Enable gets UFB speeds right

Forums › New Zealand Broadband › Enable gets UFB speeds right

JohnButt

374 posts

Ultimate Geek
+1 received by user: 47

Trusted

#138129 20-Dec-2013 19:14

Enable gets their UFB speeds better or equal to advertised speed in November.

https://www.truenet.co.nz/articles/enable-beats-chorus-uff-100mbs

sbiddle

30853 posts

Uber Geek
+1 received by user: 9996

Retired Mod

Trusted

Biddle Corp

Lifetime subscriber

#955637 20-Dec-2013 19:47

My interpretation of your data differs significantly to yours.

With the exception of the 100/50 Chorus node you have dragging down the test (which is probably an ISP handover issue), every result there meets the 100% Advertised Speed as defined by the CFH specs which is determined at Layer 2, not Layer 3.

You 100% line is in the wrong place and comparing Enable (who have increased their L2 overhead already) with Chorus (who are as part of right performing) is very unfair.

Maybe you could ask Enable (and other LFCs) about their SLAs on the speeds, something Chorus will be introducing as part of Right Performing. I've yet to see any announcements from the LFCs in regards to introducing SLAs on their EIR which is a best effort.

Haver you done any testing yet on whether providers are meeting their CIR?

JohnButt

374 posts

Ultimate Geek
+1 received by user: 47

Trusted

#955740 21-Dec-2013 10:33

Thanks Steve for your comment,

I agree with your interpretation fully, but do not agree with your conclusion which is effectively that customers should only expect the leftovers from a manufacturers product output.

The CFH specs are for the "manufacturer" i.e. the LFC's targets. I agree fully that they meet these targets, however for consumers a manufacturers output has no relevance. Imagine buying petrol by the litre and only getting 9 litres for every 10 you purchase - because a litre is lost in the delivery from the oil-well, you would not be happy.

The purchase price can, and indeed often will, include the cost of waste for a manufactured product, and I am sure the price for UFB is the same, but product is always sold meeting the retailers size measurements, not the wholesalers.

This is extremely fair, and LFCs agree! Enable actually launched their 106Mb/s product before we published our previous comment, our probes were already measuring 100Mb/s in their region a week before we published, but within December so those measurements missed our cutoff.

Yes we measure EIR and will measure CIR in future.

RunningMan

9184 posts

Uber Geek
+1 received by user: 4834

#955772 21-Dec-2013 11:55

JohnButt: [snip]
The CFH specs are for the "manufacturer" i.e. the LFC's targets. I agree fully that they meet these targets, however for consumers a manufacturers output has no relevance.

IMO the graph above, combined with the thread title imply (to the average punter that doesn't understand OSI layers) that with Chorus you don't get what you pay for, yet the statement above confirms you do.

If the layer 2 throughput meets the advertised throughput then that should be level with the 100% line on the graph.

I agree that an advertised layer 2 throughput is confusing for an end user to understand, especially when they only see about 90% of that rate when transferring their data, however, I would suggest that that is an entirely seperate issue to whether or not the connection meets the advertised throughput.

sbiddle

30853 posts

Uber Geek
+1 received by user: 9996

Retired Mod

Trusted

Biddle Corp

Lifetime subscriber

#955814 21-Dec-2013 14:08

JohnButt:
The CFH specs are for the "manufacturer" i.e. the LFC's targets. I agree fully that they meet these targets, however for consumers a manufacturers output has no relevance. Imagine buying petrol by the litre and only getting 9 litres for every 10 you purchase - because a litre is lost in the delivery from the oil-well, you would not be happy.

Based on your logic your xDSL charts should contain the xDSL speeds of the sync rate of the modem with a 100% next it, yet they don't. You only measure the real world layer 3 performance of the xDSL product.

While I agree with the confusion between layer 2 and layer 3 your first lot of UFB results did nothing but harm to the industry. Rather than explaining why the results occurred, the media simply trashed Chorus based on your findings. UFB headline speeds were created as layer 2 offering.

