4 ms·

There are easier ways to annoy HFT: "Every day for up to ten minutes near the London Stock Exchange, someone blocks signals from the global positioning system (

by Maxious 12y ago

There are easier ways to annoy HFT: "Every day for up to ten minutes near the London Stock Exchange, someone blocks signals from the global positioning system (GPS) network of satellites ... timestamps on trades made in financial institutions can be affected. " http://www.economist.com/news/international/21582288-satellite-positioning-data-are-vitalbut-signal-surprisingly-easy-disrupt-out http://www.economist.com/news/international/21582288-satelli...

minimax 12y ago
Throwing off the exchange generated timestamps doesn't mess with any HFTs. The exchanges themselves don't synchronize against one another so the timestamps are basically useless to begin with.
- kasey_junk 12y ago
  If you are timestamping in a data center and they throw off the timestamp source for the entire data center that could theoretically cause you issues.
  - easytiger 12y ago
    Anyone serious will have their own oscillator with enough Holdover to get them to sort the problem out.
    - kasey_junk 12y ago
      I think it is outlandish to suggest, but in theory you could time a GPS attack long enough and timestamping requirements tight enough to make the drift a problem. I can't think of a single real use case for that level of timestamping accuracy nor a reason you would want to mess with it. But you could.
      - easytiger 12y ago
        PLenty of things need that level of accuracy. But it isn't just about the accuracy it is about the immediate degrees of drift that can occur. With GPS time disciplining removed a crap OCXO (and there are plenty) will wander off back and forth in mad errant degrees. With this used to discipline dozens of other hosts results are not always pleasant.
      - wglb 12y ago
        Likely more important is the synchronization of time within a data center.
easytiger 12y ago
Yea that article is a bit off & largely a marketing piece. For a start, if you disrupt (rather than MITM attack) the GPS signal, the internal time distributor host (prob sending PTP & NTP) will fall back to their internal oscillator. Usually these are Rubidium or caesium. They will have a hold over of enough days to sort out the issue. Secondly, Chronos Technology are trying to sell a service where you don't need to rely on GPS but they will come to your DC with a highly portable time source recently GPS synced and discipline your caesium or rubidium oscillator every n days. They are trying to be the modern day Ruth Belville*. It is useful as an idea in cases where we don't have an easy way to set up GPS in the DC and you can live with a relative clock domain. http://en.wikipedia.org/wiki/Ruth_Belville http://en.wikipedia.org/wiki/Ruth_Belville

by Maxious 12y ago

minimax 12y ago

Throwing off the exchange generated timestamps doesn't mess with any HFTs. The exchanges themselves don't synchronize against one another so the timestamps are basically useless to begin with.

kasey_junk 12y ago
If you are timestamping in a data center and they throw off the timestamp source for the entire data center that could theoretically cause you issues.
- easytiger 12y ago
  Anyone serious will have their own oscillator with enough Holdover to get them to sort the problem out.
  - kasey_junk 12y ago
    I think it is outlandish to suggest, but in theory you could time a GPS attack long enough and timestamping requirements tight enough to make the drift a problem. I can't think of a single real use case for that level of timestamping accuracy nor a reason you would want to mess with it. But you could.
    - easytiger 12y ago
      PLenty of things need that level of accuracy. But it isn't just about the accuracy it is about the immediate degrees of drift that can occur. With GPS time disciplining removed a crap OCXO (and there are plenty) will wander off back and forth in mad errant degrees. With this used to discipline dozens of other hosts results are not always pleasant.
    - wglb 12y ago
      Likely more important is the synchronization of time within a data center.

easytiger 12y ago

Yea that article is a bit off & largely a marketing piece. For a start, if you disrupt (rather than MITM attack) the GPS signal, the internal time distributor host (prob sending PTP & NTP) will fall back to their internal oscillator. Usually these are Rubidium or caesium. They will have a hold over of enough days to sort out the issue. Secondly, Chronos Technology are trying to sell a service where you don't need to rely on GPS but they will come to your DC with a highly portable time source recently GPS synced and discipline your caesium or rubidium oscillator every n days. They are trying to be the modern day Ruth Belville*. It is useful as an idea in cases where we don't have an easy way to set up GPS in the DC and you can live with a relative clock domain. http://en.wikipedia.org/wiki/Ruth_Belville http://en.wikipedia.org/wiki/Ruth_Belville