Results 1 to 8 of 8
  1. #1

    Real world latency expectation for metro ethernet

    Hello All,

    My company is currently using Amazon Web Services and we are considering using AWS Direct Connect (aws.amazon.com/directconnect) to increase bandwidth and decrease latency.

    We have received proposals from both Level 3 and TWTC and are leaning toward TWTC because they already have fiber relatively near our location so the build fees will be much lower.

    One of our key requirements is a reduction in latency. At present, we are using a VPN connection between us and AWS and latency averages 35ms.

    How can I determine what the latency will be over the ethernet circuit before I commit to it? Naturally I will ask the vendor to comment on this, but from what I've read online most are hesitant to do so and their SLAs are fairly vague.

    The distance between us and the AWS peering point is approx 819 miles. Dividing that distance by the speed of light, I arrive at 4.81ms, which would be great. So my question is how much additional latency "overhead" is typical on a connection such as this? I assume I also have to factor some time in for the various equipment along the way to do its' job. Would it be realistic to expect sub-10ms latency?

  2. #2
    Join Date
    Aug 2003
    Location
    East Coast
    Posts
    2,082
    You can ask the ISP to provide you with a traceroute/ping from the router that would be your gateway to the IP in question. This will provide you with half of the equation. Then you can have them provide you with a realistic expectation of the latency between that router and your location.

    My metro connection latency is <1ms over a distance of 40 miles but it will depend on your provider.

  3. #3
    Join Date
    Mar 2006
    Location
    Reston, VA
    Posts
    3,131
    Metro-E should always have way less latency than a VPN.

    We see this all the time and most metro-E circuits we roll out, say in the DC metro market would see anywhere from 2ms to 5ms depending on a few factors.

    Though I have seen metro-E circuits from cogent that customers have to their colo racks take 15MS to go from washignto DC to reston VA....

    If you are 820 miles from the aws direct connect session and datacenter. I would venture to say anywhere from 8ms to 15ms is the latency you will be seeing.
    Yellow Fiber Networks
    http://www.yellowfiber.net : Managed Solutions - Colocation - Network Services IPv4/IPv6
    Ashburn/Denver/NYC/Dallas/Chicago Markets Served zak@yellowfiber.net

  4. #4
    Join Date
    Nov 2005
    Posts
    3,944
    Quote Originally Posted by Fred Weston View Post
    The distance between us and the AWS peering point is approx 819 miles. Dividing that distance by the speed of light, I arrive at 4.81ms, which would be great. So my question is how much additional latency "overhead" is typical on a connection such as this? I assume I also have to factor some time in for the various equipment along the way to do its' job. Would it be realistic to expect sub-10ms latency?

    You have to double it, it is a round trip latency, so ~10ms right there. You will also have equipment delays, so I would assume 12-15ms is the realistic minimum. This is all assuming that you have the best route possible, but likely, the fiber does not run as the crow flies.

  5. #5
    Join Date
    Aug 2004
    Location
    Kauai, Hawaii
    Posts
    3,799
    Where is most of the latency currently on your AWS route? I see a lot of latency is often generated inside the amazon datacenters during peak times, not actually getting to the datacenter.

    Make sure that it is actually the latency getting to the DC that is causing you problems before you go and sign a long term fiber contract.

  6. #6
    Join Date
    Mar 2009
    Location
    Denver, CO
    Posts
    66
    Here are some real world examples:
    282 Miles - round-trip min/avg/max = 4/5/8 ms (actual fiber length, measured via OTDR)

    All these distances are road miles:
    415 Miles - round-trip min/avg/max = 8/10/12 ms (L 3)
    388 Miles - round-trip min/avg/max = 8/10/12 ms (Zayo)
    70 Miles - round-trip min/avg/max = 1/3/4 ms (L 3)
    2100 Miles - round-trip min/avg/max = 44/48/124 ms (TWTC)

  7. #7
    Join Date
    Apr 2010
    Posts
    493
    Quote Originally Posted by Fred Weston View Post
    Hello All,

    My company is currently using Amazon Web Services and we are considering using AWS Direct Connect (aws.amazon.com/directconnect) to increase bandwidth and decrease latency.

    We have received proposals from both Level 3 and TWTC and are leaning toward TWTC because they already have fiber relatively near our location so the build fees will be much lower.

    One of our key requirements is a reduction in latency. At present, we are using a VPN connection between us and AWS and latency averages 35ms.

    How can I determine what the latency will be over the ethernet circuit before I commit to it? Naturally I will ask the vendor to comment on this, but from what I've read online most are hesitant to do so and their SLAs are fairly vague.

    The distance between us and the AWS peering point is approx 819 miles. Dividing that distance by the speed of light, I arrive at 4.81ms, which would be great. So my question is how much additional latency "overhead" is typical on a connection such as this? I assume I also have to factor some time in for the various equipment along the way to do its' job. Would it be realistic to expect sub-10ms latency?
    That is just the distance to the closest peering point for you. Where are you provisioning VM's compared to that? If your going in the wrong direction compared to your VM you could be adding latency. 35ms sounds like your going about half the US assuming normal choke points. Sub 10ms would pretty much requires a very direct route and the gear being at the far end that is expecting a lot.

    VPN's should not add anywhere near 25ms of latency. If your VM's are at that peering point your taking a suboptimal IP route (some traceroute sleuthing can tell you this) if not your going to need to move them to see any real gains.

  8. #8
    Ask both providers for a latency test from a drop point that is near your current location if they have one. It will at least give you a ballpark figure of where you will be.

    As stated before, figure out which provider will be able to provide the most direct routing as well. Build out costs will be the cheap part in the long run.

Similar Threads

  1. Comcast Metro Ethernet thoughts
    By boggin in forum Colocation, Data Centers, IP Space and Networks
    Replies: 9
    Last Post: 03-23-2012, 05:49 PM
  2. Metro Ethernet
    By mdierolf in forum Colocation, Data Centers, IP Space and Networks
    Replies: 10
    Last Post: 08-05-2010, 09:06 AM
  3. Need Rep Info for Metro Ethernet
    By dbuyer0 in forum Colocation, Data Centers, IP Space and Networks
    Replies: 7
    Last Post: 01-14-2010, 09:35 PM
  4. Metro Ethernet vs. SONET
    By OneLittleBird in forum Colocation, Data Centers, IP Space and Networks
    Replies: 19
    Last Post: 11-18-2009, 08:53 AM
  5. Metro Ethernet vs. DS3
    By falconinternet in forum Colocation, Data Centers, IP Space and Networks
    Replies: 8
    Last Post: 07-23-2008, 03:41 AM

Tags for this Thread

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •