The importance of 'ping' - latency/connection speed in trading.

Discussion in 'Order Execution' started by gotmessner, Sep 8, 2005.

  1. Hi New Guy,

    Your reasoning is quite solid. For traders not having a direct setup with an exchange, you still have to figure in the time required by the broker's computer to check the authorization and margin/funding status of the account.

    As I pointed out earlier, such discussions have been going for a long time already. Most millisecond-ping-dreamers always forget to account for these administrative processing times making a few miliseconds utterly insignificant under typical conditions.

    Be good,
    nononsense
     
    #51     Sep 23, 2005
  2. That's an interesting take... to be honest, I never considered that. I assumed that those parameters were loaded into my trading platform, and were checked as soon as I clicked the "make a bad trade" button, but now that you mention it, sometimes there is a spot of latency there.

    I'm guessing that peice of software resides on the gateways to the exchanges, within my brokers walls. The latency on that is probably tiny comapred to the internet latency just getting there (I don't direct connect for my personal account). But it is definately concievable if there is a seperate internet call to a seperate set of servers that it could double your latency (if that's how your BD is set up...).

    Thanks for the thought,

    -The New Guy
     
    #52     Sep 23, 2005
  3. tomcole

    tomcole

    My view is simply that if there are latencies along the way, the faster your order gets handled the better.

    I think its a very reasonable question to ask brokers about their systems, call centers etc.
     
    #53     Sep 23, 2005
  4. I did a ping test to IB data server

    ping gw1.ibllc.com

    I am on broadband

    Are these results good? I was told time had to be less than 1000ms, but what are the optimal values?


    Pinging gw1.ibllc.com [208.245.107.3] with 32 bytes of data:

    Reply from 208.245.107.3: bytes=32 time=1292ms TTL=107
    Reply from 208.245.107.3: bytes=32 time=96ms TTL=107
    Reply from 208.245.107.3: bytes=32 time=96ms TTL=107
    Reply from 208.245.107.3: bytes=32 time=97ms TTL=107

    Ping statistics for 208.245.107.3:
    Packets: Sent = 4, Received = 4, Lost = 0 (0% loss),
    Approximate round trip times in milli-seconds:
    Minimum = 96ms, Maximum = 1292ms, Average = 395ms

    Thanks
     
    #54     Sep 29, 2005
  5. For broadband, that's pretty good. 100 or less is very respectable, but the first one:

    Reply from 208.245.107.3: bytes=32 time=1292ms TTL=107

    is a bit of cause for concern. Do a ping -t and get a bunch of results (cntrl c to exit) and then you can see if that is a reoccuring theme over time. Even if it is, however, it might not be indicative of the problem because cisco routers (among others) will prioritize traffic and drop pings if they get busy. However, if they are that busy then there might be a problem anyway....

    Anyway, here are my results to yahoo. I beleive my set up to be about as fast as you can get right now.


    Microsoft Windows XP [Version 5.1.2600]
    (C) Copyright 1985-2001 Microsoft Corp.

    C:\Documents and Settings\comp>ping -t www.yahoo.com

    Pinging www.yahoo.akadns.net [68.142.226.46] with 32 bytes of data:

    Reply from 68.142.226.46: bytes=32 time=24ms TTL=54
    Reply from 68.142.226.46: bytes=32 time=27ms TTL=54
    Reply from 68.142.226.46: bytes=32 time=24ms TTL=54
    Reply from 68.142.226.46: bytes=32 time=26ms TTL=55
    Reply from 68.142.226.46: bytes=32 time=56ms TTL=55
    Reply from 68.142.226.46: bytes=32 time=25ms TTL=54
    Reply from 68.142.226.46: bytes=32 time=25ms TTL=54
    Reply from 68.142.226.46: bytes=32 time=25ms TTL=55

    Ping statistics for x.x.x.x:
    Packets: Sent = 8, Received = 8, Lost = 0 (0% loss),
    Approximate round trip times in milli-seconds:
    Minimum = 24ms, Maximum = 56ms, Average = 29ms

    As you can see, there's a high one in there too, but it's about double the average, vs ten times like in your example. If the majority of your times are 100 or less, you're probably doing very well.

    -The New Guy
     
    #55     Sep 29, 2005
  6. tomcole

    tomcole

    Try going to dslreports.com and checking out their resources - could be helpful!
     
    #56     Sep 29, 2005
  7. Thanks, the new ping test shows similar times, except again for the first one, but the average is 124ms
     
    #57     Sep 29, 2005
  8. alanm

    alanm

    FWIW, my API application in California, talking to an Assent server in NJ, takes 90-120ms between the time I submit an order to INET via the local API function and the time I get an "order live" ack back from the API event. This seems remarkably low to me, given that my ping times to the server are 85-90ms when the connection is "quiet". I'm certain that I'm timestamping accurately. I'm remembering that ICMP packets (used by ping) enjoy a much lower priority than other traffic, so it would appear that the actual times for "real" traffic must be quite low.

    Just the physical roundtrip distance (about 5500 miles), accounts for 30ms at the speed of light.
     
    #58     Sep 29, 2005
  9. Hi thenewguy

    How do you manage to get much better timings, are you not just using a modem/router and broadband ?

    Thanks
     
    #59     Sep 30, 2005
  10. Sorry, I should have clarified... I'm at a trading office, where I run the IT set up. We run on dual DS3's that have a very short distance to a DC that is a super POP that most everything we connect to is in also located in. From my home the times would be much, much different.

    Thanks,

    -The New Guy
     
    #60     Sep 30, 2005