Requested revision
Standard: | 802.1Qbv-2015 | Clause: | 12.29 |
Clause title: | Managed objects for scheduled traffic |
Rationale for revision
Some early implementations of IEEE Std 802.1Qbv-2015 are interpreting the times of these
managed objects as referencing the internal point of transmission selection. The delay from
this internal point to the actual point of transmit from the PHY can be significant. Since
the internal time is hidden from the management client, the delay is also hidden. This makes
it practically impossible for a management client to specify useful values for the managed objects.
In other words, under the interpretation that the managed objects represent a hidden point in time,
IEEE Std 802.1Qbv-2015 does not serve as a multi-vendor standard.
In the timestamping of IEEE Std 1588 and IEEE Std 802.1AS, as well as the latency and delay
metrics of P802.1Qcc, externally visible times are clearly specified as on-the-wire. Specifically,
these standards refer to the time that a point "passes the reference plane marking
the boundary between the network media and PHY". Many implementations of IEEE Std 802.1Qbv-2015
interpret the managed objects in this manner, and those implementations entail a proper standard.
It is worth noting that the internal registers of a Bridge will typically be programmed using
internal times. This does not contradict use of on-the-wire times in the managed objects.
The internal point of transmission selection will have variance (min/max). The delay from
the internal point to the on-the-wire point will also have variance (e.g. PHY delays).
The gate open/close times must account for such variances using a software computation
(i.e. addition/subtraction using worst-case values).
This maintenance request insists that the software computation must be performed by the
management entity in the Bridge, since that is the only entity with the essential
information.
Proposed text
Add a new subclause 12.29.2, titled "Representation of times", with the following text:
"The time values of 12.29.1 (e.g. AdminBaseTime) shall be represented at the point where
a frame passes the reference plane marking the boundary between the network media and PHY
(i.e. on-the-wire time point).
Figure 12-6 shows two points in time for scheduled traffic. The time of gate open/close
described in subclause 8.6.8.4 is used for transmission selection, and that point in time
exists internal to the Bridge (i.e. above MAC and PHY). A delay exists between this
internal time point and the time of transmit on-the-wire. For example, if transition of
a gate from closed to open immediately results in selection of a frame for transmission,
there will be a delay from that internal time of transmission selection to the start of the
frame on-the-wire. The internal time point will have variance (min/max), and the delay to
the on-the-wire time point will also have variance.
Since the Bridge contains the information needed in order to compute the min/max delay
from the internal time point to the on-the-wire time point, the Bridge is responsible
for performing the adjustments in order to represent the managed objects as on-the-wire
time.
NOTE: As an example, consider an IEEE Std 802.3 Port configured to transmit a single
traffic class with known frame lengths, and the managed objects are configured with
a cycle of open for one millisecond, then closed for one millisecond. If a continuous
burst of frames is transmitted through the Bridge to the Port, no symbol of a frame
can transmit from the PHY during the closed window (i.e. from start of Preable through
end of Frame Check Sequence)."
Insert a new Figure 12-6, titled "Points in time for scheduled traffic":
+------------------ -----------------------+
| Transmission selection for |
| scheduled traffic (8.6.8.4) | <- internal time point
| |
+---------------(EISS)---------------------+
| |
| Subclause 6.9 |
| |
+---------------(ISS)----------------------+
| |
| Subclause 8.5.1 |
| |
+---------------(ISS)----------------------+
| |
| MAC |
| |
Impact on existing networks
Originator
Name: | Rodney Cummings | Email: | Rodney.Cummings@ni.com |
Affiliation: | National Instruments | ||
Submitted: | 2017-07-05 |