Buffer Occupancy #55

rsivakolundu · 2018-10-02T22:58:16Z

Added instruction bit for buffer occupancy

mhira1 · 2018-10-03T00:01:01Z

telemetry/specs/INT.mdk

+  different number of INT hops.
+
+* The fields in the INT metadata header are interpreted the following way:
+  - Ver (4b): INT metadata header version. Should be 1 for this version.


Version must be 2 in this paragraph

mhira1 · 2018-10-03T00:02:55Z

telemetry/specs/INT.mdk

+* Buffer occupancy
+  - The build-up of traffic in the shared buffer (in bytes, cells, or packets) that the
+INT packet observes in the device while being forwarded.
+


Perhaps remove the word "shared". Reporting both queue occupancy and buffer occupancy makes sense only if the buffer is shared, but technically the buffer could be a per-queue buffer, just that in this case, both values reported will be the same. INT source does not know whether downstream switches have shared buffers or not. It can set both bits. Switches that have per-queue buffers (unlikely, switches typically use shared buffers) will just report the same value for both instructions.

May be word this differently, to say this instruction is to get buffer occupancy. Use-case is when buffer is shared.

Removed the reference to shared.

Minor editorial rewording:
The use case is when the buffer is shared between multiple queues.

Added the minor editorial change suggested.

mhira1

Minor change needed.

mhira1 · 2018-10-03T17:55:06Z

telemetry/specs/INT.mdk

+      switch(es) set the M bit based on knowledge of the network topology
+      and "Switch ID, Ingress port ID, Egress port ID" tuples in the INT
+      metadata stack.
+  - R (9b): Reserved bits.


Number of reserved bits is still 10 correct? I think this text with 9 bits is coming from your other patch in which you are using one of the reserved bits for "Sink" bit. There is no change in number of reserved bits between v1 and v2 in this patch.

jklr · 2018-10-03T18:47:27Z

telemetry/specs/INT.mdk

@@ -910,6 +915,170 @@ from (shim header length \* 4).
 For INT over Geneve it is 8 bytes subtracted from (length in Geneve tunnel
 option header \* 4).

+
+INT Metadata Header and Metadata Stack (Version = 2):


@mhira1 I was meaning to propose make a 1.x version cut before merging these changes, and move INT.mdk to version2 draft. Then we don't need to list both v1 and v2 in the same doc. How do you think?

Makes sense. Let's remove this second paragraph on v2 header then.

rsivakolundu · 2018-10-03T22:26:22Z

Ok. I will add buffer occupancy to ver 1 and remove ver 2 details.

…

-Ramesh

On Wed, Oct 3, 2018 at 12:11 PM Mukesh Hira ***@***.***> wrote: ***@***.**** commented on this pull request. ------------------------------ In telemetry/specs/INT.mdk <#55 (comment)>: > @@ -910,6 +915,170 @@ from (shim header length \* 4). For INT over Geneve it is 8 bytes subtracted from (length in Geneve tunnel option header \* 4). + +INT Metadata Header and Metadata Stack (Version = 2): Makes sense. Let's remove this second paragraph on v2 header then. — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#55 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/APOTlnEMnBSG0q12GaH5r6WnsvltcRAuks5uhQvjgaJpZM4XFAqV> .

jklr · 2018-10-03T22:35:28Z

@rsivakolundu hold on, pls. This buffer occupancy change is NOT backward compatible, right? Hence it should be added only to v2. Let's discuss version maintenance policy tomorrow and then make further changes on the PRs. I don't want you to waste time during PTO.

mickeyspiegel · 2018-10-04T02:34:59Z

telemetry/specs/INT.mdk

+* Buffer occupancy
+  - The build-up of traffic in the shared buffer (in bytes, cells, or packets) that the
+INT packet observes in the device while being forwarded.
+


Minor editorial rewording:
The use case is when the buffer is shared between multiple queues.

