[AusNOG] "It's like grandfather's axe"

[Mark ZZZ Smith]

----- Original Message -----
> From: Mark Delany <g2x at juliet.emu.st>
> To: ausnog at lists.ausnog.net
> Cc: 
> Sent: Sunday, 24 November 2013 5:19 PM
> Subject: Re: [AusNOG] "It's like grandfather's axe"
> 
> On 23Nov13, Mark ZZZ Smith allegedly wrote:
> 
>>  > And I mean a real, measurable QOS. How hard can it be to have the
>>  > customer-side NTU have a builtin QOS monitor? If the QOS light is red
>>  > the provider has a fault to fix. Simple.
>>  >?
>> 
>>  How do you measure and assure the "QoS" of the Internet, when 
> many, many parties are involved
> 
> Because the topic at hand is the last mile. There's only one party
> involved between a residence and whatever their first termination
> point is. QOS measurements are routinely tractable for this
> application.
> 

I think "QoS" is way to a general term for this. Non-technical users/customers (i.e., the general public) are likely to think it means that their ISP is making assurances about the performance of the whole of the Internet.  

I think even networking people would misunderstand what was being meant. "QoS" is commonly used to refer to IPv4 DSCP/IPv6 Traffic Class implementations, edge-to-edge across a whole network.

Both of these understanding or misunderstandings of the term "QoS" were why I asked about it.

At best, only assuring the quality of the last mile link could be described as something like "link performance monitoring" or "link quality monitoring". Anything else is open to interpretation and therefore possible confusion by customers, and the negative reputation etc. to the ISP that would come from that confusion.

>> in running it, including the end-users themselves when they install and 

> operate their CPE and the devices that they use to access the Internet?
> 
> If you're QOS monitor is looking at the media transport on the last
> mile then the customer devices are irrelevant. E.g., DSL sync errors
> and BER are a reflection solely of the link quality, not what's
> connected either side of them, or the traffic being transported.
> 
> As someone else mentioned, most DSL modems already provide a wealth of
> QOS data,

That was me, and I would never call it "QoS data".

> the suggestion is simply to codify a minimum acceptable
> level below which the NTU turns on an out-of-spec light that removes
> all the wiggle room that providers hide behind today. You know, like
> rain, strong winds, smoke, daylight savings...
> 

The service providers I've worked for haven't tried to hide behind anything. On a many occasions many of their staff have gone out of their way to resolve hard issues, raised in non-official support forums like Whirlpool, quite often in their own time, including me.

So here's a bit of history of broadband in Australia, regarding link speeds.

Telstra was the original and only broadband supplier, and they supplied 3 ADSL1 tiers - 256/64, 512/128, and 1500/512. That was it. Why did they provide no more than those speeds, even though ADSL1 could support up to 8Mbps? Oops - there it is - the "up to" thing. 

I'm guessing Telstra only provided those speeds because they could assure them universally across all local loops in Australia, regardless of where they were located.

It was only when other providers were able to install their own DSLAMs were speeds "up to" what ever the customers line supported became available. They probably choose to allow ADSL/ADSL2/ADSL2+ negotiate the best possible speed the customers line supported to provide a better service/product to the customer. Of course, as they couldn't guarantee those higher speeds, the needed to use the caveat of "up to" in their advertising, otherwise they would have been guilty of false advertising.

The alternative would have been that the providers with their own DSLAMs could have capped speeds to those that Telstra provided, and therefore would have been able assure the speeds they provided.

Given all the complaints about "up to", perhaps they should have done the latter. I certainly have preferred the former, given that I've literally been on the edge of an exchange coverage area, and still got 3Mbps on ADSL2+.

So maybe the CPE instead should have both an "out of spec" light, and a "better than spec" light, so that customers could appreciate that they were getting more than an assured universal minimum when they can.


> 
>>  > After all, if in 2020 we have "upto" 100Mb/s VDSL, but down 
> to 0MB/s
>>  > when it rains with zero recourse, how much better off are we?
>>  >?
>> 
>>  How do you know there will be zero recourse?
> 
> Status quo? A decade of history showing that many people have routinely
> experienced zero recourse? The many credible anecdotes that are posted
> here?
> The high probability that the incumbent presiding over our last

> decade will be a major player in the next decade?
> 
> Frankly I'm a bit surprised you're asking that question. Could you
> elaborate on why you think it will be different this time round? Every
> pronouncement I've heard is loud on "up to" and silent on QOS.
> 

The "up to" also has to cover Internet performance, as that can't be assured either by an ISP. While it is not actually possible to assure Internet performance, an ISP could go a long way towards doing so by ensuring that the customer's link is nearly always the bottleneck link. I don't know at what performance level that is, but I wouldn't be surprised if it is somewhere in the region of 128Kbps. I'm sure that would be unacceptable to perhaps everybody, despite the complaints about "up to".

Currently Telstra don't have any incentive to use ADSL statistics to judge the quality of a line or use them for troubleshooting. Telstra are only obligated by regulation to deliver an adequate voice service. Going any further costs them money, with no financial benefit.

You could argue they should go further in the interests of good customer service. However, as they are a monopoly supplier, they can adopt an approach of "you'll take it and like it". This sort of thing is why I don't like unnatural monopolys. It creates an incentive to provide the minimum acceptable, and nothing more. One of the things competition creates is the incentive to provide better customer service than what competitors do.

Since future NBN of either political flavour is or will be be carrying voice traffic over IP, formerly carried "out of band", the link layer broadband performance parameters will need to become parameters used to troubleshoot them. 

Regards,

Mark.