Best internet connection for scalping- trying to decide between cable and dsl ?

Discussion in 'Hardware' started by nic05, Apr 5, 2008.

  1. Rogers told me they wouldn't install a business line to a residential address aswell. I am only offered the consumer products, whilst Bell installed Business line without asking too many questions.
     
    #31     Apr 7, 2008
  2. Get a point to point T1. Any connection that propagates through the internet is way too inconsistent and unreliable for a scalper
     
    #32     Apr 7, 2008
  3. Since you are scalping, neither of them on their own is reliable enough to base your business on. If you look, rarely does even Business sold DSL or cable come with service level guarantees like a T1 would.

    So, the answer is:

    1. Get them both and get a router with autofailover. Do NOT buy a cheap router (linksys, netgear etc) as they are not reliable enough in their failover capabilities to ensure it'll work perfectly when you'd actually need it. Get at least a Peplink or better, a Cisco router.

    Also, some here have suggested load balancing. That might work if the broker would allow two IP's logged into your same account. Or perhaps your better router will take both the DSL and cable and commingle their IP's into one public IP that would work with the broker.

    2. A T1. Not as good as two different services with a fail over like #1. But the chances of a T1 going down are a lot less than DSL or cable as T1s have SLA's that ensure your ISP will work quickly to get it going again. But it still could be 4 hours. But DSL/cable is usually considered 'best effort' by the ISP meaning it'll work when it works - no promises.

    Me? Get them both and get the Peplink Balance 30 router.

    But as others have said, until you've had them both and personally seen ping times and real world usage when you're scalping, you'll never know which one's 'better.'
     
    #33     Apr 7, 2008
  4. Kap

    Kap

    I am moving soon (UK)... anyone have experience of Virgin media cable Broadbaband , fibre optic apprentley, compared to good ol ADSL.. it will be 2 km form exchange so apparently that will halve the available speed compard to cable which states its independent of distance.

    all info is good, thanks, Kap.
     
    #34     Apr 7, 2008
  5. it works much different in the UK they have fair usage policy which means for certain hours that they deem to be peak you get very little in data transfer limit. The only way i was able to get round it was to have a Business Broadband service from BT. Costs about £40 per month. No limit on that, 8 meg speeds, very reliable ( i used mine for 2 years and never had a problem).
     
    #35     Apr 7, 2008
  6. The only thing that I heard that even comes close to what you have presented is when Comcast sought to block "file-sharing" in the U.S. last Fall. Comcast blocked (hindered) some uploads when it came to applications such as BitTorrent, eDonkey, etc.

    Late last month, Comcast decided to reverse its position on blocking/hindering these types of file-sharing uploads.

    http://abcnews.go.com/Technology/wireStory?id=4535060

    Is this what you were talking about when you said that Comcast was "throttling" back some of its bandwidth?
     
    #36     Apr 7, 2008
  7. Quote from
    Landis82

    You need to do a TRACE ROUTE ( off your DOS prompt in Windows ) to the IP address of the trading/quotation server that your trading platform connects to.

    In this way, you can see what kind of "latency" exists in the route that is taken to get back to the server which provides you with your data-feed.

    More often than not, you have NO CONTROL over this kind of latency due in part because you are hostage to the Internet Superhighway and your local broadband provider has very little to do with this ( outside of their local network ).

    Concerning a trace-route, any millisecond numbers < 80 are good. Above 80-90 and you start to get a bit concerned.

    At the DOS prompt type: tracert(space)IP address of the server


    How do you read the trace route?
    I am seeing 12 different ms numbers
    Do you add them up? Or do you just take the read from the IP address you did the trace route on, which is the last read on the trace? I did it on different IP address to to compare and noticed some time outs as well. Can you please explain.

    TY
    Trader
     
    #37     Apr 7, 2008
  8. nic05

    nic05

    Well after talking to my bm today, I found out that we are no longer using a t1 at the office after all. Apparently we are trading on a business cable package now. That would explain my rather slow and fluctuating speed test results (the highest i had today was about 8000 kbps down and 488 up ) . The package they have is 10 Mbps down 2 up and the one I would get is 16 down 1.5 up. So I am assuming that I wouldn't have a problem ??
     
    #38     Apr 7, 2008
  9. GTS

    GTS

    Each hop of the traceroute shows you the latency from you to that hop. You don't add up the hops and really its the last hop (your final destination) that matters - the intermediate hops gives you can idea about where bottlenecks may be (e.g. if you see a sudden jump in the latency measurements)

    Timeouts can occur because not all devices honor icmp packets that traceroute uses - they may refuse to answer them because there are network attacks that use ICMP.

    Traceroute is a good start but some devices de-prioritize icmp packets (real traffic is given priority) so it may not give a true picture of what your end-to-end performance will be, but it is definitely a first step in evaluating what is going on.
     
    #39     Apr 7, 2008
  10. Quote from
    GTS

    Each hop of the traceroute shows you the latency from you to that hop. You don't add up the hops and really its the last hop (your final destination) that matters - the intermediate hops gives you can idea about where bottlenecks may be (e.g. if you see a sudden jump in the latency measurements)

    Timeouts can occur because not all devices honor icmp packets that traceroute uses - they may refuse to answer them because there are network attacks that use ICMP.

    Traceroute is a good start but some devices de-prioritize icmp packets (real traffic is given priority) so it may not give a true picture of what your end-to-end performance will be, but it is definitely a first step in evaluating what is going on.

    TY

    So my read for line 12, last line in test
    38ms 37 ms 38ms gw1.ibllc.com{208.245.107.3}

    The above was the last line of the trace
    to IB
    So this would be a good trace number?

    When I do a ping test to IB
    I get bytes =32 time 37,\ms TTL=111
    4 times
    I have always had good fills so I am guessing the above test are good?

    Trader
     
    #40     Apr 7, 2008