XGS-PON fiber: How does rationing work?

Werner · Jun 27, 2023

XGS-PON fiber: How does rationing work?

To get it straight:

In normal daily operation within an The customers connected to the respective splitter demand more bandwidth than the PON tree as a whole is able to provide.

As a rule, most of the time you are allocated far more than 1/32 of the total bandwidth and sometimes you even get almost the entire bandwidth of the entire PON group, which you then use up to 8.3 Gbit As an individual you can use your own individual OTO can almost exclusively at times.

However, due to large consumers or speed test junkies, it can of course occasionally happen that the freely available group bandwidth reaches its limits and then forced rationing of the bandwidth allocation to individual connections can no longer be avoided.

My question is not about the different technical implementation of the rationing in the download and the upload (in the download there is probably a changed mix in the composition of the stream for broadcast transmission, and in the upload there is probably a reduction in the time slice allocation per individual router), but rather on the Rationing mechanism in general.

Basically I think there are two possibilities:

Variant A: At the load limit, only the currently largest consumer is slowed down and all other customers continue to receive the power they are currently requesting at that time

Variant B: All customers of the PON tree are slowed down proportionally, as a percentage of their currently requested service

But there are probably also other variants and hybrid forms of it.

Does anyone here know how Swisscom has actually implemented its securely existing XGS-PON rationing mechanism?

And perhaps a second question: Is there something like a guaranteed technical minimum bandwidth for every XGS-PON router that wants to start communicating with the PON tree?

(The background to this question is the assumption that this actually has to exist, otherwise a single router that just wants to start an upload stream to the PON tree would no longer receive any free time slot to transmit its signal)

P.S.: Personally, I would actually prefer variant A, because it is polluter-pays and also prevents collateral “damage” to uninvolved customers from excessive speed test junkies or other excessive users in the same PON tree.

And as an appendix for those interested in low-level PON, a technical basic paper from Huawei on the accompanying XGS-PON features (particularly interesting are Chapter 5 “Working Principle” and Chapter 6 “Key Technologies”, which are also available from Swisscom -XGS-PON can be used at least partially in communications technology:

https://actfornet.com/ueditor/php/upload/file/20200407/1586223371421399.pdf

user109 · Jun 27, 2023

@Werner probably works like the PUS. There are only one or two fiber connections with 1Gbit/s with 255 XDSL ports. I assume that it works according to the polluter pays principle.

Let’s ask @Roger G

Roger G · Jun 28, 2023

@Werner Wow, a lot to read 😀 So a little shorter:

A PON tree has up to 64 users (not 32)
There is no restriction or rationing (today).

Small example with only 2 users on a tree: If one constantly downloads the Internet at, for example, 8gbps (constant 8gbps, so completely hypothetical) and a second one does a speed test, then they share the entire bandwidth. This means that the downloader will only have a speed of 4gbps during the speed test and the tester will only measure 4gbps.

There are already almost full PON trees, but even with so many active users, the reduced BB is only visible in the speed test. If you do the test a few minutes later, the result looks completely different. If there are streamers that constantly transmit several gbps (DS and/or US), there may be permanent restrictions. This then regulates the general terms and conditions and additional provisions.

Werner · Jun 28, 2023

@Roger G

Thanks for your feedback.

Do I now understand your example correctly that this means:

If a PON tree were actually overloaded, up to 64 customers would have to suffer proportionately in solidarity?

Roger G · Jun 28, 2023

@Werner wrote:

@Roger G

Thanks for your feedback.

Do I now understand your example correctly that this means:

If a PON tree were actually overloaded, up to 64 customers would have to suffer proportionately in solidarity?

Yes of course, but that has been the case on copper/coax for x years. Even BX has an uplink behind the FAN that limits the available bandwidth.

5018 · Jun 28, 2023

There has probably never been anyone who complained in real life. Except for certain fan boys, whose utilization according to Twitter is always at 100%. I would think of a few more # tags, but I’m avoiding them for child protection reasons.

GrandDixence · Jun 28, 2023

@Werner wrote:

If a PON tree were actually overloaded, up to 64 customers would have to suffer proportionately in solidarity?

What else can you expect from a “shared medium” that works with time slots? In overload situations, the time slots of the “Shared Medium” are divided as fairly as possible and distributed proportionately to all network participants. This is how time slot systems such as mobile communications and EuroDOCSIS (Internet connection via a television cable network) work. In the case of mobile communications, there is also a prioritization of well-paying customers. In overload situations, customers with more expensive subscriptions are allocated more time slots than customers with cheaper subscriptions.

Internet connections via fiber optics with PON (e.g. XGS-PON) are always a “shared medium” and have the corresponding disadvantages of a “shared medium”. See:

[https://mobilecommunity.ch/wbb/index.php?thread/326-salt-fiber-oder-salt-unlimited-surf-f%C3%BCr-heimnetzwerk/&postID=2526#post2526] (https://mobilecommunity.ch/wbb/index.php?thread/326-salt-fiber-oder-salt-unlimited-surf-f%C3%BCr-heimnetzwerk/&postID=2526#post2526)

Werner · Jun 28, 2023

@Anonymous

It is clear to me that bandwidths at the various nodes have always been and will continue to be limited, and that in any overall network design you still have to make additional assumptions about the oversubscription factors that are just acceptable at each node.

But what is particularly interesting with a POM tree is that for the upload, unlike the download, a common broadcast signal cannot simply be sent to all connected routers, but a centrally regulated time slice allocation mechanism is required for each individual router anyway. to prevent more than one individual router from sending an IP packet upstream at the same time, because if this were allowed in the PON architecture, the individual upload signals would be different from several Routers in the optical splitter overlap each other and ultimately cause the entire upload to collapse in the entire PON tree.

So there must undoubtedly be a central allocation mechanism in every POM tree (including Swisscom’s XGS-PON) on the OLT (or perhaps one level higher in the network), which, at least in the upload, provides the individual routers with their allocated individual bandwidths directly actively controls.

So nothing about queues, or everyone on the PON tree is simply a little slower, but purely due to the PON architecture, there must be an active rule-based control in the upload.

My question is now very simple:

According to which rules is the mandatory bandwidth allocation for individual customers implemented in Swisscom’s XGS-PON and are these also fair from the perspective of the “PON community” of an individual PON tree?

By the way, what would certainly work in the upload would be to simply allocate to each PON participant their purely mathematical share of total bandwidth/number of participants, which would then, however, mean that XGS-PON would never be faster in the upload for the individual participant approx. 140 Mbit/sec., which is definitely not the case in reality.

This simply justifies the fact that there is certainly a centrally controlled active resource allocation management in the POM tree, which actively and continuously regulates which individual router is allowed to actively use how many upstream timeslots at any given time.

GrandDixence · Jun 28, 2023

In the

https://www.itu.int/rec/T-REC-G.9807.1

I haven’t read through this material and tried to understand it because the topic doesn’t interest me enough…

Herby · Jun 28, 2023

Oh @Werner

Do you really think that @Roger G will give “your answer” here?

Since it is still public 😉

Does every user need the full “bandwidth/speed” on the Internet or only when needed?

Which should probably be guaranteed by this.

OT: Sorry

IMO:

Evolution.

I call it “Maslow’s pyramid”.

We “here” are already way beyond that.

That’s why such discussions can/do take place at all.

I’m just wondering when the time will come until…

…and only then can we continue.

Werner · Jun 28, 2023

@5018 wrote:

There has probably never been anyone who has complained in real life. Except for certain fan boys, whose utilization according to Twitter is always at 100%. I would think of a few more # tags, but I’m avoiding them for child protection reasons.

Now just straight from business administration student to business administration student, I actually spontaneously perceive your statement as a purely polemical comment, sorry for that.

It has absolutely nothing to do directly with Swisscom and XGS-PON, but purely with the question of how scarce goods are distributed “fairly and at least somewhat fairly”.

So it could just as easily be about the distribution of scarce gas or medicines 🙂

From a market economic point of view, the allocation mechanism for scarce goods is the price, i.e. in principle an auction, but if all consumers pay the same regulated price for the same product anyway, you have to commit to some rules at all times when demand exceeds supply , who should now receive the demanded good and in what quantity.

My original question is aimed at these specific allocation rules and the answer “there is always enough of everything anyway” is really not helpful, as this statement has often proven to be completely wrong.

