These forums have been archived and are now read-only.

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

Player Features and Ideas Discussion

 
  • Topic is locked indefinitely.
 

Idea: Auto logoff after 20-30 minutes of idle time
Author
Previous Topic Next Topic
mkint
#81 - 2017-06-01 13:21:31 UTC
Kassimila wrote:
Cade you keep talking about the individual nodes, when I'm talking about the cluster as a whole. There are two different types of "CPUs". Physical, and virtual. Without knowledge of the specific hardware/software configuration of the CCP cluster it's hard to discuss specifics. Hence my request for more information from CCP.

A single system 'node' can have MANY virtual/physical cpus assigned to it. That's in fact how they reinforce a node. There is also the fact that each solar system with a player in it gets 1 or a fraction of one CPU. So lets say hypothetically 300 systems have a single inactive player in them right now. Lets also say the minimum fraction of a cpu you can get is .1 . Kicking these players off the server would free up 30 physical CPUs that could then be assigned to active systems with players in them. So while these players won't show up in the 'server load' graphs you are linking, they are in fact stealing 30 cpus worth of power doing absolutely nothing.

The second part of CPU theft is idle players in Jita, even though they aren't 'doing' anything chat windows are still being sent to their clients. If they left their market window open, that has to be refreshed, etc. Is it a huge load? Probably not, but why not reclaim it for the people actually playing the game?

so much citation needed.

The CPU virtualization can only happen with the code written in C++. Not the python code that runs Dogma where ALL the lag and TiDi happen. You are so wrong in your understanding and your obtuse arguing, it hurts. You read a total of 2 performance devblogs. Out of the dozens where they explain the architecture, how different parts of the server interact with each other and how they are programmed, what causes load etc. Seach both old and current EVE forums for the posts on python. Then just abandon this thread. It's a bad idea and you're just pooping the bed.

Maxim 6. If violence wasn’t your last resort, you failed to resort to enough of it.
Kassimila
Deep Core Mining Inc.
Caldari State
#82 - 2017-06-01 23:28:08 UTC
mkint wrote:
Kassimila wrote:
Cade you keep talking about the individual nodes, when I'm talking about the cluster as a whole. There are two different types of "CPUs". Physical, and virtual. Without knowledge of the specific hardware/software configuration of the CCP cluster it's hard to discuss specifics. Hence my request for more information from CCP.

A single system 'node' can have MANY virtual/physical cpus assigned to it. That's in fact how they reinforce a node. There is also the fact that each solar system with a player in it gets 1 or a fraction of one CPU. So lets say hypothetically 300 systems have a single inactive player in them right now. Lets also say the minimum fraction of a cpu you can get is .1 . Kicking these players off the server would free up 30 physical CPUs that could then be assigned to active systems with players in them. So while these players won't show up in the 'server load' graphs you are linking, they are in fact stealing 30 cpus worth of power doing absolutely nothing.

The second part of CPU theft is idle players in Jita, even though they aren't 'doing' anything chat windows are still being sent to their clients. If they left their market window open, that has to be refreshed, etc. Is it a huge load? Probably not, but why not reclaim it for the people actually playing the game?

so much citation needed.

The CPU virtualization can only happen with the code written in C++. Not the python code that runs Dogma where ALL the lag and TiDi happen. You are so wrong in your understanding and your obtuse arguing, it hurts. You read a total of 2 performance devblogs. Out of the dozens where they explain the architecture, how different parts of the server interact with each other and how they are programmed, what causes load etc. Seach both old and current EVE forums for the posts on python. Then just abandon this thread. It's a bad idea and you're just pooping the bed.


Please stop, you have NO IDEA what you're talking about. ESXi is doing the virtualization. Any code, Java, Python, C++, etc can be run on servers using ESXi. Google it.
Cade Windstalker
#83 - 2017-06-02 03:01:07 UTC
Kassimila wrote:
mkint wrote:

so much citation needed.

