These forums have been archived and are now read-only.

The new forums are live and can be found at https://forums.eveonline.com/

EVE General Discussion

 
  • Topic is locked indefinitely.
 

Can't log into Jita

First post First post
Author
Previous Topic Next Topic
Dr Slaughter
Rabies Inc.
#61 - 2012-04-29 08:14:08 UTC
Malcanis wrote:
Dr Slaughter wrote:
that's why I used quotes around the word 'instance'.



I could use quotes round the word 'unicorn' to describe my brother's cat; it would still be the wrong word. Instance has a specific meaning within the context of MMOs, as does zoning. Quote marks won't turn one into the other.

I liked your definition of zoning. Perhaps I have also made some sort of additional punctuation errors you can correct for me thus showing how clever you are on internet forums?
Alavaria Fera
GoonWaffe
#62 - 2012-04-29 08:25:17 UTC
CCP Explorer wrote:
CCP Goliath wrote:
Jita has a population cap in place to solve the very issue of server load. If you keep trying to log in you will eventually get there.
The current population cap on Jita (2200 pilots) and the "Jita-customised" Time Dilation parameters on the Jita node are intended to keep Jita playable. We are actively monitoring Jita and we are trying to keep it right on the edge with as high a population cap as possible.

Wow. Can you tell us anything about what makes Jita-customised TiDi different from fleet fight Tidi?

Does it help prevent those chaps from making it really hard to load the Jita 4-4 undock grid?

Triggered by: Wars of Sovless Agression, Bending the Knee, Twisting the Knife, Eating Sov Wheaties, Bombless Bombers, Fizzlesov, Interceptor Fleets, Running Away, GhostTime Vuln, Renters, Bombs, Bubbles ?
Francisco Bizzaro
#63 - 2012-04-29 08:34:20 UTC  |  Edited by: Francisco Bizzaro
Just curious, does the architecture allow them to put the Jita 4-4 station onto a separate cluster node?

The in-station and in-space games are disconnected enough that this would be a natural place to split the work, and allow for an effective doubling of the local population (maybe more for the in-station node, if it is less cpu hungry).

I've never heard of them doing this, so I assume it would require some significant re-engineering.
Malcanis
Vanishing Point.
The Initiative.
#64 - 2012-04-29 08:37:10 UTC
Dr Slaughter wrote:
Malcanis wrote:
Dr Slaughter wrote:
that's why I used quotes around the word 'instance'.



I could use quotes round the word 'unicorn' to describe my brother's cat; it would still be the wrong word. Instance has a specific meaning within the context of MMOs, as does zoning. Quote marks won't turn one into the other.

I liked your definition of zoning. Perhaps I have also made some sort of additional punctuation errors you can correct for me thus showing how clever you are on internet forums?


Sure. I'll get back to you after breakfast if that's OK.

"Just remember later that I warned against any change to jump ranges or fatigue. You earned whats coming."

Grath Telkin, 11.10.2016
Rico Minali
Federal Navy Academy
Gallente Federation
#65 - 2012-04-29 08:37:36 UTC
Col Arran wrote:
CFC are stopping me playing how I want to play, please ban them all!



Fixed your post for you, see? You didnt need all those words after all.

Trust me, I almost know what I'm doing.
CCP Explorer
C C P
C C P Alliance
#66 - 2012-04-29 09:23:11 UTC
Alavaria Fera wrote:
CCP Explorer wrote:
CCP Goliath wrote:
Jita has a population cap in place to solve the very issue of server load. If you keep trying to log in you will eventually get there.
The current population cap on Jita (2200 pilots) and the "Jita-customised" Time Dilation parameters on the Jita node are intended to keep Jita playable. We are actively monitoring Jita and we are trying to keep it right on the edge with as high a population cap as possible.

Wow. Can you tell us anything about what makes Jita-customised TiDi different from fleet fight Tidi?

Does it help prevent those chaps from making it really hard to load the Jita 4-4 undock grid?
Time Dilation was born from the need to better support large-scale fleet fights in 0.0. While we have had Time Dilation kick in in low-sec then this is the first time that Time Dilation kicks in for any sustained periods* in high-sec and frankly speaking this is a technical gold mine for us. Big smile

* Time Dilation is not always active in Jita right now, it only kicks in when there is significant activity. Most of the time right now then it is at 95-100% CPU and TiDi at 100% with TiDi bursting occasionally to 50-33%.

The CPU profile and general usage profile of Jita has always been very different from the rest of the universe. There are 4 large groups of profiles: 0.0 fights, mission hubs, market hubs, general solar system.

We configured TiDi to deal with 0.0 fights and those parameters mostly work for Jita but not quite. As a specific example then TiDi has to kick in very aggressively and very quickly during large fleet jump-ins. This use case doesn't apply to Jita but rather we have to deal with Crime Watch monitoring (which also applies to low-sec but not 0.0) and CONCORD response (which doesn't apply to low-sec or 0.0) for short-burst smaller engagements (but potentially multiple such activities). Different code and different usage profile and so we are having TiDi delay the activation until it's clear that the engagement is going to be sustained (e.g., parties in high-sec war in Jita) or that there are multiple suicide ganks on-going at the same time.

Erlendur S. Thorsteinsson | Senior Development Director | EVE Online // CCP Games | @CCP_Explorer
CCP Explorer
C C P
C C P Alliance
#67 - 2012-04-29 09:27:50 UTC
Francisco Bizzaro wrote:
Just curious, does the architecture allow them to put the Jita 4-4 station onto a separate cluster node?

The in-station and in-space games are disconnected enough that this would be a natural place to split the work, and allow for an effective doubling of the local population (maybe more for the in-station node, if it is less cpu hungry).

I've never heard of them doing this, so I assume it would require some significant re-engineering.
From an architectural standpoint then it's possible to split stations from solar systems. This is how it was in the beginning of EVE but over time dependencies have been added and it's a fair amount of work to split them again. As a part of multi-player Incarna then we were working on splitting them up again.

The CPU profile differs between different types of systems. In most systems then stations comparatively consume almost no CPU with CPU cycles being spent on in-space simulation and damage tracking, whereas Jita is inventory-operation heavy and the stations (4-4 in particular) consume a lot of CPU.

Erlendur S. Thorsteinsson | Senior Development Director | EVE Online // CCP Games | @CCP_Explorer
Mallak Azaria
Caldari Provisions
Caldari State
#68 - 2012-04-29 09:47:06 UTC
Yet another butthurt OP.

Your tears would be ruining my forums immersion, if they weren't so tasty.

This post was lovingly crafted by a member of the Goonwaffe Posting Cabal, proud member of the popular gay hookup site somethingawful.com, Spelling Bee, Grammar Gestapo & #1 Official Gevlon Goblin Fanclub member.
RDevz
State War Academy
Caldari State
#69 - 2012-04-29 10:07:21 UTC
Col Arran wrote:

1. You may not take any action that imposes an unreasonable or disproportionately large load on the System.

Stated here plain and clear if the system is unable to accept new connections then I would say that is an unreasonable or disproportionately large load.


I'm in Jita. Occasionally I shoot something. I can't help that one hundred of my closest friends and associates have also decided to be in Jita, occasionally shooting something. I especially can't help that hundreds of vultures, alts, and associated hangers-on that are completely unaffiliated to my alliance have decided to follow me around and try to loot my wrecks and whore on my killmails.

~
Dr Slaughter
Rabies Inc.
#70 - 2012-04-29 12:38:53 UTC
CCP Explorer wrote:
Francisco Bizzaro wrote:
Just curious, does the architecture allow them to put the Jita 4-4 station onto a separate cluster node?

The in-station and in-space games are disconnected enough that this would be a natural place to split the work, and allow for an effective doubling of the local population (maybe more for the in-station node, if it is less cpu hungry).

I've never heard of them doing this, so I assume it would require some significant re-engineering.
From an architectural standpoint then it's possible to split stations from solar systems. This is how it was in the beginning of EVE but over time dependencies have been added and it's a fair amount of work to split them again. As a part of multi-player Incarna then we were working on splitting them up again.

