersion avg
+        cd_min                       = FLOAT                            ; chromatic dispersion min
+        cd_max                       = FLOAT                            ; chromatic dispersion max
+        dgd_avg                      = FLOAT                            ; differential group delay avg
+        dgd_min                      = FLOAT                            ; differential group delay min
+        dgd_max                      = FLOAT                            ; differential group delay max
+        sopmd_avg                    = FLOAT                            ; second order polarization mode dispersion avg
+        sopmd_min                    = FLOAT                            ; second order polarization mode dispersion min
+        sopmd_max                    = FLOAT                            ; second order polarization mode dispersion max
+        pdl_avg                      = FLOAT                            ; polarization dependent loss avg
+        pdl_min                      = FLOAT                            ; polarization dependent loss min
+        pdl_max                      = FLOAT                            ; polarization dependent loss max
+        osnr_avg                     = FLOAT                            ; optical signal to noise ratio avg
+        osnr_min                     = FLOAT                            ; optical signal to noise ratio min
+        osnr_max                     = FLOAT                            ; optical signal to noise ratio max
+        esnr_avg                     = FLOAT                            ; electrical signal to noise ratio avg
+        esnr_min                     = FLOAT                            ; electrical signal to noise ratio min
+        esnr_max                     = FLOAT                            ; electrical signal to noise ratio max
+        cfo_avg                      = FLOAT                            ; carrier frequency offset avg


@jaganbal-a we don't need to copy these again from section 2.1.5. A reference to 2.1.5 is sufficient. can you remove this ?

removed the redundant fields and added link.

jaganbal-a · 2023-10-19T19:26:11Z

@jaganbal-a can you please help to add the code PRs? Thanks.

@zhangyanzhao
sonic-platform-common: sonic-net/sonic-platform-common#402
sonic-platform-daemons :sonic-net/sonic-platform-daemons#390
sonic-utilities : sonic-net/sonic-utilities#2927

jaganbal-a · 2023-10-20T15:19:58Z

@jaganbal-a can you please help to add the code PRs? Thanks.

@zhangyanzhao , I am working on the code changes and I will post once I am complete with unit testing. But this PR is for the high level design. so please approve if you do not have any further question on the HLD.

please check the description for the code PRs.

prgeor self-assigned this Mar 1, 2023

prgeor requested review from keboliu and mihirpat1 March 1, 2023 19:03

keboliu reviewed Mar 6, 2023

View reviewed changes

qinchuanares reviewed Mar 17, 2023

View reviewed changes

doc/platform_api/CMIS_and_C-CMIS_support_for_ZR.md Outdated Show resolved Hide resolved

doc/platform_api/CMIS_and_C-CMIS_support_for_ZR.md Outdated Show resolved Hide resolved

doc/platform_api/CMIS_and_C-CMIS_support_for_ZR.md Outdated Show resolved Hide resolved

qinchuanares reviewed Mar 17, 2023

View reviewed changes

doc/platform_api/CMIS_and_C-CMIS_support_for_ZR.md Outdated Show resolved Hide resolved

qinchuanares reviewed Mar 17, 2023

View reviewed changes

doc/platform_api/CMIS_and_C-CMIS_support_for_ZR.md Outdated Show resolved Hide resolved

jaganbal-a added 3 commits April 16, 2023 23:28

Update CMIS_and_C-CMIS_support_for_ZR.md

ee69977

Update CMIS_and_C-CMIS_support_for_ZR.md

0604af7

Update CMIS_and_C-CMIS_support_for_ZR.md

64c949f

zhangyanzhao force-pushed the patch-5 branch from 2f10e5f to 64c949f Compare April 17, 2023 06:28

eddyk-nvidia reviewed Apr 17, 2023

View reviewed changes

doc/platform_api/CMIS_and_C-CMIS_support_for_ZR.md Show resolved Hide resolved

jaganbal-a added 3 commits April 18, 2023 10:53

Update CMIS_and_C-CMIS_support_for_ZR.md

e5e1e63

Update CMIS_and_C-CMIS_support_for_ZR.md

4578e77

Update CMIS_and_C-CMIS_support_for_ZR.md

4a1663e

Update CMIS_and_C-CMIS_support_for_ZR.md

09b6e3e

Addressing the OCP review comments

zhangyanzhao requested changes May 9, 2023

View reviewed changes

jaganbal-a closed this May 10, 2023

jaganbal-a reopened this May 10, 2023

Update CMIS_and_C-CMIS_support_for_ZR.md

6c31d76

7.Performance Monitoring - 400-G ZR module

jaganbal-a added 9 commits June 28, 2023 11:24

Update CMIS_and_C-CMIS_support_for_ZR.md

d8e372c

Improvement

Update CMIS_and_C-CMIS_support_for_ZR.md

f04b9b1

Update CMIS_and_C-CMIS_support_for_ZR.md

5955831

Update CMIS_and_C-CMIS_support_for_ZR.md

d455c0e

Update CMIS_and_C-CMIS_support_for_ZR.md

f8c389c

Update CMIS_and_C-CMIS_support_for_ZR.md

b7ed93d

Update CMIS_and_C-CMIS_support_for_ZR.md

8f53d55

Update CMIS_and_C-CMIS_support_for_ZR.md

491b4f2

Update to the flow diagram

Update CMIS_and_C-CMIS_support_for_ZR.md

d036420

jaganbal-a mentioned this pull request Jul 18, 2023

Fix VDM freeze and unfreeze needed for PM stats collection #386 sonic-net/sonic-platform-common#387

