DOCSIS 4.0 ESD Demo

[why_not]

Low latency DOCSIS hat nichts mit Segmentgrößen zutun....

LLD kommt nicht for free sondern benötigt signaling zwischen CMTS und CPEs, dementsprechend verringert sich die verfügbare Bandbreite mit jedem CPE mehr im Segment, daher lässt sich das nicht Sinnvoll einsetzen in großen Segmenten.

[Flole]

Und wie viel Overhead denkst du verursacht irgendein signaling bei LLD sodass es die derzeitigen Segmente an ihre Grenzen bringen würde? Und welches signaling meinst du überhaupt, könntest du vielleicht die korrekten Begriffe wie sie im Standard verwendet werden ebenfalls verwenden, dann weiß man auch was du meinst.

[why_not]

Und wie viel Overhead denkst du verursacht irgendein signaling bei LLD sodass es die derzeitigen Segmente an ihre Grenzen bringen würde? Und welches signaling meinst du überhaupt, könntest du vielleicht die korrekten Begriffe wie sie im Standard verwendet werden ebenfalls verwenden, dann weiß man auch was du meinst.

Bei Segment größen von 1200+ und den im PDF angestrebten Latenzen von ~1ms mehr als ein Segment an upload bereitstellt.
Also LLD mit PGS

[Flole]

Ein Segment stellt über 200Mbit/s an Upload bereit. Und das alles willst du nur für signaling benutzen? PGS zum Beispiel wäre ja aber auch kein signaling mehr....

[denunge]

There is some amount of overhead involved with PGS, because of the unsolicited nature of the grants (the modem may receive more grants than it can use = wasted). PGS is not a fixed and static thing, though. It is configurable by the operator (target latency). The CMTS can also optimize it, by looking at the traffic pattern from the modem.

PGS is optional. LLD can be activated without it (you cannot reach 1ms without it, though).

[Flole]

But that's not signaling traffic, PGS does not cause noticeable signaling traffic. Not to mention that proactively assigned timeslots will not need a Request-Cycle so that overhead is reduced there.

[denunge]

Yes, the bandwidth requests from the modem will be reduced a bit, but the amount of bandwidth needed for that is tiny compared to how much wasted grants can eat up.

DOCSIS has always (until now) been about using the bandwidth as optimally as possible (many modems must share not that much upstream bandwidth). For example, contention slots, where the modem can request bandwidth, is a shared ressource (bandwidth efficient). It also mean that these requests can collide with requests from other modems, so they must be retransmitted (all in the name of saving bandwidth).

[SenfKabelHer]

There is some amount of overhead involved with PGS, because of the unsolicited nature of the grants (the modem may receive more grants than it can use = wasted). PGS is not a fixed and static thing, though. It is configurable by the operator (target latency). The CMTS can also optimize it, by looking at the traffic pattern from the modem.

PGS is optional. LLD can be activated without it (you cannot reach 1ms without it, though).

GPON as it's been used and will be used for years to come cannot reach 1 ms either. It's consumer grade infrastructure we talk about. It's all about "good enough" and cost efficiency there. And that's why I assume docsis is just in the middle of its lifespan with another ≈20 years ahead.

[denunge]

To compare DOCSIS to GPON... GPON has always (until now) had more upstream bandwith than DOCSIS systems, and with fewer customers to share it, so they could implement a scheduling system from day 1 that was more latency efficient.

[SenfKabelHer]

Both are on the market and compete for customers that do compare. I for one have ftth access in my home, but I don't use it, docsis is enough for me (not enough as in I feel like giving up on something, but go for the eco option anyway, like economy class in a plane, but it has literally all I need).

The comparison is a market reality for customers. I just don't pay a "ftth premium" just like that.