Latency issues

[Raist]

In some cases, the last mile ISP's (your ISP in this case) may not be aware of the congestion further upstream. They should be aware if it is happening at the more localized exchange though. It happens often simply because the peering agreement is not set up with enough bandwidth to handle all the traffic they are trying to send through the exchange. Peering is often basically just an agreement that the companies using the exchange will share the costs of maintaining the exchange and such so they can all use it. There are a lot of "public" exchanges that have so many people coming through them that they simply get overloaded, and then there are the more "private" ones where money may actually exchange hands for access (which ideally would cut the congestion, but even then sometimes they don't purchase enough bandwidth). So, in many cases there actually is something an ISP can do to reduce the congestion--but it may mean spending money to do so, so it doesn't happen.

Your ISP peers with both companies. Which one you are assigned is determined by the policies in play, both for the peering agreement and for how routing is analyzed/implemented. They can manually create routing entries or they can use routing protocols that try to calculate/predict which routes may put forth the best effort (where the Autonomous System Numbers typically come into play--to share route details). Either system can run into problems, and simply get you stuck on a bad path. Things happen... bad weather, sudden shifts in usage patterns--sometimes the network just can't be adjusted fast enough and the pipes get clogged. I've looked up some of the calculated metrics between me and Montreal in the past and seen some routes report anywhere from 130-800+, but know some of them almost never run high enough to be pushed that far up the priority list when I've been on them. The values probably got assigned in a period when traffic was heavy, but then things evened out and those tables are no longer accurately reflecting the states of those routes. When it works right, it's amazing... but when it gets it wrong, it can be pretty bad.

Not sure exactly what Time Warner Cable eventually wound up doing for us, but they appear to have us locked into some specific paths through the Carolina's now. They may just be be putting entries directly into the routing tables through our area...who knows for sure. I've notice when my route changes it is pretty much the same groups of IP's that get recycled again and again. We bombarded them with a TON of data back at the 2.0 launch, and again last fall. This may have led to them adding entries to direct us through some specific gateways for getting to SE's network in Montreal. It's sort of an old-school way of routing. There are ways to over-ride the default routing with static paths in the routing tables. For instance, they could add entries in routers in Charlotte, NC to send all packets destined for SE's Canadian IP's (or Ormuco's network in Montreal) to always go through a specific router that would direct that traffic differently than all the other traffic going through Charlotte and up the East Coast. It can be some serious micro-managing of the network to do this, so getting it done can be tricky--but it is possible if you have people with the resources to implement the changes.

And that is the problem in some of these situations. The people you typically first talk to on the phone can't look into these options, much less make the changes. You have to get moved to higher levels of support/authority to get to someone that can properly investigate and recommend these types of changes. But yes, it should be possible to get them to assign alternate routing somehow...but you need the right people involved to make it happen. The industry standard for that level of support is Tier3, but some smaller companies may simply refer to them as Engineering or their Network Administrators. I've found I fair better getting to more advanced people when I use the online support options (email and such). Using phone support can get quite frustrating... constantly on hold, switching you back and forth to different people. Sometimes it seems like they do it on purpose in the hope you will give up trying.

If you do manage to get someone higher up that is actually interested in trying to do something about it, they would need to know the main block of addresses for Montreal, which is 199.91.189.xxx. There are other addresses we also hit periodically in Japan (mostly during log-in and initial loading of the game), but for the gameplay portion on the Canadian servers, they fall in that 199.91.189.1 /24 subnet.