The CPU profile differs between different types of systems. In most systems then stations comparatively consume almost no CPU with CPU cycles being spent on in-space simulation and damage tracking, whereas Jita is inventory-operation heavy and the stations (4-4 in particular) consume a lot of CPU.

I remember there was a time when most of the SOL services had been implemented in a monolithic way and you worked hard splitting them up (with some successes such as market etc.). That was back when the GIL was causing you scalability problems and the x86 architecture had run out of an ever increasing number of Ghz.

There was talk about making each SOL service distributable and using Infiniband RDMA to keep everything in sync. What happened to the HPC cell and this work?
CCP Explorer
C C P
C C P Alliance
#71 - 2012-04-29 14:25:58 UTC
Dr Slaughter wrote:
CCP Explorer wrote:
Francisco Bizzaro wrote:
Just curious, does the architecture allow them to put the Jita 4-4 station onto a separate cluster node?

The in-station and in-space games are disconnected enough that this would be a natural place to split the work, and allow for an effective doubling of the local population (maybe more for the in-station node, if it is less cpu hungry).

I've never heard of them doing this, so I assume it would require some significant re-engineering.
From an architectural standpoint then it's possible to split stations from solar systems. This is how it was in the beginning of EVE but over time dependencies have been added and it's a fair amount of work to split them again. As a part of multi-player Incarna then we were working on splitting them up again.

The CPU profile differs between different types of systems. In most systems then stations comparatively consume almost no CPU with CPU cycles being spent on in-space simulation and damage tracking, whereas Jita is inventory-operation heavy and the stations (4-4 in particular) consume a lot of CPU.
I remember there was a time when most of the SOL services had been implemented in a monolithic way and you worked hard splitting them up (with some successes such as market etc.). That was back when the GIL was causing you scalability problems and the x86 architecture had run out of an ever increasing number of Ghz.

There was talk about making each SOL service distributable and using Infiniband RDMA to keep everything in sync. What happened to the HPC cell and this work?
We eventually went a different route since that technology didn't fit what we needed.

We have in the past few years moved services off the location nodes that don't pertain to the solarsystems. Early examples of that were the market and corp/alliance services. We implemented new things such as the EVE Voice service, EVE Gate Service, Fleet Finder, Bulk Data Delivery Service, Incursions Service and Planetary Interaction Manager in this manner and then moved all character-related services to different nodes, services such as character lookup, mail, calendar, contracts, etc.

We then replaced the network layer with a new (multi-threaded) layer in C++ (called CarbonIO), which then avoids the GIL. We used in features in CarbonIO for synchronising Time Dilation timestamps between each node and the clients on that node. Using CarbonIO we also made improvements to how the space simulation sends update packets to all clients.

With significant simplification then the location nodes now handle the space simulation (Destiny), damage tracking (Dogma), aggressions (Crime Watch) and inventory operations (Inventory).

Erlendur S. Thorsteinsson | Senior Development Director | EVE Online // CCP Games | @CCP_Explorer
Audix
Ministry of War
Amarr Empire
#72 - 2012-04-29 14:57:32 UTC
Col Arran wrote:
This is unacceptable, I am not able to log into the Jita system because it is unable to accept new connections. This is 100% unacceptable and is affecting my experience.

I wanted to log into Jita to get some trading done in these times of trouble but guess what? I can't. I expect a resolution immediately, CCP had more than enough time to sufficiently reinforce the Jita node and they did not. Also I do not care if it is merely a political statement, for the lulz or whatever.


6. CONDUCT

A. Specifically Restricted Conduct

Your continued access to the System and license to play the Game is subject to proper conduct. Without limiting CCP's rights to control the Game environment, and the conduct of the players within that environment, CCP prohibits the following practices that CCP has determined detract from the overall user experience of the users playing the Game.

1. You may not take any action that imposes an unreasonable or disproportionately large load on the System.


Stated here plain and clear if the system is unable to accept new connections then I would say that is an unreasonable or disproportionately large load.


