Pages: [1] :: one page |
|
Author |
Thread Statistics | Show CCP posts - 2 post(s) |
monkfish1234
Caldari Attempted Ninjury 0ccupational Hazzard
|
Posted - 2011.02.16 13:10:00 -
[1]
I'm quite intruiged by the process of reinforcing nodes currently in use by CCP. Currently it is only possible during downtime move an area to one of the beefed up nodes in preparation for a large known fight.
It leaves me to wonder what exactly needs to be done to make this happen, and why a reinforced node cannot be prepared, run in parralelle then switched to whilst the server is live?
I am presuming that the majority of the issue would be with moving all of the active sessions and all of the db information they bring, but I am surprised that some sort of session changing process could not be implemented, similar to a mass gate activation. I'm aware this would probably mean a stasis period where modules could not be activated both before and after the switch.
I guess with the current state of fleet combat this is something ccp (i presume gridlock) should be looking at if they arent already. Uemon was a very good example of how things can escalate quickly, A form sent 24hrs in advance just isn't always practical.
I'd really like to see a dev blog about the process as it is (the drakes of destiny and other lag blogs have been very interesting) ------------------------------------------------
Please resize your signature to the maximum allowed of 400 x 120 pixels.StevieSG
|
Hirana Yoshida
Behavioral Affront
|
Posted - 2011.02.16 13:31:00 -
[2]
The plan is to allow for impromptu reinforcements at some point, but I don't think they have the hard-/software in place yet .. think it mentioned in one of the "TQ Level UP!" threads a while back.
Low-sec cannot be reinforced, period. They share nodes among themselves and with high-sec. There is some technical mumbo-jumbo involved that was also mentioned previously.
The problem that arose in Uemon had nothing to do with node not being reinforced and everything to do with the care-sec citizens throwing 130 SC's and Goddess knows how many carriers into one system. Blobs cause lag, blobs using drones cause more lag. This is the way its been since forever so it can hardly come as a surprise to anyone.
|
Amaroq Dricaldari
Amarr Universal Deathdealing Militia Fusion Alliance
|
Posted - 2011.02.16 13:36:00 -
[3]
How aobut this? Low Sec systems and High-Traffic areas will get a sub-server connected to the main server. They are still run by the main server, but the sub server for that area improves performance in that area, wether it is a High-Traffic area like Jita, or a Low-Sec region with alot of alliance battles. -- As an Amarr Defector, I chose to become a Mercenary and Industrialist. I also have one goal in mind: Create a new age of peace and prosperity for all four empires. |
monkfish1234
Caldari Attempted Ninjury 0ccupational Hazzard
|
Posted - 2011.02.16 13:56:00 -
[4]
Ahh i must have missed where they mentioned that.
I'm aware currently a specific non-null system cannot just be picked out from one node and moved to another. I was using Uemon as an example of how quickly an engagement can escalate.
I guess another idealistic situation would be for some description of sub-nodes that cover a system or a smaller group that can be detached from it's node and moved without breaking everything. Again this very much depends on the current structuring of the nodes what each systems dependencies are.
As I think was already mentioned in another thread lowsec is very much differant to null as there are things such as criminal timers, standing etc etc to take into account. when you add that ontop of the load produced by a battle of that size bad things happen.
EVE has obviously evolved a long way since alot of the core code was dreamt up and bit by bit things are changing. I figure it's only a matter of time before systems etc become more modular and load balancing can be managed better, CCP have already show they are inovators in these fields.
I would love to see more dev blogs about the technical ins and outs of EVE ------------------------------------------------
Please resize your signature to the maximum allowed of 400 x 120 pixels.StevieSG
|
monkfish1234
Caldari Attempted Ninjury 0ccupational Hazzard
|
Posted - 2011.02.16 14:12:00 -
[5]
Originally by: Amaroq Dricaldari How aobut this? Low Sec systems and High-Traffic areas will get a sub-server connected to the main server. They are still run by the main server, but the sub server for that area improves performance in that area, wether it is a High-Traffic area like Jita, or a Low-Sec region with alot of alliance battles.
That is essentially how it is run now, a node will only handle a certain number of systems. ( i presume this is based on average trafic as each system on a node arent always connected/near each other afaik).
The problem comes when a 'sub-server' gets far more trafic than normal. In these cases there needs to be a way to move what is currently on that node onto a more powerful server. This can;t really be done Live becasue there are so many database transactions etc happeneing per second that by the time the db was copied to another server the data would be well out of date.
