Could the Americans switch off GPS ?

[jrc1983]

Nasa have just indicated that there is intense electromagnetic activity related to a "hole" in the sun's corona. They anticipate problems with GPS signals and G4 reception over the next two months.


http://fr.news.yahoo.com/trou-à-surface-soleil-perturbe-portables-124924259.html

An interesting point. It seems that all the "GPS is infallible and anyone who says different is a nut" people have forgotten that it only takes a single solar event to knock the entire system out of kilter, and that applies equally to any satellite based navigation system.

 

[grumpy_o_g]

You'll also be disappointed to discover that Google's entire (in house developed) NoSQL equivalent, upon which a lot of their infrastructure now runs is entirely synched by GPS too then.......

Its not as clear cut as people would like to make out.... there is now so much interdependency between the big players on the network, that only a few having major problems could cause a significant effect on many others.... its OK saying that switches can handle poor clocks... but not when it gets to core switching between different AS environments, when speeds and volumes, and hence timing become a much bigger issue.... and a lot of the big players are not only massive content and service providers, but also key parts of the core switching network...

So, yes, a few global telcos do still maintain their own timing services, but a lot of mid tier telcos and service providers don't.... and they are just as critical to a balanced service nowadays.

All the Banks I know do not rely on GPS timing though they do use it. If they relied on it they'd fail to meet regulatory requirements and also be bankrupted the first time GPS failed to provide a time signal to them whilst other Banks were able to carry on trading. The Telco's we use certainly don't rely on GPS alone either (at least that's what they tell us) as they are subject to the same rules by virtue of us using them at least for certain critical circuits. If we don't actually need a time signal for trading reasons then the clock in an NTP server is usually good enough for us frankly. Eventually someone will notice servers starting to get out of synch and raise a low priority ticket for it to be fixed it goes wrong. It's pretty rare and very poor practice for someone that has timing requirements to rely on GPS when there's so many off-the shelf solutions for radio and GPS. Google may have one or two services where they solely use GPS but I'm not aware of any of their really big server farms where that's the case.

To all those saying the US can't turn off GPS - of course they can. They may have promised they won't and I'd struggle to think of a circumstance where they'd want to but they can and, if it suited their military in a Time of War situation they probably would. Just as would any other country. Lose New York, LA, Boston, Washington and Seattle or switch off GPS - it's not a very tricky decision, I just can't think of a situation where GPS would make that difference.

And if there's ever a war big enough to turn off GPS I'd want to know exactly where I was and how to get to where I wanted to go thanks. Far more so than if my life possibly didn't depend on it.

 

[lustyd]

All the Banks I know do not rely on GPS timing though they do use it. If they relied on it they'd fail to meet regulatory requirements and also be bankrupted the first time GPS failed to provide a time signal to them whilst other Banks were able to carry on trading. The Telco's we use certainly don't rely on GPS alone either (at least that's what they tell us) as they are subject to the same rules by virtue of us using them at least for certain critical circuits. If we don't actually need a time signal for trading reasons then the clock in an NTP server is usually good enough for us frankly. Eventually someone will notice servers starting to get out of synch and raise a low priority ticket for it to be fixed it goes wrong. It's pretty rare and very poor practice for someone that has timing requirements to rely on GPS when there's so many off-the shelf solutions for radio and GPS. Google may have one or two services where they solely use GPS but I'm not aware of any of their really big server farms where that's the case.

To all those saying the US can't turn off GPS - of course they can. They may have promised they won't and I'd struggle to think of a circumstance where they'd want to but they can and, if it suited their military in a Time of War situation they probably would. Just as would any other country. Lose New York, LA, Boston, Washington and Seattle or switch off GPS - it's not a very tricky decision, I just can't think of a situation where GPS would make that difference.

And if there's ever a war big enough to turn off GPS I'd want to know exactly where I was and how to get to where I wanted to go thanks. Far more so than if my life possibly didn't depend on it.

Specifically what are they/you using instead? I'm assuming that most of them don't buy their own atomic clocks, and NTP servers are not only less reliable but generally use GPS because it's cheap. Radio sources are just as easy to jam so you've got me curious.

 

[AntarcticPilot]

Not that can be timed accurately.... atomic clocks are great.... but are on the network, with therefore unreliable certainty over latency.... only any use if they are local... and too many aren't, so would cause an avalanche of outages that would eventually take everything down... very few telcos for example now have their own clocks, so all bandwidth (inc telehouses with core traffic switching) would probably fall over fairly quickly.

The NTP4 protocol allows time sharing across the internet with millisecond accuracy; plenty good enough for most purposes. Applications requiring higher accuracy will already have arranged synchronized time-sources without relying on GPS.

 

[grumpy_o_g]

Specifically what are they/you using instead? I'm assuming that most of them don't buy their own atomic clocks, and NTP servers are not only less reliable but generally use GPS because it's cheap. Radio sources are just as easy to jam so you've got me curious.