mickeyspiegel · 2018-10-04T02:35:49Z

telemetry/specs/INT.mdk

@@ -1279,6 +1289,7 @@ struct headers {
    int_egress_tstamp_t         int_egress_tstamp;
    int_level2_port_ids_t       int_level2_port_ids;
    int_egress_port_tx_util_t   int_egress_port_tx_util;
+    int_b_occupancy_t		int_b_occupancy;


nit: fix indent

mickeyspiegel · 2018-10-04T02:36:23Z

telemetry/specs/INT.mdk

@@ -1610,6 +1621,11 @@ control EgressDeparserImpl(packet_out packet,
            ck.add({hdr.int_egress_port_tx_util.egress_port_tx_util});
        }

+ 	if (hdr.int_b_occupancy.isValid()) {


nit: fix indent

mickeyspiegel · 2018-10-04T02:36:34Z

telemetry/specs/INT.mdk

@@ -1656,6 +1672,7 @@ control EgressDeparserImpl(packet_out packet,
        packet.emit(hdr.int_egress_tstamp);
        packet.emit(hdr.int_level2_port_ids);
        packet.emit(hdr.int_egress_port_tx_util);
+	packet.emit(hdr.int_b_occupancy);


nit: fix indent

Fixed minor editorial comment from Mickey

mhira1

Ramesh, I looked through the edits again, marked some typos. Also, at the previous meeting, we decided to include both buffer ID and buffer occupancy into the 32-bit space. Please make that change as well.

telemetry/specs/INT.mdk

mhira1 · 2018-10-16T22:33:28Z

telemetry/specs/INT.mdk

+  - The build-up of traffic in the buffer (in bytes, cells, or packets) that the
+INT packet observes in the device while being forwarded. Use case is when buffer is
+shared between multiple queue.
+* Instantaneous buffer occupance


typo: occupance -> occupancy

telemetry/specs/INT.mdk

Added some text to queue occupancy and buffer occupancy description about how the YANG model shall define the format and units of these fields to enable switch implementation specific details.

jklr

@rsivakolundu thanks for the PR. I suggested several editorial changes. Please check the last two nit comments from Mukesh.

telemetry/specs/INT.mdk

jklr · 2018-12-04T18:00:59Z

telemetry/specs/INT.mdk

@@ -1219,6 +1229,10 @@ header int_egress_port_tx_util_t {
    bit<32> egress_port_tx_util;
 }

+header int_b_occupancy_t {
+    bit<32> q_occupancy;


Please make it consistent with the (example) layout provided in line 866, which has 8b buffer id.

Let us discuss the semantics of Queue and Buffer occupancy. I will make it consistent after the discussion.

Added Buffer ID (8 bits) and Buffer Occupancy (24 bits)

telemetry/specs/INT.mdk

Addressed some of the comments and changes proposed. Semantics of the queue and buffer occupancy still need to be addressed.

Changes to address all review comments.

Corrected a line spacing issue

jklr · 2018-12-06T21:35:39Z

@rsivakolundu your latest commit merged domain specific extension to this branch/PR!

rsivakolundu · 2018-12-06T22:39:21Z

@jk - I don't see it in my branch that way.

…

-Ramesh

On Thu, Dec 6, 2018 at 1:35 PM JK Lee ***@***.***> wrote: @rsivakolundu <https://github.com/rsivakolundu> your latest commit merged domain specific extension to this branch/PR! — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#55 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/APOTlvrMZOGqzRitwlRhnS6cGJ0eSVmMks5u2Y2tgaJpZM4XFAqV> .

jklr · 2018-12-06T23:35:25Z

@rsivakolundu hmm I think there is a github bug. The bottom most commit in this PR's commit tab showed domain specific changes 2 hours ago. I double checked at that time.
It's not like that any more.