The CPU virtualization can only happen with the code written in C++. Not the python code that runs Dogma where ALL the lag and TiDi happen. You are so wrong in your understanding and your obtuse arguing, it hurts. You read a total of 2 performance devblogs. Out of the dozens where they explain the architecture, how different parts of the server interact with each other and how they are programmed, what causes load etc. Seach both old and current EVE forums for the posts on python. Then just abandon this thread. It's a bad idea and you're just pooping the bed.


Please stop, you have NO IDEA what you're talking about. ESXi is doing the virtualization. Any code, Java, Python, C++, etc can be run on servers using ESXi. Google it.


mkint is still correct though, because one of the major limitations to Eve right now is that that Python stack is completely single threaded per node, and the underlying core architecture of the code means that one system is the minimum that it can be broken down into at the moment.

So having more virtual CPUs doesn't help, it's all about absolute processor speed, and having random AFK players scattered around has zero impact on that because an AFK player has zero impact on the physics and tasklet queues because *they're not doing anything*.

Again, go poke through some of the old videos and dev posts about Eve's architecture and you'll understand why AFK players are somewhere below "CPU intensiveness of AT ships exploding" on CCP's list of things that induce TiDi or lag of any kind.
mkint
#84 - 2017-06-02 03:23:24 UTC
Cade Windstalker wrote:
Kassimila wrote:
mkint wrote:

so much citation needed.

The CPU virtualization can only happen with the code written in C++. Not the python code that runs Dogma where ALL the lag and TiDi happen. You are so wrong in your understanding and your obtuse arguing, it hurts. You read a total of 2 performance devblogs. Out of the dozens where they explain the architecture, how different parts of the server interact with each other and how they are programmed, what causes load etc. Seach both old and current EVE forums for the posts on python. Then just abandon this thread. It's a bad idea and you're just pooping the bed.


Please stop, you have NO IDEA what you're talking about. ESXi is doing the virtualization. Any code, Java, Python, C++, etc can be run on servers using ESXi. Google it.


mkint is still correct though, because one of the major limitations to Eve right now is that that Python stack is completely single threaded per node, and the underlying core architecture of the code means that one system is the minimum that it can be broken down into at the moment.

So having more virtual CPUs doesn't help, it's all about absolute processor speed, and having random AFK players scattered around has zero impact on that because an AFK player has zero impact on the physics and tasklet queues because *they're not doing anything*.

Again, go poke through some of the old videos and dev posts about Eve's architecture and you'll understand why AFK players are somewhere below "CPU intensiveness of AT ships exploding" on CCP's list of things that induce TiDi or lag of any kind.

I just want to add *why* they made the deliberate decision to NOT use CPU virtualization: because the added overhead would hurt performance more than help. (Source: devblogs/devquotes)

Maxim 6. If violence wasn’t your last resort, you failed to resort to enough of it.
Bjorn Tyrson
Brutor Tribe
Minmatar Republic
#85 - 2017-06-02 06:14:51 UTC  |  Edited by: Bjorn Tyrson
Posting in stealth "nerf afk cloaking" thread #2179 of June 2017

Edit: original post was made in may so it's #9276539
perseus skye
Republic University
Minmatar Republic
#86 - 2017-06-02 12:03:04 UTC  |  Edited by: perseus skye
Kassimila wrote:
perseus skye wrote:
Personally a afk person is someone who can be found ,cloaked or not and killed where as someone logged off cannot be so it's a no from me .

Example a cloaky camper managed to decloak while afk and my T1 frigate got a 900m tenge kill plus pod


This thread is not about cloaky campers. Please stay on topic.


It wasn't a cloaky camper it was in fact a cloaky booster and it is on topic as this is exactly what you want to nerf 😆

Also no again
Desmios Sanguis
10th Crusade
#87 - 2017-06-04 12:07:43 UTC
bad idea.

I AFK in eve all the time and do not want to have go through the process of logging back in, especially while market trading.