Becasue of this, they need a way to freeze the inspace activity (chat etc works differantly I think). then it can be ported and restored on a stronger node and the lag moster is defeated. The major issues are things like the time it would take to perform this, which could be quite long if there are 1000+ in system (even before fighting) and actually making the switch whithout needed to kick every one in the node.(mass loggin back in causes pretty high load in itself) ------------------------------------------------
Please resize your signature to the maximum allowed of 400 x 120 pixels.StevieSG
|
|
CCP Veritas
|
Posted - 2011.02.16 15:30:00 -
[6]
Hi there guys.
A few corrections before I get into the meat:
- There's nothing special about low/high sec when it comes to node reinforcment. Jita is reinforced constantly, for instance.
- We have the ability (and use it fairly often) to take a loaded solar system off of a shared node and put it by itself at run-time. It's a messy process and sometimes fails catastrophically, so there has to be very good reason to do so.
- Uemon was already on its own node for that fight
- Uemon was not overloaded because of drones, at least not mostly. We've isolated a bulk of the load to systems that are not present in nullsec.
The reason remapping is so messy is, unfortunately, quite simple - there's a mountain of code that was not designed with that ability in mind. A team over in Core made solid progress toward cleaning it up last fall, but after a few months of effort they've had to put it on the back burner for other tasks.
|
|
monkfish1234
Caldari Attempted Ninjury 0ccupational Hazzard
|
Posted - 2011.02.16 15:54:00 -
[7]
Originally by: CCP Veritas Hi there guys.
A few corrections before I get into the meat:
- There's nothing special about low/high sec when it comes to node reinforcment. Jita is reinforced constantly, for instance.
- We have the ability (and use it fairly often) to take a loaded solar system off of a shared node and put it by itself at run-time. It's a messy process and sometimes fails catastrophically, so there has to be very good reason to do so.
- Uemon was already on its own node for that fight
- Uemon was not overloaded because of drones, at least not mostly. We've isolated a bulk of the load to systems that are not present in nullsec.
The reason remapping is so messy is, unfortunately, quite simple - there's a mountain of code that was not designed with that ability in mind. A team over in Core made solid progress toward cleaning it up last fall, but after a few months of effort they've had to put it on the back burner for other tasks.
Thanks for the response,
I guess the main battle is improving the process of separating a system from a shared node. if that was a process less prone to catastrophic failiure then looking forward to a time when it could be done live would be more realistic. (altho i'd hate to be the one responsible for transporting the 1000's of sessions worth of data to another server)
If you ever get a chance it would be really cool if you could poke someone over at core to put out a dev blog about the mysteries of reinforcment.
I presume there will be one at a later date regarding what happened in uemon as i'm aware that is a separate issue. ------------------------------------------------
Please resize your signature to the maximum allowed of 400 x 120 pixels.StevieSG
|
Jokerface666
Amarr
|
Posted - 2011.02.16 15:58:00 -
[8]
Originally by: monkfish1234
Originally by: CCP Veritas Hi there guys.
A few corrections before I get into the meat:
- There's nothing special about low/high sec when it comes to node reinforcment. Jita is reinforced constantly, for instance.
- We have the ability (and use it fairly often) to take a loaded solar system off of a shared node and put it by itself at run-time. It's a messy process and sometimes fails catastrophically, so there has to be very good reason to do so.
- Uemon was already on its own node for that fight
- Uemon was not overloaded because of drones, at least not mostly. We've isolated a bulk of the load to systems that are not present in nullsec.
The reason remapping is so messy is, unfortunately, quite simple - there's a mountain of code that was not designed with that ability in mind. A team over in Core made solid progress toward cleaning it up last fall, but after a few months of effort they've had to put it on the back burner for other tasks.
Thanks for the response,
I guess the main battle is improving the process of separating a system from a shared node. if that was a process less prone to catastrophic failiure then looking forward to a time when it could be done live would be more realistic. (altho i'd hate to be the one responsible for transporting the 1000's of sessions worth of data to another server)
If you ever get a chance it would be really cool if you could poke someone over at core to put out a dev blog about the mysteries of reinforcment.
I presume there will be one at a later date regarding what happened in uemon as i'm aware that is a separate issue.
Yeah a devblog about that would be awesome. /emote wants to know more bout it! I'm looking forward for a automated reinforcement system, where the hamsters do not have to run around with usb sticks with light speed.
BR w00t w00t wtfpwnage train |
|
CCP QC
|
Posted - 2011.02.16 16:24:00 -
[9]
Originally by: monkfish1234
If you ever get a chance it would be really cool if you could poke someone over at core to put out a dev blog about the mysteries of reinforcment.