jk · 2018-12-07T07:32:52Z

@rsivakolundu @jklr != @jk

rsivakolundu added 10 commits June 21, 2018 15:03

Update INT.mdk

3771848

Update INT.mdk

10355fd

Update INT.mdk

fe346e4

Update INT.mdk

6f4c8e8

Update INT.mdk

2e253b0

Update INT.mdk

34b55ff

Update INT.mdk

54c9158

Update INT.mdk

3289ea4

Update INT.mdk

c32c475

Update INT.mdk

3c0cdd3

mhira1 reviewed Oct 3, 2018

View reviewed changes

Update INT.mdk

84f520a

mhira1 reviewed Oct 3, 2018

View reviewed changes

mhira1 requested review from mickeyspiegel, mhira1 and jklr October 3, 2018 18:26

jklr reviewed Oct 3, 2018

View reviewed changes

Update INT.mdk

63ce414

mickeyspiegel reviewed Oct 4, 2018

View reviewed changes

rsivakolundu changed the title ~~Shared buffer~~ Buffer Occupancy Oct 4, 2018

rsivakolundu added 3 commits October 3, 2018 21:29

Update INT.mdk

bc3d729

Fixed minor editorial comment from Mickey

Update INT.mdk

7131207

Update INT.mdk

748ca28

mhira1 reviewed Oct 16, 2018

View reviewed changes

rsivakolundu added 2 commits October 18, 2018 09:04

Added Buffer ID field

ae57749

Update INT.mdk

5b3a6f5

Added some text to queue occupancy and buffer occupancy description about how the YANG model shall define the format and units of these fields to enable switch implementation specific details.

jklr approved these changes Dec 4, 2018

View reviewed changes

Update INT.mdk

19a6456

Addressed some of the comments and changes proposed. Semantics of the queue and buffer occupancy still need to be addressed.

rsivakolundu added 2 commits December 6, 2018 11:57

Update INT.mdk

6f725b4

Changes to address all review comments.

Update INT.mdk

5f7ac99

Corrected a line spacing issue

jklr merged commit 1ce57dd into p4lang:master Dec 8, 2018

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Buffer Occupancy #55

Buffer Occupancy #55

rsivakolundu commented Oct 2, 2018 •

edited

Loading

mhira1 Oct 3, 2018

mhira1 Oct 3, 2018

mhira1 Oct 3, 2018

rsivakolundu Oct 3, 2018

mickeyspiegel Oct 4, 2018

rsivakolundu Oct 4, 2018

mhira1 left a comment

mhira1 Oct 3, 2018

jklr Oct 3, 2018 •

edited

Loading

mhira1 Oct 3, 2018

rsivakolundu commented Oct 3, 2018 via email

jklr commented Oct 3, 2018

mickeyspiegel Oct 4, 2018

mickeyspiegel Oct 4, 2018

rsivakolundu Oct 4, 2018

mickeyspiegel Oct 4, 2018

rsivakolundu Dec 6, 2018

mickeyspiegel Oct 4, 2018

rsivakolundu Oct 4, 2018

mhira1 left a comment

mhira1 Oct 16, 2018

rsivakolundu Dec 6, 2018

jklr left a comment

jklr Dec 4, 2018

rsivakolundu Dec 4, 2018

rsivakolundu Dec 6, 2018

jklr commented Dec 6, 2018

rsivakolundu commented Dec 6, 2018 via email

jklr commented Dec 6, 2018

jk commented Dec 7, 2018 •

edited

Loading

Buffer Occupancy #55

Buffer Occupancy #55

Conversation

rsivakolundu commented Oct 2, 2018 • 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

mhira1 left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

jklr Oct 3, 2018 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

rsivakolundu commented Oct 3, 2018 via email

jklr commented Oct 3, 2018

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

mhira1 left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

jklr left a 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

jklr commented Dec 6, 2018

rsivakolundu commented Dec 6, 2018 via email

jklr commented Dec 6, 2018

jk commented Dec 7, 2018 • edited Loading

rsivakolundu commented Oct 2, 2018 •

edited

Loading

jklr Oct 3, 2018 •

edited

Loading

jk commented Dec 7, 2018 •

edited

Loading