Pacific Fibre ceases operations

Forums › New Zealand Broadband › Pacific Fibre ceases operations

1 | 2 | 3 | 4 | 5

mercutio

1392 posts

Uber Geek

#666994 3-Aug-2012 14:11

ubergeeknz:
I will point out why this testing is flawed:

1. It looks like you are tracerouting from US back to NZ. Routing is not always the same both ways, so this might not mean much for a connection initiated from NZ.

That's probably why ihug had lower latency. Connects initiated in NZ could take a different path, but it's more common to have assymetric traffic, with one path going via AU and one via Hawaii then to depend on which end initiates.

2. Who's to say that the www is on their network core or that it has any bearing on residential customer traffic routes

True - but I don't have the IP's of any residential connections.

A better to pick a server in the US - or several - and trace routes from residential connections on each ISP with good local performance.

Let's also be mindful that not all ISPs have interconnects everywhere, so the best path to two servers in LA might be very different.

That'd mean having residential connections on multiple ISP's, which I don't have.

Although ISP's don't have interconnects everywhere this is coresite/any2ix, which is one of the major interconnects between NZ and the world. Although one little issue is that any2ix is located in both southern and northern california, and vocus are only peering in northern california. Which is why the 2nd hop shows pings of just under 9 msec. Connectivity to facebook for instance may be a little better, as it is also in northern california.

But basically the concern of higher pings due to going via Australia is already happening to some extent. And could happen more.

Globalgateway can have high pings sometimes too. But don't show an australian hop.

traceroute to www.trademe.co.nz (202.162.73.2), 64 hops max, 40 byte packets
1 174.136.111.233 (174.136.111.233) 0.930 ms 0.887 ms 0.962 ms
2 ge0-15.as01.lax07.mzima.net (67.199.135.101) 0.550 ms 0.527 ms 0.522 ms
3 69.174.120.182 (69.174.120.182) 17.210 ms 17.496 ms 17.805 ms
4 f0-0.pabr1.netgate.net.nz (198.32.176.97) 9.116 ms 9.31 ms 9.26 ms
5 ae0.sjbr2.global-gateway.net.nz (203.96.120.73) 9.603 ms 9.434 ms 9.458 ms
6 so5-0-2.akbr4.global-gateway.net.nz (202.50.232.33) 166.1 ms 162.230 ms 166.113 ms
7 ae5-2.akbr5.global-gateway.net.nz (202.50.232.78) 166.488 ms 166.486 ms 166.492 ms
8 trademe-int.akbr5.global-gateway.net.nz (202.50.233.114) 167.839 ms 167.727 ms 167.404 ms
9 www.trademe.co.nz (202.162.73.2) 166.815 ms 158.339 ms 158.434 ms

Mzima is a pretty major tier 2 transit provider. And has a suboptimal path too.

GoodSync

GoodSync. Easily back up and sync your files with GoodSync. Simple and secure file backup and synchronisation software will ensure that your files are never lost (affiliate link).

Ragnor

8192 posts

Uber Geek

Trusted

#666996 3-Aug-2012 14:16

stevenz: Cable seems like a fairly sure way to make money, a public offering might've attracted interested, heck, even I might've been keen to put a bit in.

Regardless, I'm still pretty happy with $70-ish getting me 120GB/month @ 2MB/sec via Telecom. UFB is of no interest to me personally but I'm obviously not the target audience.

If it was a sure bet it would have attracted enough funding, it didn't.

ubergeeknz

3344 posts

Uber Geek

Trusted

Vocus

#667001 3-Aug-2012 14:20

Just making sure people realise that it's not really a comparison of different ISP routing to the US -- re-reading I was a little err blunt :)

Like most things these days, the main limiting factor is not technology but commercials - who's buddy with who, how much does a direct to US link with low latency cost vs AU->US, etc etc.

Anyway, for sure some competition in international transit would go a long way to bringing costs down and should overall improve performance for users in NZ accessing overseas content.

Another good way is to have more CDN's here to deliver the bulk of content.

But you will never get latency super low to the US due to physical constraints, and the difference between going via AU and not is fairly negligible in the scheme of things.

