Doggydaddy wrote on Sep 12^th, 2014 at 4:42pm:



Server wide crippling lag when Dragon appears.

Random 24 people instances... with their own version of Greek Freeze crippling lag. (1) see below for what I think is the reason.
Some gets full pretty fast, others ends up being empty and it's faster to exit area ( at the risk of loosing the key ) and enter another one than wait for it to eventually fill.
as hinted in the line just above : keys not being given back if you exit before completion.

(1) reason of Dragon instance crippling lag :

The game is originally built around 6 people instances and a few 12 people instances. All of them private.  ( one party/raid = one instance )

The dragon in Mabar is an exception it's a 24 people semi-private instance ( one dragon instance = several parties with members that can be in yet another dragon instances.

Public instances ( say Marketplace ) are more or less limited to 50-70 people before they start spawing a second version of themselves. That's automatic and loadbalancing will put people entering the public instance in the less loaded one.

now I need to talk a bit at how the game client and the game server interract with each other.

When somebody does an action, the game client perform the action, and sends to the server the information that player X has performed said action. ( also sent to the server are the result of the action like the new location of the character )
The server then send that information to all the other game clients in the same instance.

Now we have what we all call crippling lag... actually it's a network/game client/Game server issue.
Because what happens when the dragon appears is that we have a massive monster, that change rooms every 2/3 seconds, and that does area of effect damages ( and take area of effect damages from 24 people ). All these, as well as all the location updates for 24 people, the Summons, the Pets, the Dragon, the Gargoyles.... The Action result for all that mess, as well as the hit ponits, spell points and more have to be sent back and forth between the Server and the 24 game clients... Then the game clients have to interpret all that  and display it... So what happens is that we just overload our links with Network stuff ( creating some delay ) overload a single thread ( representing the dragon instance ) on the game server, and overload the single thread/single process of the game client that needs to get all the effect/animations and such extracted out of the hard disk ( you know the optimized DAT files we all love. ). The result is crippling lag. Since the server is less in overload than the rest of the infrastructure the Dragon and the Gargoyles keep moving ( and hitting things ) while we all stand frozen by the network/game client overload.

Since the biggest chokepoint is the Dat files there's no real workaround... they can't do shit about it, so after trying so hard for so long to keep it alive they finally decided that it was going to be cheaper and faster to scrap the whole thing and create a new event. I won't mourn it's passing.

Terebinthia wrote on Sep 13^th, 2014 at 5:41am:


This.

One of the ways to avoid crippling lag in Mabar was to wait for the other instances to start and wait at least 2-3 minutes. Waiting for them to complete was even better. Then the last guy pulls the lever in his room and off we go. That resulted in more successes than fails. Organizing this with random puggers beforehand is the hardest part though. Some pugs are super retarded.

Many interesting theories in this thread.  However, I was in several instances of the dragon fight which were maxed with players and completed successfully with little performance issues.  And I remember the first mabar dragon raids a few years ago that as I recall ran pretty smooth (aside from the maban issue.)  But the game back then was 1/2 the size, or even smaller, than it is now, with a greater population.  Makes sense to me that there were some software bugs introduced (game wide) since mabar's inception, and possibly even as a result of the changes the devs made to the event itself since the beginning as well.

24 toon instances shouldn't be a problem.  They just schmegged something along the way, and they probably don't know what it was.  Maybe even schmegged a few things, because I think there are multiple issues involved (some that Flav pointed out.  And we continue to see some of these performance issues in other quests/raids to this day.

Gonna miss it, despite the shitty lag at time.
Wraps and cloak are not that horrid and 5 tome for few hours fun is cool. Also seen some people/guilds you never knew even existed on ze server.
Enjoyed it much more than other events.

Good riddance.

Invis on the altar always helped prevent a wipe.

Persiflage wrote on Sep 14^th, 2014 at 9:29am:

Lament....

Aeolwind wrote on Sep 14^th, 2014 at 9:41am:


Why?  Did a puppy die?

Only in the movie is it called the Lament Configuration, also "Lemarchand's box".  I am far more familiar with the Hellbound Heart, upon which the movie was based, and it's "Lemarchand's Configuration" in there.

We now return you to your normal progra...  Wait.  Thread derails ARE normal programming.

Aeolwind wrote on Sep 12^th, 2014 at 3:53pm:


Quote:

DropBear wrote on Sep 15^th, 2014 at 12:38am:


I suspect you'll fine a number of dead things up the dev's collective asses.

DropBear wrote on Sep 15^th, 2014 at 12:38am:

No need to worry there; the cockknuckles have Boots/Leap/Step timers to "fix".