Closed

jaganbal-a pushed a commit to jaganbal-a/PM-stats-sonic-platform-daemons that referenced this pull request Jul 21, 2023

Signed-off-by: Jaganathan Anbalagan <jaganbal@cisco.com>

b66a0f9

Refer to the HLD for problem description and design. HLD: sonic-net/SONiC#1258

jaganbal-a mentioned this pull request Jul 21, 2023

400G-ZR : Time window based PM statistics collection sonic-net/sonic-platform-daemons#390

Open

longhuan-cisco mentioned this pull request Jul 26, 2023

Add transceiver pm CLI for advanced window-based performance monitoring sonic-net/sonic-utilities#2927

Open

jaganbal-a requested review from zhangyanzhao, qinchuanares, keboliu and eddyk-nvidia August 18, 2023 18:55

prgeor reviewed Oct 18, 2023

View reviewed changes

jaganbal-a mentioned this pull request Oct 19, 2023

Implement Global CLI to enable performance monitoring for coherent optics. sonic-net/sonic-platform-daemons#402

Open

jaganbal-a closed this Oct 20, 2023

jaganbal-a reopened this Oct 20, 2023

jaganbal-a added 2 commits October 25, 2023 13:23

Update CMIS_and_C-CMIS_support_for_ZR.md

5e59008

Update CMIS_and_C-CMIS_support_for_ZR.md

213872f

jaganbal-a requested a review from prgeor October 25, 2023 21:58

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Add Performance monitoring update to CMIS_and_C-CMIS_support_for_ZR.md #1258

Add Performance monitoring update to CMIS_and_C-CMIS_support_for_ZR.md #1258

jaganbal-a commented Feb 9, 2023 •

edited

Loading

linux-foundation-easycla bot commented Feb 9, 2023 •

edited

Loading

prgeor commented Mar 1, 2023

prgeor commented Mar 1, 2023

prgeor commented Mar 1, 2023

zhangyanzhao commented Apr 17, 2023 •

edited by linux-foundation-easycla bot

Loading

zhangyanzhao commented Apr 17, 2023

zhangyanzhao commented Apr 18, 2023

jaganbal-a commented Apr 18, 2023 •

edited

Loading

zhangyanzhao left a comment

jaganbal-a commented May 10, 2023

bmridul commented May 24, 2023

zhangyanzhao commented Jun 14, 2023

prgeor Sep 12, 2023

jaganbal-a Oct 24, 2023

prgeor Sep 12, 2023

jaganbal-a Oct 25, 2023 •

edited

Loading

prgeor Sep 12, 2023

jaganbal-a Oct 19, 2023

prgeor Sep 12, 2023

jaganbal-a Oct 25, 2023

prgeor Sep 12, 2023

jaganbal-a Oct 25, 2023

jaganbal-a Oct 25, 2023

prgeor Sep 12, 2023

prgeor Sep 19, 2023

jaganbal-a Oct 25, 2023

jaganbal-a commented Oct 19, 2023

jaganbal-a commented Oct 20, 2023

		The 60sec time window are filled as it is read from module if the PM interval is 60sec, the 60sec sample collected from module is then sampled for 15mins and 24Hr window every 60seconds.


		<img src ="https://user-images.githubusercontent.com/97986478/236910604-bbb52e77-5c62-4a7e-b64d-ff883d9b57d3.png" width=60% height=50%>

Add Performance monitoring update to CMIS_and_C-CMIS_support_for_ZR.md #1258

Are you sure you want to change the base?

Add Performance monitoring update to CMIS_and_C-CMIS_support_for_ZR.md #1258

Conversation

jaganbal-a commented Feb 9, 2023 • edited Loading

linux-foundation-easycla bot commented Feb 9, 2023 • edited Loading

prgeor commented Mar 1, 2023

prgeor commented Mar 1, 2023

prgeor commented Mar 1, 2023

zhangyanzhao commented Apr 17, 2023 • edited by linux-foundation-easycla bot Loading

zhangyanzhao commented Apr 17, 2023

zhangyanzhao commented Apr 18, 2023

jaganbal-a commented Apr 18, 2023 • edited Loading

zhangyanzhao left a comment

Choose a reason for hiding this comment

jaganbal-a commented May 10, 2023

bmridul commented May 24, 2023

zhangyanzhao commented Jun 14, 2023

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

jaganbal-a Oct 25, 2023 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

jaganbal-a commented Oct 19, 2023

jaganbal-a commented Oct 20, 2023

jaganbal-a commented Feb 9, 2023 •

edited

Loading

linux-foundation-easycla bot commented Feb 9, 2023 •

edited

Loading

zhangyanzhao commented Apr 17, 2023 •

edited by linux-foundation-easycla bot

Loading

jaganbal-a commented Apr 18, 2023 •

edited

Loading

jaganbal-a Oct 25, 2023 •

edited

Loading