[TERA PC & Console] En Masse is closing, but TERA lives on! We will continue to support TERA PC (NA) and TERA Console until service is transferred. Stay tuned for more information.
[TERA Console] The Grotto of Lost Souls update (v85) is now live! Read the patch notes here: https://bit.ly/TERACon_v85
[TERA PC] The 64-bit update (v97) is now live. Check out all the changes delivered on August 11 here: https://bit.ly/tera64_patchnotes
[TERA PC & CONSOLE] Summerfest Part 2: The Beach Bash is on from August 11 until September 1! Participate in event activities to earn tokens redeemable for costumes, consumables, mounts, and more! Details: https://bit.ly/tera_sf20
Can you move tera to another datacenter? please?
I don't really care where you put it, as long as wherever you move it is not controlled by centurylink(savvis). Their peering has been notoriously bad for quite a few years, and since load balancing stopped happening, has given us problems more or less constantly.
Lots of people don't really understand peering, and that's reasonable! It's very complicated. A simplified way of thinking of it is that a peering location is like a freeway on-ramp. Whether the freeway is there or not doesn't matter if there's no way to get on it.
What's been happening with centurylink is that they don't peer properly with many other providers (their on-ramps are in bad places), leading to situations where a user in chicago connecting to a server in chicago may be routed through atlanta because that's the nearest peering node which will accept their connection. Indeed, if we look at a traceroute from cogent's chicago node to tera's servers, we see them being routed through atlanta at present. Not only does this add latency (at least 30ms in this case) but it also adds quite a few hops to the route, which leads to less stability - more opportunities for packet loss or latency spikes.
This isn't a problem that centurylink is intent on fixing, they have no reason to as it would cost them money and not get them more money. I also get the impression, since this has been problematic for months, that their datacenter also does not have any interest in re-enabling load balancing with zayo. The only solution I see in that case is for tera is to move to a datacenter not controlled by centurylink.
Lots of people don't really understand peering, and that's reasonable! It's very complicated. A simplified way of thinking of it is that a peering location is like a freeway on-ramp. Whether the freeway is there or not doesn't matter if there's no way to get on it.
What's been happening with centurylink is that they don't peer properly with many other providers (their on-ramps are in bad places), leading to situations where a user in chicago connecting to a server in chicago may be routed through atlanta because that's the nearest peering node which will accept their connection. Indeed, if we look at a traceroute from cogent's chicago node to tera's servers, we see them being routed through atlanta at present. Not only does this add latency (at least 30ms in this case) but it also adds quite a few hops to the route, which leads to less stability - more opportunities for packet loss or latency spikes.
This isn't a problem that centurylink is intent on fixing, they have no reason to as it would cost them money and not get them more money. I also get the impression, since this has been problematic for months, that their datacenter also does not have any interest in re-enabling load balancing with zayo. The only solution I see in that case is for tera is to move to a datacenter not controlled by centurylink.
0
Comments
In any case, the real problem is: who are you going to choose that's going to be better? CenturyLink just bought Level 3, so that's another major backbone under their control, and it's not like any of the other U.S. providers are so much more willing to invest. If they change it might be better for you personally, but it could easily be worse for other people. So there's no single obvious answer.
(Another major MMO just recently moved their NA servers from one datacenter to another, and some people had their ping more than double as a result, and it became basically unplayable for part of their overseas playerbase due to horrible routing -- not to mention a much higher apparent occurrence of DDoS since the move. Of course, some others are neutral or slightly better too...)
https://tera-forums.enmasse.com/forums/general-discussion/topics/Australianand-others-latency-support-thread?page=2
https://tera-forums.enmasse.com/forums/general-discussion/topics/Latency-In-Tera-WHY
http://support.enmasse.com/answer_topics/ping-problem-where-is-the-server-hosted
(That first link even shows you a traceroute, and... it's still Savvis to Chicago even in 2012.)
Upgrading and stabilizing our backbone is not something anyone wants to do.