Hyperoptic Announce Prices for their 100Mb/s-1Gbp/s Services *DELETED*

Post deleted by Sadoldman

"All tests have shown that customers' throughput on the Hyperoptic fibre network will be 95% of their connection speeds."

Hands up any reader with genuine GbE experience who's ever seen 95% of advertised speed in any real circumstances.

Hands up anyone who thinks this 95% claim is at all credible, even before you start thinking about other bottlenecks in the system.

95% is a stupid claim to make. Will only lead to customer support issues with customers not willing to pay the bill coz they are getting only getting 94% from speedtest.net!

In reply to a post by Anonymous:

"All tests have shown that customers' throughput on the Hyperoptic fibre network will be 95% of their connection speeds."

Hands up any reader with genuine GbE experience who's ever seen 95% of advertised speed in any real circumstances.

Hands up anyone who thinks this 95% claim is at all credible, even before you start thinking about other bottlenecks in the system.

If you re-read the key passage (see below), I think it is actually talking about the maximal throughput any given line can possibly attain in the best case (e.g. 95% of 1Gbps, 5% disappears with protocol overheads and other related stuff). I _don't_ think they are referring to the ability for all customers connected to sustain 95% throughput at all times (implying they have a gargantuan backhaul).

However, the sentence seems (deliberately?) badly worded, and gives the impression that they are.

All tests have shown that customers' throughput on the Hyperoptic fibre network will be 95% of their connection speeds.

I think they are saying something akin to 'You buy a 20Mbps ADSL connection, you might only get 20% of that due to line length, attenuation, etc, but with hyperoptic you get 95% of xxMbps)'. Which happens to be no miracle given that they're doing FTTH and they wouldn't have the same distance related headaches.

Anyway, seems fishy, I'd like to hear them clarify it.

Theses are the same guys who founded BE?

I honestly cant see the point of 1Gig connections right now. Maybe in 10 years sure. I'm fairly happy with my unlimited 6Mb. If I was getting 40Mb or more I could see me being happy for many years as I can't imagine what I would need that bandwidth for at this moment in time.

One thing that I do like the idea of with a pure Fibre network is ping times. They should drop to sub 10ms for anywhere in the UK. And it might even make playing people in the US and other far away countries the realm of possible for the first time especially if they are on a FTTH network too.

Edited by LeJimster (Thu 15-Sep-11 02:49:44)

Too much like spam ...the inclusion of the link did it for me.

Already down to 9ms on FTTC. You're not going to see much more of a reduction on a residential network. Ping times to New York are 90ms. Perfectly playable.

In reply to a post by LeJimster:

I honestly cant see the point of 1Gig connections right now. Maybe in 10 years sure. I'm fairly happy with my unlimited 6Mb. If I was getting 40Mb or more I could see me being happy for many years as I can't imagine what I would need that bandwidth for at this moment in time.

One thing that I do like the idea of with a pure Fibre network is ping times. They should drop to sub 10ms for anywhere in the UK. And it might even make playing people in the US and other far away countries the realm of possible for the first time especially if they are on a FTTH network too.

For really twitchy games where people insist on very low latencies (<=50ms) it'll never be possible, given that light travels slower through fibre than a vacuum, and the fastest dedicated link from UK (slough) to NY (which is dedicated for expensive realtime financial trading, and therefore massively lower latency than any standard link) is 65.7ms.

Fun PDF:
http://www.hiberniagfn.com/documents/ExpectedLatency...

I ping New York at about 75ms from home I think. Be connection with fastpath in London and a 5ms first hop. Fibre might help me drop 4ms off that, though from the traceroutes I've seen from people on the BT trial, they still have a 4ms first hop.

In reply to a post by driz:

I ping New York at about 75ms from home I think. Be connection with fastpath in London and a 5ms first hop. Fibre might help me drop 4ms off that, though from the traceroutes I've seen from people on the BT trial, they still have a 4ms first hop.

I'd be very surprised (but pleased to hear) if it was that low. Using pingtest.net I get approximately 120ms at best (or to an east coast EC2 instance the same).