As we move beyond 100Mbps your testing is going to become pretty well meaningless anyway at least from a layer 1 and layer 2 level, because right now (with the exception of the odd teething issue that has occurred) there are no bottlenecks anywhere in the UFB networks because these networks are being engineered to handle capacity far in excess of what they have right now. Chorus in particular have changed their network architecture quite fundamentally from the year 1 and year 2 builds, and this architecture will make it easier to smaller customer splits on GPON if required to deliver faster speeds.

JohnButt

374 posts

Ultimate Geek
+1 received by user: 47

Trusted

#955899 21-Dec-2013 16:30

Yes we are measuring the layer 3 product and reporting on it, nothing else matters for end consumers. In practice, Layer 3 is the product which is advertised, not layer 2.

I struggle with your accusation that my earlier report "did nothing but harm to the industry". A mirror to bad practices cannot be harmful. If harm was done, then the industry needs to look at itself, we only report the results of testing at the consumer. In this case we have grouped consumers into those where the wholesaler is common.

The advertised speeds are 100Mb/s and 30Mb/s, convince me that there is an ISP that advertises otherwise.

Nowhere is a consumer product advertised with a speed of 94Mb/s or 28Mb/s, which is what you get on layer 3 with a layer 2 speed of 100Mb/s or 30Mb/s respectively.

Just for simple comparison, Snap is the first to have significant users and their website explains the 100Mb/s product as "Offering speeds of up to 100Mbps download and up to 50Mbps upload, this is the fibre package that is ideal for those people who like to upload a lot of files and videos to the cloud and consume their TV content online." No mention of layer 2

As for meaningless measurement Steve, you need to read Orcon's FAQ where they explain the impact of the internet on fibre speeds quite reasonably. You could learn something.

sbiddle

30853 posts

Uber Geek
+1 received by user: 9996

Retired Mod

Trusted

Biddle Corp

Lifetime subscriber

#955913 21-Dec-2013 17:05

JohnButt: Yes we are measuring the layer 3 product and reporting on it, nothing else matters for end consumers. In practice, Layer 3 is the product which is advertised, not layer 2.

Nobody mentions layer 2 or layer 3 anywhere in reference to speed, it's nothing but an assumption. Yes it may be a logical assumption, but it's still an assumption. It's not Chorus or any of the LFCs fault that CFH set dimensioning at a layer 2 speed, but it's also not your average home user's fault that they don't understand the OSI stack. This is something that's changing, but I still think it's wrong of you to infer there are problems with UFB, Chorus or LFCs when there isn't.

There does still seem to be some ingress policing issues for some ISPs due to the differences between Chorus using Metro Ethernet Forum standards and other providers using TR156 / TR101

The new Chorus SLAs that will be offered once contention is introduced across the whole Chorus network are vastly superior to the current frame loss figures that everybody is following which can have negative impacts on TCP performance. Knowing that Chorus will guarantee 77Mbps or greater 95% of the time is far more meaningful. I've yet to hear whether the LFCs will copy this directly, but it will be good if they do performance ends up being equal across every provider.

Geekzone support

Support Geekzone with one-off or recurring donations Donate via PressPatron.

Foiler

220 posts

Master Geek
+1 received by user: 10

#955930 21-Dec-2013 18:06

I wonder what the Jitter and Packet Loss stats are like? A tiny amount of either of those might be more to worry about than ONLY getting 95% of the full rate.

sbiddle

30853 posts

Uber Geek
+1 received by user: 9996

Retired Mod

Trusted

Biddle Corp

Lifetime subscriber

#955940 21-Dec-2013 18:42

Foiler: I wonder what the Jitter and Packet Loss stats are like? A tiny amount of either of those might be more to worry about than ONLY getting 95% of the full rate.

