America vs. Singapore (or Gateway vs. Instance)

I play from the USA, so naturally I chose the America gateway. In the last two weeks I have been noticing my ping can jump up +100ms (its usually stable between 40-50ms). It seems completely random.

However, yesterday I started noticing that my instances were sometimes located in Singapore. For me that explains the 100ms increase in latency... but I'm still left wondering why?

Are the gateways just load-balanced login servers, and you have a separate instance load-balancer that is misconfigured? I went from 90% of a level down to 10% before I called it quits today, and made note of my 150ms ping.

I run a pf based firewall/router so I can run sophisticated logging or testing if any Dev needs me to :) I would love to get back into the game but 150ms ping is worse than than the ISDN I had 15 years ago...

-DH

"

Thomas wrote:

If you're connecting to the America gateway then the only reason you should end up in a Singapore instance is if you had joined a party with someone who was on the Signapore gateway.

Has this been happening in solo play, or only when partying with people?

Also which IPs are you getting increased latency on? You should be able to find them in the <install dir>/logs/Client.txt log file, the lines look like this:

2012/09/14 16:30:59 4207752 56 [INFO Client 3544] Connecting to instance server at 173.193.53.194:43148

Thomas thank you for the fast response! I will check the Client.txt and report back the IPs the next time I see the spike.

Regarding solo play, I will make some more detailed observations since I'm playing a character that desyncs way less than my Shadow :)

-DH

Thomas, I just went to take a quick peak at the Client.txt and it was 32mb (I've been playing since July 27th). I assume it's safe to delete this file, and that it will rebuild itself on the fly the next time I run the client.

Based on my cursory evaluation, I would like to recommend that the client be updated with a UI element to toggle logging of the chat channels.

32mb is huge for a text file, and the vast majority of mine appears to be global chat. I will delete it before my next play session, and during that session I will try to keep chat to a minimum while looking for the "instance hopping" during solo.

I would also recommend that when PoE goes "live" that some official form of "instance hopping" prevention be implemented. It's extremely hard to diagnose as a player, when not all players are forthcoming about their IRL locations.

-DH

Thomas, so far I can't replicate the issue solo.

This morning I confirmed my huge lag was from joining a party. Almost died at the first group of monsters, checked F1 meters, then asked in group where everyone was. The party I joined had players from France and India; using the Singapore gateway.

I think this is leading to serious frustrations, and probably some bug reports that aren't actually bugs; just artifacts of poor global implementation.

I suggest quarantining instances to the gateways. Leave global chat global, but if you're going to allow global users to share instances, there should be a HUGE disclaimer about how terrible the play experience is going to be due to 100-200ms increase in latency.

As it is now, someone hits the Notice Board looking for a party, joins up, has no immediate indication they just jumped across the Earth and are about to have a terrible play session.

-DH

Thomas, I've been doing a little more testing and I wanted to let you know that it appears joining a party lead by a player on a different Gateway causes the town instances to jump overseas as well.

Played with some nice German fellows this morning, when I noticed the lag I asked where they were from... I said my goodbyes and sprung a portal back to town: the latency stayed with me even though I had left the party *before* entering my TP back to town.

Jumping waypoints, I figured out that towns have their own instances too, but you can't make your own. Luckily the servers sorted things out for me auto-magically so when I returned to town the latency was back down to ~40ms :)

-DH