|
Author |
Thread Statistics | Show CCP posts - 24 post(s) |
Jessica Danikov
Clan Shadow Wolf Fatal Ascension
141
|
Posted - 2013.12.03 17:04:00 -
[1] - Quote
Dev Blogs like this make me reconsider my current job... sounds like you guys have fun problems to work on. |
Jessica Danikov
Clan Shadow Wolf Fatal Ascension
141
|
Posted - 2013.12.03 17:44:00 -
[2] - Quote
Kale Freeman wrote:Given that you know that inter node jumps are the best because you split the work between two nodes, wouldn't it be better to work with node pairs. Apply the exact same algorithm to break down all the systems into equal size groups, but then right at the end take the group of systems and split them across a pair of nodes so as to create as many inter node jumps as possible.
That's starting to sound like a Color problem. Ultimately you want to maximise the number of local nodes without them touching themselves, while minimising the global distribution of any individual colour. This does the best to maximise the high-performance transitions while keeping TiDi relatively local to its cause. |
Jessica Danikov
Clan Shadow Wolf Fatal Ascension
141
|
Posted - 2013.12.04 02:20:00 -
[3] - Quote
GeeShizzle MacCloud wrote:Jessica Danikov wrote: That's starting to sound like a Colour problem. Ultimately you want to maximise the number of local nodes without them touching themselves, while minimising the global distribution of any individual colour. This does the best to maximise the high-performance transitions while keeping TiDi relatively local to its cause.
from what i can understand i think he means the opposite... cut up the load balancing down to the final level then split the final level across 2 nodes ensuring maximum contact surface between the 2 servers. instead of creating a line between the localised cluster of systems you create a chessboard style distribution to maximise the amount of server to server jumps rather than intra server jumps (that are considered inferior to discrete server to server jumps) you have the localisation at a cluster wide scale, then when u get to the singular node scale it switches to a high diffusion model. the problem would be how to keep the remapping from morphing the diffused local distribution from becoming localised through time.
Yes, that's the naive first solution I was thinking of, but it's a colour problem as 'checkboarding' the graph isn't possible with two colours- you will end up having white adjacent to white nodes (take three systems all connected to each other for the simplest example- White, White, Black). The whole idea of colouring is to minimise the numbers of colours needed, so you're coming up with the minimum number of involved nodes to ensure every edge is a change in node (for most cases, you will not likely need more than 3 colours/nodes).
If you colour the universe map then load balance each colour individually, you should end up with a similar load distribution but with every jump an inter-node jump. The best aspect of this is the colouring only has to be done once. |
Jessica Danikov
Clan Shadow Wolf Fatal Ascension
141
|
Posted - 2013.12.04 07:10:00 -
[4] - Quote
Being able to migrate solar systems live between nodes would be close to the holy grail of load balancing as you can harden a node incrementally by moving other nodes off it or move a system to a reinforced node when something like Asakai happens.
The other big leap would be the ability to exploit multiple cores/threads per solar system (doesn't help with load balancing so much as make the CPU ceiling and thus the load potential of an individual system much higher == more players in the same system together).
Both of these would be significant technical challenges, potentially a significant re-architecture in the application, so I wouldn't hold my breath for them happening any time in the foreseeable future. |
Jessica Danikov
Clan Shadow Wolf Fatal Ascension
144
|
Posted - 2013.12.07 13:20:00 -
[5] - Quote
Andy Koraka wrote:Maybe I'm misunderstanding something, but as far as I can tell this will only have a negative effect on the quality of game play in regards to already painful fleet combat.
Frankly I don't remember the last time I was in a full fleet and there wasn't heavy Ti-Di. Every time a solitary 250 man fleet jumps a gate the system spikes to 10% tidi for 30-45 seconds. Even if every fleet fight was on an individual reinforced node (reinforced nodes are the exception, not the rule) the issue of gate Tidi is going to be exponentially worse under the new regional scheme since every individual fleet in the area traveling to (or from) the combat system is going to be sequentially triggering gate lag on the same node. It's going to be a particularly painful change given the recent quality of life hits to the majority of fleet ships, there's nothing fun or engaging about staring at a warp tunnel for 10 minutes per system the entire trip home.
As far as the metagame is concerned, even without a published node map it's going to be exploited. For example in a defensive Sov war, if most of a region is on the same node it's not going to be hard to find a linked system by trial and error and dock/undock repeatedly to cascade the entire node (most of a region in the current scheme) into a sustained 10% tidi to discourage siege fleets from grinding structures.
Yes the old system wasn't perfect, but the guy ratting in an empty system halfway across EvE could have just moved over to a different system and continued ratting. Maybe this is the right solution for Empire where loads are usually steady from day to day but it's the wrong approach in Nullsec.
The changes made haven't done much to change this problem significantly- both systems create large areas of connected systems that are all on a single node, the new one just ignores constellation boundaries and balances the (predicted) load across nodes better, while also ensuring all solar systems on a node are fairly local to each other. At worst, it may make the contiguous spaces a little larger.
The static mapper could do a lot more for this issue by striping nodes if the difference between intra-node and inter-node jumps really is significant (especially when scaled up) and the efforts to do so should be fairly minimal. If not, the Brain in a Box is going to be the next big advance in that area. |
Jessica Danikov
Clan Shadow Wolf Fatal Ascension
150
|
Posted - 2013.12.12 19:20:00 -
[6] - Quote
CCP Prism X wrote:I do believe everything else has been covered already in previous posts.
So is there going to be any investigation into whether static striping of nodes will give better performance for large fleets moving through systems because of the inter- vs. intra- node thing? |
|
|
|