Frame delay variation and loss form part of the CFH specs.

buildinggeekrotation

13 posts

Geek
+1 received by user: 5

#956416 23-Dec-2013 10:02

JohnButt: The advertised speeds are 100Mb/s and 30Mb/s, convince me that there is an ISP that advertises otherwise.

Nowhere is a consumer product advertised with a speed of 94Mb/s or 28Mb/s, which is what you get on layer 3 with a layer 2 speed of 100Mb/s or 30Mb/s respectively.

Seriously? Your article is misleading and reads like there is an issue with the LFCs/Chorus' layer 2 products. Steve has explained well why this is not true. And this comment that you try to back up your point with is so far off the mark. Do you understand that the retailers are advertising these layer 3 speeds, not the LFCs/Chorus? The LFCs/Chorus/CFH call a 100/50 a 100/50 layer 2 - and deliver on that.

The RSPs should be advertising rates THEY can deliver at a layer 3 level, as is the case around the world. LFC/Chorus shouldn't be blamed if RSPs are advertising rates they can't deliver on.

Let's look at a simple example of a wood mill. They supply dry timber to a wood retailer (Mitre 10 for example). Mitre10 stores the wood outside for whatever reason and it gets wet. They sell it as dry wood and the customer complains because the wood is wet....who is to blame, whose advertising is misleading? The wood suppliers?....surely not, they supplied the specified dry wood. It was the retailer that left it out in the rain.....

Sorry for the bad example everyone, but you get my point.

plambrechtsen

1948 posts

Uber Geek
+1 received by user: 459
Inactive user

#956447 23-Dec-2013 10:43

I think the other main problem in the article is it will be used as a "sound byte" for those less technical and may be rather influential in the whole argument.

If your graph showed the 100% line at 94mb as that is what you are capable of testing using l3 technologies which is what you are testing and had Enable at 110% or whatever. Then with a title such aa "Enable deliver service above what they are contractually obliged to provide" then I don't think there would be so much argument about it.

Otherwise I suggest you get a handover and some ONTs and test and report back on the l2 services the LFCs provide.

There is too much variance between RSPs such as nationwide back haul and international capacity that those two items are now the determining factor rather than the first l2 hop.

To do anything else is dishonest and will be used as the ongoing beat up of Chorus who are doing nothing other than providing service they have been contractually obliged to provide with CFH.

This is my own personal view on the matter and not that of my employer.

JohnButt

374 posts

Ultimate Geek
+1 received by user: 47

Trusted

#956652 23-Dec-2013 18:24

Let me clarify in simple terms

Enable promotes 100Mb/s for residential users - here;
http://www.enable.net.nz/why-choose-fibre/home.html

UltrafastFibre promotes 100Mb/s for residential users - here;
http://www.ultrafastfibre.co.nz/fibre-and-its-benefits/how-fast-is-fibre

Chorus/CFH calls the 30 and 100Mb/s new services "bandwidth profiles" in their Bitstream 2 Service http://www.chorus.co.nz/file/24162/chorus-ufb-services-agreement-service-description-for-bitstream-2.pdf

I cannot find at any stage an LFC explaining that the speed you will actually get is subject to the limitations of Level 3 over Level 2.

As stated earlier, Enable have replaced their offering with a higher Level 2 speed to achieve the advertised speeds at Level 3. TrueNet have measured that change and reported success.

Geekzone support

Support Geekzone with one-off or recurring donations Donate via PressPatron.

plambrechtsen

1948 posts

Uber Geek
+1 received by user: 459
Inactive user

#956690 23-Dec-2013 20:07

JohnButt: Let me clarify in simple terms

Enable promotes 100Mb/s for residential users - here;
http://www.enable.net.nz/why-choose-fibre/home.html

UltrafastFibre promotes 100Mb/s for residential users - here;
http://www.ultrafastfibre.co.nz/fibre-and-its-benefits/how-fast-is-fibre

Chorus/CFH calls the 30 and 100Mb/s new services "bandwidth profiles" in their Bitstream 2 Service http://www.chorus.co.nz/file/24162/chorus-ufb-services-agreement-service-description-for-bitstream-2.pdf

I cannot find at any stage an LFC explaining that the speed you will actually get is subject to the limitations of ~~Level~~Layer 3 over ~~Level~~Layer 2.