5018 · Jun 28, 2023

I think this is a very theoretical discussion. There were also bottlenecks with VDSL. The utilization was simply monitored and when a threshold was reached, further expansion was carried out. That happened very quickly. You learn from it. If it really is. If bottlenecks were to occur, then there would definitely be countermeasures.

All the documents that the Federal Council has published are much more interesting today. I read it for a long time and sometimes I couldn’t stop being amazed. 🙂

Werner · Jun 28, 2023

@5018

OT: I never had any doubts that we understood each other well and this will certainly continue to be the case.

I may have delved too deeply into the technical background of I still think it’s a legitimate question.

P.S. By the way, I started thinking about it myself because I met someone who has an automated process on his PC to create a statistical graph of his Internet L connection around the clock, which generates his XGS-PON -Anschluss every 15 minutes for the runtime of 30 seconds each for maximum upload and download speeds, and assumes that he can’t affect his neighbors at all, since he only uses his own Anschluss and yes In any case, only the bandwidth is received that no one else needs and that would otherwise just “lay around” unused…

5018 · Jun 28, 2023

Yes, there may be 1-2-3 customers. But I bet the others on the tree never noticed. Just think about the speed you need when surfing.

And @Werner: but sure.

I was really impressed by the Federal Council’s ideas about what speeds customers will need in the next 1-5 years. They must have smoked something. But good reasoning to justify the subsidies.

Werner · Jun 28, 2023

@5018 wrote:

…..

I was really impressed by the Federal Council’s ideas about what speeds customers will need in the next 1-5 years. They must have smoked something. But good reasoning to justify the subsidies.

Governor, c’est prévoir

When I started my own entry into global computer communication with a 9,600 baud modem and before the introduction of the World Wide Web, I couldn’t imagine where we would still be with ours after just over 30 years The ever-increasing communication needs have already arrived 🙂

Roger G · Jun 28, 2023

@GrandDixence wrote:

In the

https://www.itu.int/rec/T-REC-G.9807.1

That’s exactly the basic document, thanks for the link. Anyone who wants to delve deeper into the subject will also find the various options and requirements for the ONT. I don’t do it, we have the right guys who have the knowledge.

Roger G · Jun 28, 2023

@Werner wrote:

P.S. By the way, I started thinking about it myself because I met someone who has an automated process on his PC to create a statistical graph of his Internet L connection around the clock, which generates his XGS-PON -Anschluss every 15 minutes for the runtime of 30 seconds each for maximum upload and download speeds, and assumes that he can’t affect his neighbors at all, since he only uses his own Anschluss and yes In any case, only the bandwidth is received that no one else needs and that would otherwise just “lay around” unused…

@Werner You are welcome to tell him that we have already recognized him and have him on the watchlist 😁 Also send him the link to the general terms and conditions and special provisions (use for normal private purposes… special applications only by arrangement with Swisscom, etc.).

user109 · Jun 28, 2023

@GrandDixence wrote:

In the

https://www.itu.int/rec/T-REC-G.9807.1

I haven’t read through this material and tried to understand it because the topic doesn’t interest me enough…

@GrandDixence @Roger G I would like to read the document but I don’t have access to it.

Werner · Jun 28, 2023

@Roger G

Of course I’m happy to do that, but since Swisscom itself makes the 10 gig speed test server available to everyone without any restrictions on use, it might not be easy to convey 🙂

Of course, it would be easier if I could also credibly tell him at the same time that, contrary to what he has previously thought, his speed test statistics also eat away at the currently running bandwidth of each of his PON neighbors as unexpected collateral damage. 🙂

Roger G · Jun 29, 2023

@user109 wrote:

@GrandDixence wrote:

In the

https://www.itu.int/rec/T-REC-G.9807.1

I didn’t read through this reading material and try to understand it because the topic doesn’t interest me enough…

@GrandDixence @Roger G I would like to read the document but I don’t have access to it.

Umpf, yes sorry, I didn’t realize that you have to be an ITU member. So I can’t just share it publicly. Ev. You can find it online, but you probably have to register there too. An example:

[ REC-G-9807-1-201606-I-PDF-E#)