|Standard:||IEEE Std. 802.1BA-2011||Clause:||6.5|
|Clause title:||Meeting latency targets for SR classes A and B|
Rationale for revision
IEEE 802.1Q has introduced the combination of CBS and TAS (compare Figure 8-14 in 802.1Q-2018). This has lead in turn to a re-calculation of the (oper)IdleSlope for CBS in clause 18.104.22.168 of 802.1Q-2018 (see under d) there). In effect it means that a closed TAS-Gate will hold up a CBS stream for an ‘arbitrary’ amount of time. This is not currently accounted for in the math of section 6.5 in 802.1BA. It further more results in “larger” bursts than was anticipated by the original CBS/.1BA considerations. Worst case it may mean, that a (dynamic) reservation request for a CBS queue, which is behind a TAS-Gate, must lead to a change in the TAS control-list. It furthermore means that the 75% recommendation of 34.3.1 in 802.1BA now only applies to the gate-open-time of the TAS-Gate. As one may want to compensate and simplify this configuration, it would be good if the math in section 6.5 of 802.1BA would differentiate between the actually used BW and the reserved BW.
Include the effects of TAS-Gate closure into the math of section 6.5 in 802.1BA. Allow used and reserved bandwidth to be different in the math of section 6.5 in 802.1BA. e.g. similar to this: Max Latency = tGate + tDevice + tMaxPacketSize+IPG +(tAllStreams – tStreamPacket+IPG) × Rate/MaxAllocBand + tStreamPacket tAllStreams = (UsedBandwidth × tInterval) / tRate tGate = ?tbd?
Impact on existing networks
Using IEEE Std. 802.1BA-2011 together with implementations of IEEE Std. 802.1Q-2018 may lead to unexpected latency behaviour.
|Affiliation:||Ethernovia, BMW, Ruetz System Solutions|