Ping times to foreign data server

Discussion in 'Data Sets and Feeds' started by rsi80, Mar 30, 2010.

  1. rsi80



    What would be the highest (worst) latency acceptable for the data server of a foreign market?

    250ms? 300ms? 350ms? 400ms? 450ms? 500ms?

    What if latency varies between a lower level (say, 300ms) and a higher one (say, 500ms)? Would you find such an internet connection usable?

    It is assumed that trading is discretionary and non-scalping.

    Further, is a packet loss rate not exceeding 1% acceptable?

    Thanks for your feedback.

    Fiber propagation is approx 5ms / 1000km

  3. rsi80


    Thanks for your reply.

    My query concerns what level of latency would be acceptable in a real-world environment from a trader's perspective. Just want to get some practical feedback from fellow ETers.
  4. That so vague, it's all relative. Last time I checked from Denver to Chi I am about 50ms and that fine. If you are in Europe to the US 200ms is reasonable. If you are collocated at exchange 1-2 ms. Lookup and download pingplotter and test it out.
  5. At a Colo full order process best case is 25ms. (Quote - Order - Clear Credit - Execute - fill & clear). If your playing the sub second HFT game using market orders you should be in a colo. If your staging limit orders it makes little difference as your orders are queued. Your price quotes are delayed by your one way net latency.

  6. rsi80


    Again, thanks for your feedback, guys.

    Nothing remotely like HFT here. :) Just another run-of-the-mill discretionary trader.

    Reading between the lines (of your comments), it seems that any latency not exceeding 500ms may be acceptable in practical terms. Fair comment?
  7. Latency is relatively moot if you are simply staging limit orders, they queue on the exchange order book. Your quotes may be delayed by 250ms (1/2 of the 500ms round trip). You will have issues if you are constantly canceling and replacing orders or perceived slippage when utilizing market orders. All issues are fairly manageable if you understand the constraints.

  8. rsi80


    Thanks for the clear, helpful pointers, PocketChange.
  9. I've traded ES from Asia with 300ms-400ms ping times to my broker's US servers. I monitor on 5m bars and have noticed no difference in terms of partials or slippage on 20 contracts from when I trade in the States where I usually get ping times under 100ms. For me latency is not the problem.

    The problem comes from packet loss which causes disconnects to the servers and I have to log back on. The problem is intermittent but it is rare I get through an evening without getting disconnected at least once. I trade only the first 2 hours or so of the regular session.

    I don't know how to measure % packet loss so I can't help you there.
  10. JackR


    If you are interested in packet loss as well as ping time you can download PingPlotter from Scroll to the bottom of the download page from a limited freeware version.

    Pingplotter will show you the links to your broker along with ping time and any packet loss at each router along the way. Remember that the way the internet (TCP) protocol works is that every router along the way checks the message (actually it checks each packet - small segments of your message) for errors and requests that a bad packet be resent from the upstream router before it sends the packet to the next router downstream from it. Thus, you may see packet loss in the middle of the route and none further downstream. That is because the packet has been retransmitted. This will cause a delay in the receipt of your message but will probably not show up in the ping times unless things are really bad.

    #10     Mar 31, 2010