75ms to US not unreasonable, and have seen Be users with pings to our servers in london with an average of around the 7ms mark, i.e. a lot lower than most ISP's

In reply to a post by driz:

I ping New York at about 75ms from home I think. Be connection with fastpath in London and a 5ms first hop. Fibre might help me drop 4ms off that, though from the traceroutes I've seen from people on the BT trial, they still have a 4ms first hop.

Sounds unlikely - typical trans-atlantic routes from London to NY will be near 70ms and at best might achieve 60ms Add in you links though to that, plus routing in NY to the target will add 20-25ms

Can you provide a screenshot of from PingPlotter to prove me wrong?

Edited by MHC (Thu 15-Sep-11 11:12:34)

What would you expect for the specific London Telehouse/Telecity to the NY equivalent? We always worked on 75 to 80ms and we had several Gigabits of capacity on various links.

In reply to a post by MrSaffron:

75ms to US not unreasonable, and have seen Be users with pings to our servers in london with an average of around the 7ms mark, i.e. a lot lower than most ISP's

Many of them could plausibly have only been a short distance away from your test servers though.

For instance, given that using an expensive high priority business service from Verizon designed for low transatlantic latency achieved at best 78ms it seems unlikely to me.

http://www.verizonbusiness.com/about/network/latency/

In reply to a post by MHC:

Sounds unlikely - typical trans-atlantic routes from London to NY will be near 70ms and at best might achieve 60ms Add in you links though to that, plus routing in NY to the target will add 20-25ms

Can you provide a screenshot of from PingPlotter to prove me wrong?

I can't do that right now (as I'm at work), but here's a traceroute from a VPS in London:

traceroute to qw.quakephil.com (69.31.82.226), 30 hops max, 60 byte packets
1 212.111.33.229 (212.111.33.229) 0.790 ms 0.833 ms 0.874 ms
2 212.111.33.233 (212.111.33.233) 1.287 ms 1.342 ms 1.425 ms
3 te3-1-border76-01.lon2.telecity.net (217.20.44.217) 0.949 ms * *
4 * * *
5 195.50.113.1 (195.50.113.1) 0.998 ms 1.033 ms 1.026 ms
6 ae-34-52.ebr2.London1.Level3.net (4.69.139.97) 1.365 ms 1.174 ms 0.934 ms
7 ae-42-42.ebr1.NewYork1.Level3.net (4.69.137.70) 69.617 ms ae-43-43.ebr1.NewYork1.Level3.net (4.69.137.74) 69.652 ms ae-42-42.ebr1.NewYork1.Level3.net (4.69.137.70) 69.618 ms
8 ae-61-61.csw1.NewYork1.Level3.net (4.69.134.66) 70.262 ms ae-91-91.csw4.NewYork1.Level3.net (4.69.134.78) 79.120 ms 79.063 ms
9 ae-23-70.car3.NewYork1.Level3.net (4.69.155.69) 70.354 ms 70.310 ms 70.301 ms
10 PILOSOFT-IN.car3.NewYork1.Level3.net (4.53.88.10) 69.506 ms 69.722 ms 70.086 ms
11 69.172.128.1 (69.172.128.1) 70.068 ms 69.957 ms 69.941 ms
12 qw.quakephil.com (69.31.82.226) 69.807 ms 70.283 ms 70.066 ms

And here's my ping to bbc.co.uk from home (which I have saved in email drafts ;p)

Pinging bbc.co.uk [212.58.241.131] with 32 bytes of data:
Reply from 212.58.241.131: bytes=32 time=5ms TTL=250
Reply from 212.58.241.131: bytes=32 time=5ms TTL=250
Reply from 212.58.241.131: bytes=32 time=5ms TTL=250
Reply from 212.58.241.131: bytes=32 time=5ms TTL=250

I can hopefully combine the two about 630pm tonight!

In reply to a post by driz:

In reply to a post by MHC:

Sounds unlikely - typical trans-atlantic routes from London to NY will be near 70ms and at best might achieve 60ms Add in you links though to that, plus routing in NY to the target will add 20-25ms

Can you provide a screenshot of from PingPlotter to prove me wrong?

