NBR: Telecom exec: Pacific Fibre needs a geography lesson
The NBR published another critical headline today. Chris Keall was the writer again.
Telecom exec: Pacific Fibre needs a geography lesson
Chris spoke with Telecom Wholesale chief Matt Crocket:
His message, in short, is that Southern Cross has a load of unlit fibre, and has been reducing prices by around 20% per annum for the past few years.
While it is good that Southern Cross prices have dropped over time, we would contend that it is mostly in response to competitive threats – like that of Kordia’s proposed cable to Australia and that there is still plenty of scope left to reduce prices. Some evidence of those high prices later in the article:
So why do things slow down during the evening rush-hour, as Pacific Fibre co-founder Mark Rushworth highlighted yesterday?“Because so much of traffic goes from the US here at peak hours and a lot of people trying to use it at once. People [ISPs] could buy more capacity on Southern Cross to address that – and they have been – but to make it completely uncontended, quite frankly, is not realistic.”
That pretty much says that the ISPs are unwilling to buy more capacity because the price is too high. And that’s exactly the problem we are trying to solve.
We also see the demand for capacity rising sharply as the Australian and New Zealand fibre initiatives move forward and want to make sure that they are able to use a competitive market to source international bandwidth.
Now we turn to the headline topic:
Earlier, in an email, Mr Crockett said of Pacific Fibre’s claim it’s Auckland-LA route would have lower latency, “If they think there is a more direct, lower latency route then they must know something about the seabed and broader geography of the Pacific that we’ve missed!”
There is – the direct distance from New Zealand to the USA is substantially shorter than the distance from New Zealand to Hawaii to the USA – the route that the Southern Cross cable takes. (The amusing Southern Cross map may not make it that obvious though)
Southern Cross’s shortest route from New Zealand is 8002 km to Hawaii and a further 4125 km to the USA mainland for a total (exclusing land-based cables) of 12127km. In contrast the direct route to LA or San Francisco from Auckland is about 10500 km. That’s a lot shorter, and also there is no lag associated with a landing station in Hawaii.
We hear again that this is not easy:
Mr Crockett said he could write a book about the unexpected logistical problems companies encounter in the world of sub-sea cabling – and another about the resource consent battles Southern Cross has been through, especially with its US landing.
Indeed the submarine cabling industry has these horror stories, and Southern Cross has a particular shocker – they were let down by a partner that failed to get the appropriate permits to land in the USA. We are aware of these stories and are working with and will always work with experienced campaigners that have been through a few of them. Permitting is very early on our project plan.
The critical nature of the NBR article aside, I think Matt Crockett’s point was although the point-to-point length is clearly shorter, can the cable actually be laid along this path or are there obstacles (trenches, under-sea mountain ranges etc.) that make it a longer or more expensive pathway. Like a lot of other industry observers, I sincerely hope not.
I concur with Gerards point. All bashing aside, its unrealistic to compare a theoretical straight line with an actual laid route. Further, even if you could achieve the theoretical straight line, based on the SX stated transit time of about 65msec, a 10,000km route versus a 12,000km one would only reduce it to about 55msec. I don’t think the saving of 1/10th of a second is the delay that users are concerned with.
While there would be some gamers who would salivate at even a 10ms gain.
In any case, the landing in Hawaii probably add’s a resonable amount of of latency. Also while I haven’t read the article, how accurate is Southern Cross’s stated 65 msec?
I’ve never seen anything close to that in traceroutes. Checking again now I’m seeing about 115msec between the last visible local hop and the next hop (first visible overseas/LA hop). I’m pretty sure this is consistent with what I’ve seen before. This is early Easter Monday morning so I doubt there’s much saturation of any link but I’ll concede a high priority customer may be able to get better latency if they ask for it from Southern Cross. So let’s say a 20ms reduction. That’s still 95msec which is a lot more then their stated 65msec.
This isn’t my area of expertise but I believe one of the big factors is the many hops inbetween (amplifiers etc). Note that a 12000km route only takes about 40ms at the speed of light. The actual distance travelled will be futher then that for the reasons already mentioned but I don’t think this accounts for the difference.
In other words I suspect Southern Cross’s stated latency is more theoretical then in practice, Leaving plenty of room where Pacific Fibre can improve particularly being newer and every little bit helps and is important (although I somewhat agree it’s perhaps a little misleading to concentrate solely on the distance).
Intended to say “Well there would be some…..”
also I intended to say amplifiers, silent routers etc.
Anyway one more point I wanted to make that a cable going a relatively different route would be beneficial. As someone with connections to Malaysia, I have some personal knowledge of how bad it was when the many of the East and South East Asian international links were cut. Problem being while there are quite a few, they nearly all seem to go thorough the same earthquake prone area of Taiwan so it’s rather easy for one hit to take out a lot. I don’t know much about the actual route planned but if it’s fairly different from the SCC route having this sort of redundancy would be good. We already do have some by the connection to Australia but additional can’t hurt. Of course they do all land in volcanic Auckland but as I live in Auckland myself if something happens why do I care if the rest of the country is cut off from the world?
@nil I would watch those assumptions about latency. Gaming aside (no small industry in itself), I can put forward a scenario I’ve encountered professionally, frequently:
Multi-tiered RIAs (applications in the browser) present a degraded User experience when bandwidth and latency are low. There are a lot of simple tricks for dealing with bandwidth issues, from minimising the required client-side information to streaming in the background. However latency is much harder to ‘trick’ – the application responsiveness can only be as fast as the round trip, or the client has to ‘pretend to know’ the server response ahead of time and then correct itself when it finally gets the canonical data.
The impact of high-latency connections thus requires additional content-creator efforts such as:
- Duplicate code in client and server (this is considerable extra effort to maintain)
- Hosting at mutliple geographic locations to present local users a better experience
Not only that, content-creators on the “Internet Mainland” of Americas, Europe and Asia have such a lions share of the global population that us, adrift in Oceania, are less likely to be compensated for in the design of online software. If our link times drop, then our user experience with much of the software out there would start to come into line with the content-creators original intent.
Sorry I meant that as @Samuel