As stated earlier, Enable have replaced their offering with a higher ~~Level~~Layer 2 speed to achieve the advertised speeds at ~~Level~~Layer 3. TrueNet have measured that change and reported success.

Call me pedantic but it's Layer not Level as in OSI Layer 2(Data Link) which the LFC provides and Layer 3(Network) which the RSP Provides followed by 4(Transport), 5(Session), 6(Presentation) and 7(Application).

But again you seem to be missing the point. All LFCs only provide Layer 2 service at the agreed to rate, and they have no interest or need to provide Layer 3 services as that's not their job. Plus they wholesale service to retail service providers who then in turn sell it to end customers it isn't their job to advertise to end customers .

The people you should have issue with are the RSPs who sell Retail service at 100mb or 30mb downstream rates when that is never achievable due to Layer 2/3 and depending on which RSP if they offer IPoE or PPPoE service as with PPPoE they loose another 8 bytes per packet from the PPPoE header thus in turn giving them slightly slower speed than IPoE subscribers.

But then you would also need to take into account the CPE used along with the customers devices connected. As we have recently seen that can go horridly wrong.

Talkiet

4819 posts

Uber Geek
+1 received by user: 3934

Trusted

#956715 23-Dec-2013 21:51

Calm down everyone... Here's my read (personal read, NOT necessarily a Telecom view)

The LFCs are doing what they are contracted to do. Saying or insinuating that Chorus for example is underperforming because your probes show less than 100mbit L3 throughput is just wrong. If you want to have a go at someone for that, then it's the RSPs that should shoulder the responsibility of delivering the service they advertise.

The RSPs are currently in a tough spot because of precisely the above point. No-one wants to advertise a 96mbps or 28mbps product (because all the others are going to advertise 100/30).

FWIW I have done the testing myself mentioned earlier (get a handover, do the throughput testing at L2) and the services worked to spec.

There's an awful lot of stuff going on at the moment to make all of this clearer to consumers, and the last thing we need is inaccurate, imprecise interpretation of test results that do nothing to properly inform the decisions that need to be made.

John, if you're going to "test the LFCs" then you should be testing at L2 using appropriate technology. If you are going to test the end to end Retail Services, then by all means use your L3 tests and compare the results to what the RETAIL SERVICE PROVIDERS claim.

Cheers - N

ps. Just to absolutely confirm, this is my personal opinion, I'm on holiday.

Please note all comments are from my own brain and don't necessarily represent the position or opinions of my employer, previous employers, colleagues, friends or pets.

sbiddle

30853 posts

Uber Geek
+1 received by user: 9996

Retired Mod

Trusted

Biddle Corp

Lifetime subscriber

#956718 23-Dec-2013 22:04

The biggest problem going forward is that Truenet testing is going to be largely irrelevant once we go beyond 100Mbps because it's going to be an attempt to measure something that will sit outside the margin of error.

Right now you're testing both xDSL and UFB services that have absolutely no contention at the layer 2 level. This will change with Right Performing (and we can assume similar plans from the LFC's) and will move towards delivering SLA's to customers once contention occurs on the GPON network.

Chorus will guarantee L2 speeds of 77Mbps on >95% of tests and 109Mbps on >80% of tests once these SLAs are introduced and L2 dimensioning is changed. Clearly real word L3 speeds will be below this and depend whether the RSP is using IPoE or PPPoE.

Your testing in the future needs to make it very clear that SLAs may exist for the product that is being tested.

I'm interested in your view as to whether you see UFB as broken when Gigatown finally happens and with a 16 way split using your current testing methodology you see users with speeds fluctuating from 500Mbps right up to 1Gbps. What do you believe a SLA should be for a 1Gbps user with a 16 way split on GPON?