I can't do that right now (as I'm at work), but here's a traceroute from a VPS in London:

traceroute to qw.quakephil.com (69.31.82.226), 30 hops max, 60 byte packets
1 212.111.33.229 (212.111.33.229) 0.790 ms 0.833 ms 0.874 ms
2 212.111.33.233 (212.111.33.233) 1.287 ms 1.342 ms 1.425 ms
3 te3-1-border76-01.lon2.telecity.net (217.20.44.217) 0.949 ms * *
4 * * *
5 195.50.113.1 (195.50.113.1) 0.998 ms 1.033 ms 1.026 ms
6 ae-34-52.ebr2.London1.Level3.net (4.69.139.97) 1.365 ms 1.174 ms 0.934 ms
7 ae-42-42.ebr1.NewYork1.Level3.net (4.69.137.70) 69.617 ms ae-43-43.ebr1.NewYork1.Level3.net (4.69.137.74) 69.652 ms ae-42-42.ebr1.NewYork1.Level3.net (4.69.137.70) 69.618 ms
8 ae-61-61.csw1.NewYork1.Level3.net (4.69.134.66) 70.262 ms ae-91-91.csw4.NewYork1.Level3.net (4.69.134.78) 79.120 ms 79.063 ms
9 ae-23-70.car3.NewYork1.Level3.net (4.69.155.69) 70.354 ms 70.310 ms 70.301 ms
10 PILOSOFT-IN.car3.NewYork1.Level3.net (4.53.88.10) 69.506 ms 69.722 ms 70.086 ms
11 69.172.128.1 (69.172.128.1) 70.068 ms 69.957 ms 69.941 ms
12 qw.quakephil.com (69.31.82.226) 69.807 ms 70.283 ms 70.066 ms

And here's my ping to bbc.co.uk from home (which I have saved in email drafts ;p)

Pinging bbc.co.uk [212.58.241.131] with 32 bytes of data:
Reply from 212.58.241.131: bytes=32 time=5ms TTL=250
Reply from 212.58.241.131: bytes=32 time=5ms TTL=250
Reply from 212.58.241.131: bytes=32 time=5ms TTL=250
Reply from 212.58.241.131: bytes=32 time=5ms TTL=250

I can hopefully combine the two about 630pm tonight!

That is cheating . Running from a data centre directly to another data centre. I was talking about home grade bb connections.

In reply to a post by dustofnations:

That is cheating . Running from a data centre directly to another data centre. I was talking about home grade bb connections.

Yeah, but I ping the bbc datacentre at 5ms from home. So the theory (and practice, I'm fairly sure!) goes that I get about 75ms from home to NY!

But you'll have to wait 7 hours

In reply to a post by MrSaffron:

75ms to US not unreasonable, and have seen Be users with pings to our servers in london with an average of around the 7ms mark, i.e. a lot lower than most ISP's

On a somewhat related note (more to do with the OP before it was deleted). Have you tried to get an interview with these HyperOptic guys to see what they're supposedly up to? It'd be nice to hear some real technical details from them rather than marketing fluff which is pretty much useless.

I'd like to hear proper information about their IPv6 support, how they intend to support such huge bandwidth connections, what real world speeds people should expect during peak times, what their peering arrangements are given the massive connection speeds.

Other nice info would be national vs international bandwidth. For instance some fast ISPs abroad give high speeds within the country and more limited speed internationally (for obvious costs reasons).

If large numbers of home users suddenly had gigabit connections and hammered them, even LINX would fall on its face!

Edit: They don't even seem to be on the LINX members list, I'd have expected to see them there by now as they are supposedly beginning to connect people up.

Edited by deleted (Thu 15-Sep-11 12:13:37)

So where is the server in the US? I must be co-located with the transatlantic termination.

One of the NY terminations that our traffic would flow through is 195.66.224.130 currently that is giving an average of 94ms which if I remove the 12ms to get to the UK end still gives 82 ...


And looking at your figures again, you must be in one of the London centres - so 70 ms is what I suggest for the transatlantic hop.

It is a little better than one I was using last week - San Diego to TBB was around 700ms !

Edited by MHC (Thu 15-Sep-11 11:28:39)

In reply to a post by MHC:

So where is the server in the US? I must be co-located with the transatlantic termination.