The best solution for this is protocols that handle latency well (infrequent handshaking), and in the case of websites, reducing the number of assets needing to be requested (hence reducing the number of browser requests). Unfortunately for us, content providers in the USA don't tend to worry too much about the user experience for overseas users.

mercutio

1392 posts

Uber Geek

#667003 3-Aug-2012 14:22

ubergeeknz: Just making sure people realise that it's not really a comparison of different ISP routing to the US -- re-reading I was a little err blunt :)

Like most things these days, the main limiting factor is not technology but commercials - who's buddy with who, how much does a direct to US link with low latency cost vs AU->US, etc etc.

Anyway, for sure some competition in international transit would go a long way to bringing costs down and should overall improve performance for users in NZ accessing overseas content.

Another good way is to have more CDN's here to deliver the bulk of content.

But you will never get latency super low to the US due to physical constraints, and the difference between going via AU and not is fairly negligible in the scheme of things.

The best solution for this is protocols that handle latency well (infrequent handshaking), and in the case of websites, reducing the number of assets needing to be requested (hence reducing the number of browser requests). Unfortunately for us, content providers in the USA don't tend to worry too much about the user experience for overseas users.

I wouldn't say that 60 msec is neglible.

One way through Australia isn't nearly as bad as two way through Australia. And I think at the same time it's important to have "direct" connections in the US to many locations, and to minimise latency at that end.

Part of the issues seem to happen to do with southern/northern california. There are two entry points for the southern cross cable. One of them is between Los Angeles and San Jose. But it doesn't seem common to have direct links to both Los Angeles and San Jose. The other one is way up by Seattle. And I think that leg is sometimes used increasing latency even without going via Australia.

There's a little map at: http://www.southerncrosscables.com/public/Backhaul/default.cfm

hellonearthisman

1819 posts

Uber Geek

Trusted

#667231 3-Aug-2012 20:51

Maybe Pacific Fibre should rebrand to to Pacific Tachyon and build a tachyonic detector and generator in NZ.
Even though the particals don't seem to travel faster than light, they can travel in a straight line, cutting through the Earth core and reducing the time it takes vs going the long way around the surface.
Saving 11oooKm vs 9oooKm NZ to the UK at 19oooKm could be cut to 12oooKm

I guess that will be happening as fast as new cable to Australia happens.

There is a big demand to connect to the USA, but as the big content people expand they will/have build servers in Australia and so that would shift the demand some what.

A publish share float would have been good, or there was talk at one time that P_C would use some kind of micro financing. 250ooo people chipping in $400 would have been a good chunk of the money.

DonGould

3892 posts

Uber Geek

#667236 3-Aug-2012 21:10

hellonearthisman: There is a big demand to connect to the USA,

Where?

Most of the reading I've done this week suggests there is currently an over supply.

Promote New Zealand - Get yourself a .kiwi.nz domain name!!!

Check out mine - i.am.a.can.do.kiwi.nz - don@i.am.a.can.do.kiwi.nz

daverobb

7 posts

Wannabe Geek

Trusted

TelstraClear

#667344 4-Aug-2012 07:49

mercutio:

Well yeah - but I think there is actually a lot of hidden latency on the internet now days.

What is this hidden latency of which you speak? You can't hide latency, it's obvious when it's there, because packets take longer to get somewhere.

mercutio:

Some people say that ping/traceroute is inaccurate cos icmp is depriortised. But sometimes whole tcp streams can be deprioritised (or go across higher latency path), rate-shaped(or have enough packet loss to limit transfer speeds below that of which the connection accessing). You can do a traceroute, then do an actual connection and end up on a different path.

So you may ping and get 125 msec ping, then do a tcp connection and get 135 msec.

That doesn't happen very often at the protocol level. Yes, there are occasionally load balancing things going on which will use a couple of different paths, usually some sort of flow-based hashing. But it's not "TCP one way, everything else the other", because TCP is 80%+ of the traffic, so it doesn't make much difference if you put all the TCP traffic down one path!

mercutio:

But this kind of traffic shifting could become even more common with "bulk" traffic being pushed through Australia and interactive traffic being pushed straight. Or low-value customers being pushed through Australia.