For stuff that needs an accurate time signal such as most of our low-latency trading apps Radio is the only realistic alternative. Yes it can be jammed but they are the only two methods of meeting regulatory requirements without laying dark fibre. In fairness, if a situation arose where GPS was being jammed on a large scale or had been switched off trading would probably have been suspended anyway.

 

[grumpy_o_g]

The NTP4 protocol allows time sharing across the internet with millisecond accuracy; plenty good enough for most purposes. Applications requiring higher accuracy will already have arranged synchronized time-sources without relying on GPS.

And one day the regulatory authorities will catch up with it The trading reqs are very tight but we're looking closely at NTP4.

 

[laika]

All the Banks I know do not rely on GPS timing though they do use it. If they relied on it they'd fail to meet regulatory requirements and also be bankrupted the first time GPS failed to provide a time signal to them whilst other Banks were able to carry on trading.

The couple that I actually know what their NTP infrastructure looks like use only GPS in their top NTP stratum. They obviously have a few distributed over at least 3 continents. Personally I think this is silly (what harm in at least one radio clock?), but don't want to go off on a rant about the staff you're left with when you sack all the expensive IT folk and outsource everything...

Depending on what regulatory requirements you mean, they tend to be fairly loose (in computer timing terms). The NASD for example needs timestamps within 3 seconds from UTC. With decent hardware at the top of the NTP tree it should take months (at least) to drift far enough to cause a regulatory problem, and that's assuming your GPS was taken out in all geographic locations.

I think we have this thread annually, so anticipating the misunderstandings from last time, we're talking about an internal NTP infrastructure underneath some internally hosted GPS receiver/NTP server devices. "The Internet" is involved.

Boaty angle? Well I'm sat in the boatyard cafe writing this and working out whether to revive my plan of using the PPS output from a GPS chip I have to turn boat's raspberry pi into an NTP server...

Edit: Yes and anticipating further comments, the low latency stuff uses PTP rather than NTP, usually with dedicated GPS clocks...

Later edit: should emphasise that I'm not questioning grumpy_o_g's assertion, but clearly we worked in different banks

 

[grumpy_o_g]

The couple that I actually know what their NTP infrastructure looks like use only GPS in their top NTP stratum. They obviously have a few distributed over at least 3 continents. Personally I think this is silly (what harm in at least one radio clock?), but don't want to go off on a rant about the staff you're left with when you sack all the expensive IT folk and outsource everything...

Depending on what regulatory requirements you mean, they tend to be fairly loose (in computer timing terms). The NASD for example needs timestamps within 3 seconds from UTC. With decent hardware at the top of the NTP tree it should take months (at least) to drift far enough to cause a regulatory problem, and that's assuming your GPS was taken out in all geographic locations.

I think we have this thread annually, so anticipating the misunderstandings from last time, we're talking about an internal NTP infrastructure underneath some internally hosted GPS receiver/NTP server devices. "The Internet" is involved.

Boaty angle? Well I'm sat in the boatyard cafe writing this and working out whether to revive my plan of using the PPS output from a GPS chip I have to turn boat's raspberry pi into an NTP server...

Edit: Yes and anticipating further comments, the low latency stuff uses PTP rather than NTP, usually with dedicated GPS clocks...

Later edit: should emphasise that I'm not questioning grumpy_o_g's assertion, but clearly we worked in different banks

Certainly the low-latency and HFT systems need a lot less - we're talking sub 3 mSecs. I'll qualify this to say I'm talking about Investment Banks but these rules should apply to all trading. Even a manual trade has a strict latency requirement in low mSecs to avoid the risk of two people inadvertently buying/selling the same stock. It not absolute time stamps that are the problem but being able to prove latency from one point to another.

 

[lw395]

All those waffling on that the US could not turn off GPS due to the impact on aviation etc, might care to remember that the US grounded all aircraft for a while after 9/11. That was not even in a state of war IIRC.
It's possible that the impact on trading systems would be messy, but that's just bad planning on the part of IT cowboys, plus a large measure of non-understanding.
But the markets would go ape if we had another 9/11 and an excuse to switch them off for a bit might be welcome?
As others have said, it's very possible to jam GPS and all the other satnav systems over big areas.
Are the Yanks still telling us that drone was brought down in Iran by spoofing GPS? And are we believing them?

 

[laika]

It not absolute time stamps that are the problem but being able to prove latency from one point to another.

I'd be interested what regulations you mean (or is this just a question of when the models break down and your quants start throwing things?), but surely if we're no longer talking about drift from UTC I think that is a different topic (and we've already got pretty far away from anything boaty)

 

[grumpy_o_g]

I'd be interested what regulations you mean (or is this just a question of when the models break down and your quants start throwing things?), but surely if we're no longer talking about drift from UTC I think that is a different topic (and we've already got pretty far away from anything boaty)

I'll reply by pm as you're right, is very non-boaty.