One of the NY terminations that our traffic would flow through is 195.66.224.130 currently that is giving an average of 94ms which if I remove the 12ms to get to the UK end still gives 82 ...


And looking at your figures again, you must be in one of the London centres - so 70 ms is what I suggest for the transatlantic hop.

It is a little better than one I was using last week - San Diego to TBB was around 700ms !

Yep and that tallies with the Verizon figures. Given a home bb connection isn't usually going to be pushed through highest priority peering it would be amazing to see such low figures as presented earlier.

Anyway, I was working with an EC2 instance in Tokyo recently, and it was torturous. Even using SSH was noticeably laggy due to the huge latency. It feels like working through molasses .

70 to 80ms

So poster is not in the impossible area

I doubt many Be users live in Docklands, which ironically is a fairly bad area for broadband.

Tis true MrSaffron that the Canary Wharf part of docklands is very bad for broadband

However Be have been on the poplar exchange for some time - I was/am one of their first customers there. So I imagine Be have a fair number of E14 customers

The interesting thing with HyperOptic is that they target business/residential premises directly rather than per area/exchange

I've been moaning to BT/Virgin for many years re lack of fibre where I live - Annoyingly BT have FTTC in my area - but not to my development.

Hyperoptic have done a site survey where I live so all being well I'll be one of their first customers too

I'll feed back my expieriences once(if) connected

Edited by deleted (Thu 15-Sep-11 12:05:17)

I average 87ms to NY (69.31.82.226) on a residential Bethere connection with interleaving. I used to be on fastpath, and average 6ms to BBC and 75-90ms to much of the US east coast. This was fairly consistent too.

Edited by deleted (Thu 15-Sep-11 13:47:24)

Global Crossing are at 65ms for London-NY according to this recent article in the Telegraph., which also describes a new link that will shave 6ms off this.

However, the gamers will have to be able to outbid the hedge funds to get their hands on this.

At best 60ms is just possible assuming absolute minimal delays in teh equipment, perfect fibre, a perfect route &c.

The refractive index for fibre is between 1.5 and 1.6 (a vacuum is 1.0) and with the speed of light being 3x10⁸ms^-1 it gives a speed of light in fibre of (3/1.5)x10⁸ms^-1 or 2x10⁸ms^-1 The Hibernian Express cable is 6021km long - so forward and return needed for a successful "ping" assume 12x10⁶m

12x10⁶m = 6x10^-2s or 60ms
2x10⁸ms^-1

In reply to a post by Zadeks:

Already down to 9ms on FTTC. You're not going to see much more of a reduction on a residential network. Ping times to New York are 90ms. Perfectly playable.

TBH that could be improved on and should be, as latency to the USA from other neighboring EU countries can be and is half that, why Peering agreements here in the uk suck compared to them as does the current state of affairs with proper infrastructure for low latency / high speed bb

In reply to a post by driz:

But you'll have to wait 7 hours

7 hours later, the goods:

Tracing route to qw.quakephil.com [69.31.82.226]
over a maximum of 30 hops:

1 <1 ms <1 ms <1 ms dodgy.router [10.0.1.254]
2 6 ms 6 ms 6 ms 87-194-48-1.bethere.co.uk [87.194.48.1]
3 * * 6 ms 10.1.2.218
4 6 ms 6 ms 5 ms 208.178.245.81
5 6 ms 5 ms 5 ms ldn-b5-link.telia.net [213.248.81.73]
6 6 ms 6 ms 5 ms ldn-bb1-link.telia.net [80.91.252.205]
7 75 ms 74 ms 75 ms nyk-bb1-link.telia.net [80.91.248.203]
8 76 ms 75 ms 75 ms nyk-b4-link.telia.net [80.91.250.97]
9 75 ms 76 ms 76 ms pilosoft-ic-122192-nyk-b1.c.telia.net [213.248.88.182]
10 75 ms 76 ms 75 ms qw.quakephil.com [69.31.82.226]

For reference, I'm in southwest London.

In reply to a post by MHC:

So where is the server in the US? I must be co-located with the transatlantic termination.

To be honest, I'm not sure. But it's a server set up for the game QuakeWorld, which has a few hundred remaining players mostly concentrated around northern europe. I suspect that the server admin may have chosen its location based on increasing the likelihood of having europeans play with the tens of northern american based players left.