Who are these low-value customers you speak of? It's not business, because they generally pay more. It's not residential, because they complain a lot on geek zone

, and it's certainly not government/enterprise.

I think anyone who's been in the industry long enough to remember the satellite links we used to use to shift data into NZ also remembers efforts to push some traffic types one way and some the other. It's a pain in the butt, because you have to control both ends of the link; you can't just do it at the BGP level, so you need routing infrastructure in the US, Aus, or wherever, to decide what goes down what path. That's expensive, and annoying to maintain.

--David

Yes, I work for TelstraClear.

daverobb

7 posts

Wannabe Geek

Trusted

TelstraClear

#667348 4-Aug-2012 08:04

DonGould:

Ok first up you seem to be making the assumption that I think traffic draw is going to continue to come from .us and push to same. I don't.

No, I'm not assuming that at all. I said that if you used PF to get to Aus, and then some other capacity to get to the US/anything-else-which-goes-via-the-US, performance is going to be worse than going direct (via SCCN, or some other cable if someone builds another NZ-US link).

I also said, and this is relevant to NZ-Asia traffic (which is almost certainly going via AU):

daverobb: Having a cable dedicated to NZ-AU traffic isn't going to be any different to the existing SCCN capacity; I think it very unlikely that it would be cheaper

To which you replied:

DonGould:

Yes it will be different. To get to PPC1, AJC or Endv, you're not going to have to touch SCCN at all.

So you're saying it (another cable NZ-AU) would be different because it's not SCCN. Sorry, I thought we were talking about performance and pricing differences, not branding.

Whomever you buy capacity from, be it SCCN, PF, or DaveRobbInternationalFibre, unless they own capacity all the way from NZ-Asia, you're going to hop off in Sydney, and then onto another system.

DonGould:

The problem at present is that when you buy capacity from SCCN you then have to get it off SCCN, to a peering point then onto the other cables, which the SCCN sales guy knows and will cut a deal just to use the capacity with them direct to .us. This stuff has to be an obvious competitive issue.

I don't see how it's a problem that you're able to get SCCN to cut a deal, that sounds like a good thing to me! And to do this (apparently), we don't even need another fibre out of NZ, we just need them to exist from AU-US.

DonGould:

Ok, isn't this where traffic management technology has to come in? Also, how are people going to know that the traffic is heading to .au --> .us if it's layer 2?

It's not layer-2 all the way from your PC to the far end. And even if it was, you can still exchange frames with the other end, which allows you to determine the path length. There's no magic way to hide increased RTT, there Just Isn't.

DonGould:
Is anyone going to notice is you push torrent traffic via .au?

Yes. And it's annoying (read expensive) to do so.

--David

Yes, I work for TelstraClear.

webwat

2036 posts

Uber Geek

Trusted

#669118 7-Aug-2012 16:18

hellonearthisman: I was hoping they would just connect to OZ, stage one. Then look at the figures of the functioning cable and then look at a 2nd cable to the states.

Just another OZ connection would be great it would keep the industrial players real.

I wonder if Kordia can buy the bits of Pacific Fiber to complete that ozzy cable job.

Would be great if Kordia bought the remains of Pacific Fibre, but I wonder if duplicating the main SCC destinations, and resulting perceptions of directly competing with it, was the reason for investment drying up.

I also think the data hosting industry needs to be a key driver of international cables. Networks of any kind are driven by whats happening at their key destinations, not so much the preferences of end users. At the moment data centres in NZ are growing as fast as the AU market, and many NZ services are still hosted in the US. If you want to start a big cloud service or something, you probably want it to be really well interconnected rather than just having redundancy to the same IXs in the same places.

I guess you can't avoid an Auckland landing if you want to connect NZ, but Asian markets are growing while US not su much. Pacific Fibre might have focussed more on new routes to places SCC doesnt go by perhaps doing legs to Auckland-Melbourne and Auckland-Guam. To entrench their advantage, a big data centre build might create opportunities to get early customers launching trans-tasman hosting (such as the odd cloud accounting service).

Time to find a new industry!

1 | 2 | 3 | 4 | 5