QQ o7o7o7o7o7oo7o7o7o7oo7o7o7 m8m8m8m8m8mm8m8m8m8mm8m8
Lapine Davion
Outer Ring Applied Logistics
#73 - 2012-04-30 04:10:47 UTC
Sigurd Sig Hansen wrote:


http://s144.photobucket.com/albums/r188/buddahcjcc/?action=view&current=JitaBurns.jpg

that is why lol


Hey man, you're down with the clown, aren't you?

[b]Don't worry about posting with your main!  Post with your brain! "Never attribute to malice that which is adequately explained by stupidity."[/b]
Son IamaDerp
Federal Navy Academy
Gallente Federation
#74 - 2012-04-30 04:20:01 UTC  |  Edited by: Son IamaDerp
Lapine Davion wrote:
Sigurd Sig Hansen wrote:


http://s144.photobucket.com/albums/r188/buddahcjcc/?action=view&current=JitaBurns.jpg

that is why lol


Hey man, you're down with the clown, aren't you?
You're insulting clowns. He's a Juggalo, whoopwhoopmagnets.
Dr Slaughter
Rabies Inc.
#75 - 2012-04-30 08:52:28 UTC
CCP Explorer wrote:
We eventually went a different route since that technology didn't fit what we needed.

We have in the past few years moved services off the location nodes that don't pertain to the solarsystems. Early examples of that were the market and corp/alliance services. We implemented new things such as the EVE Voice service, EVE Gate Service, Fleet Finder, Bulk Data Delivery Service, Incursions Service and Planetary Interaction Manager in this manner and then moved all character-related services to different nodes, services such as character lookup, mail, calendar, contracts, etc.

We then replaced the network layer with a new (multi-threaded) layer in C++ (called CarbonIO), which then avoids the GIL. We used in features in CarbonIO for synchronising Time Dilation timestamps between each node and the clients on that node. Using CarbonIO we also made improvements to how the space simulation sends update packets to all clients.

With significant simplification then the location nodes now handle the space simulation (Destiny), damage tracking (Dogma), aggressions (Crime Watch) and inventory operations (Inventory).

Thanks for the catch-up Explorer! While you're showing us what's under the hood I'm going to ask a couple more questions :)

Has all of this impacted the proxies at all? I assume you simply limit the number of clients that can be connected to each proxy node rather than operate them in a NLBS fashion?

What's CCP Warlock up to now?
CCP Explorer
C C P
C C P Alliance
#76 - 2012-04-30 16:59:34 UTC  |  Edited by: CCP Explorer
Dr Slaughter wrote:
CCP Explorer wrote:
We eventually went a different route since that technology didn't fit what we needed.

We have in the past few years moved services off the location nodes that don't pertain to the solarsystems. Early examples of that were the market and corp/alliance services. We implemented new things such as the EVE Voice service, EVE Gate Service, Fleet Finder, Bulk Data Delivery Service, Incursions Service and Planetary Interaction Manager in this manner and then moved all character-related services to different nodes, services such as character lookup, mail, calendar, contracts, etc.

We then replaced the network layer with a new (multi-threaded) layer in C++ (called CarbonIO), which then avoids the GIL. We used in features in CarbonIO for synchronising Time Dilation timestamps between each node and the clients on that node. Using CarbonIO we also made improvements to how the space simulation sends update packets to all clients.

With significant simplification then the location nodes now handle the space simulation (Destiny), damage tracking (Dogma), aggressions (Crime Watch) and inventory operations (Inventory).
Has all of this impacted the proxies at all? I assume you simply limit the number of clients that can be connected to each proxy node rather than operate them in a NLBS fashion?
These changes reduced the CPU load on the proxies but also changed the profile to be slightly non-linear. The overall gain was that at 100% CPU load then approx. 5% more clients could be connected to each proxy but at the target 80% CPU load then approx. 25% more clients could be connected. We then bought better hardware for the proxies and eventually reduced the number of proxies significantly, from 24-27 proxies (on 11-12 blades) to 16 proxies (on 4 blades) today. The load balancer operates on "fewest number of connections" and there is a safe-guard limit on each proxy.

Erlendur S. Thorsteinsson | Senior Development Director | EVE Online // CCP Games | @CCP_Explorer