"Other nice info would be national vs international bandwidth. For instance some fast ISPs abroad give high speeds within the country and more limited speed internationally (for obvious costs reasons).

If large numbers of home users suddenly had gigabit connections and hammered them, even LINX would fall on its face!

Edit: They don't even seem to be on the LINX members list, I'd have expected to see them there by now as they are supposedly beginning to connect people up. "

No, you're thinking far too sensibly. You'll never get a marketing job.

Were you around when Be started up? One link from them to the outside world. No resilience, and not much capacity. Or something like that, I wish I could remember enough details to find a definitive reference.

I don't know what they are using in HE, but non-zero dispersion fibre (NZDF) is commonly available with a group index of 1.47 - enough to shave-off another millisecond across the Atlantic (or 953 microsec in this particular case, if we're picky); for low latency applications I'm seeing people pushing for 1.462 using the latest Corning glass.

Very interesting, somebody needs to invent a fibre cable that can let light pass through at closer to light speeds. If it was at true light speed we would see a link of sub 20ms from the UK to USA. That would be awesome. 60-70ms would be fine aslong as its stable for *some* games. But never for First Person Shooters, especially competitive FPS'.

Speaking of Fibre I was reading this article last night, which holds promise for "faster than light" technology in the future. The scientist is incredibly freaky looking LOL. But I won't hold that against him if he's onto something +)

http://www.livescience.com/790-light-travels-faster-...

In reply to a post by LeJimster:

If it was at true light speed we would see a link of sub 20ms from the UK to USA.

We're talking about round-trip times, so ping times would be a little over 40ms compared to the current 65. A shade under 59ms is possible with current low-index fibre.

Strictly speaking, it's group index rather than simple refractive index that matters for propagation of information. Various techniques, like the one in the link you provide, can produce weird phase effects, but the speed of a wave group carrying information is still no faster than the speed of light in a vacuum, and usually slower.

Quite a few in Poplar which serves a lot of Docklands and surrounding area

Was the first Be exchange to go live

Right, I didn't factor in the return trip *doh* =)

was it really the first?? did not know that

Apparently Hyperoptic have 10 launch sites in mind of which my development is one (if we sign up)

In reply to a post by JonathanCassar:

was it really the first?? did not know that

Apparently Hyperoptic have 10 launch sites in mind of which my development is one (if we sign up)

Do you know where they are?

Only my one (Lockes Wharf E14) - I was told when I phoned my development was one of ten. I'm not sure if all ten are firm. For example we haven't signed off yet but arent too far from doing so

Apparently if we sign off soon the install could be as early as next month...

St David's square E14 has just signed off on getting hyperoptic installed too

In reply to a post by Anonymous:

St David's square E14 has just signed off on getting hyperoptic installed too

its the same place Lockes Wharf is the Development Name!!

Any time line of rolling out in St David Square.. I will be moving in there soon.. and would not like to get on with a new operator like bt etc and directly move to high speed BB instead..

In reply to a post by MrSaffron:

75ms to US not unreasonable, and have seen Be users with pings to our servers in london with an average of around the 7ms mark, i.e. a lot lower than most ISP's

But only if they live within a specified distance, it's down to their points of presence PoP, & distance to that, nothing more,
Unfortunately for those who live north of the M25 we will never see sub 10ms to the same location, as they don't have PoP s in many other places , so is are a bit behind ,as other networks do have POP s up and down the uk

In reply to a post by Anonymous:

In reply to a post by MrSaffron:

75ms to US not unreasonable, and have seen Be users with pings to our servers in london with an average of around the 7ms mark, i.e. a lot lower than most ISP's

But only if they live within a specified distance, it's down to their points of presence PoP, & distance to that, nothing more,
Unfortunately for those who live north of the M25 we will never see sub 10ms to the same location, as they don't have PoP s in many other places , so is are a bit behind ,as other networks do have POP s up and down the uk

I would suggest that distance to the servers themselves and any peering and transit is also an issue given we're talking about RTTs to servers. For better or worse most transit and peering is concentrated in London and the South East.