I consider myself poked =)
As CCP Veritas said, separating a live solar system from its host server is a messy process. While I dream of the day where I can freeze all of you for a couple of minutes and transport the system to a dedicated machine, an enormous amount of work remains to be done for this to become reality.
|
|
monkfish1234
Caldari Attempted Ninjury 0ccupational Hazzard
|
Posted - 2011.02.16 16:56:00 -
[10]
Originally by: CCP QC
Originally by: monkfish1234
If you ever get a chance it would be really cool if you could poke someone over at core to put out a dev blog about the mysteries of reinforcment.
I consider myself poked =)
As CCP Veritas said, separating a live solar system from its host server is a messy process. While I dream of the day where I can freeze all of you for a couple of minutes and transport the system to a dedicated machine, an enormous amount of work remains to be done for this to become reality.
hopefully we'll get there one day :)
Just wish I was good enough to be able to come help realise the dream.
Thanks again for the response, also I don't think you guys get anywhere near the recognition you deserve for alot of what goes on within eve.
------------------------------------------------
Please resize your signature to the maximum allowed of 400 x 120 pixels.StevieSG
|
|
Amaroq Dricaldari
Amarr Universal Deathdealing Militia Fusion Alliance
|
Posted - 2011.02.16 17:30:00 -
[11]
Originally by: CCP QC As CCP Veritas said, separating a live solar system from its host server is a messy process. While I dream of the day where I can freeze all of you for a couple of minutes and transport the system to a dedicated machine, an enormous amount of work remains to be done for this to become reality.
All you really need to do is have the nodes reinforced with a bunch of other nodes and have the High-Traffic systems such as Jita transferred to an area on the Main server connected to a special sub-server during downtime.
Also, I don't know exactly where to put it on the forums, but I have some Technological Ideas to improve Server Performance (and efficiency). -- As an Amarr Defector, I chose to become a Mercenary and Industrialist. I also have one goal in mind: Create a new age of peace and prosperity for all four empires. |
Hirana Yoshida
Behavioral Affront
|
Posted - 2011.02.16 18:51:00 -
[12]
Originally by: CCP QC ...
What's stopping you? We live in a world where people take sick days to wait for a plumber and idle for hours listening to crappy music to get <insert name> customer service.
Surely the blobbier parts of eve would love to have a big "Please Hold" sign pop-up, if it means that the subsequent battle does not cause ulcers to burst and facial indentation to form in the furniture.
|
AnonyTerrorNinja
Minmatar Buggers' Advanced Interstellar Transport
|
Posted - 2011.02.16 22:08:00 -
[13]
It wouldn't affect only the blobbers, it would affect anyone and everyone that happens to be on that same node that's being shared by x systems when y system needs to get moved to its own node.
At worst, it could even mean the entire server needs to get 'frozen' for the fight, since we don't know exactly what the extent of freizure is --- Sig removed till it's relevant again. |
Hirana Yoshida
Behavioral Affront
|
Posted - 2011.02.16 22:57:00 -
[14]
Originally by: AnonyTerrorNinja It wouldn't affect only the blobbers...
Would require a redistribution of the nodes/systems before hand obviously.
For low you could probably slap entire regions (or more) onto individual nodes, decoupling them from high is a given. Population in low is not that high so people affected by the 'Please Hold' would be minimal.
Having been quite a few of the "time for a cuppa in between cycles" fights I can safely say that any measure that can mitigate it is worth almost any sacrifice .. if a loaded system is not detached then everyone on node can suffer so I doubt anyone will object to an "unscheduled bio-break". But as the CCP guy said, bashes of that magnitude are pretty rare in low to begin with so chances are nothing will be done at all.
|
prospector oen
|
Posted - 2011.02.17 02:00:00 -
[15]
i think most are upset because 2 large 00 alliances tried blobbing titans and whined when they got blown up, ccp still hasnt canned anti blobism. This could be solved in part by wormholes 4-8 hour timeframes apwning more often directly form high and low sec and other 00 into 00 and low sec.
Resistant atm they is
|
CO Bradley
|
Posted - 2011.02.17 03:02:00 -
[16]
Originally by: CCP Veritas Hi there guys.
A few corrections before I get into the meat:
- There's nothing special about low/high sec when it comes to node reinforcment. Jita is reinforced constantly, for instance.
- We have the ability (and use it fairly often) to take a loaded solar system off of a shared node and put it by itself at run-time. It's a messy process and sometimes fails catastrophically, so there has to be very good reason to do so.
- Uemon was already on its own node for that fight
- Uemon was not overloaded because of drones, at least not mostly. We've isolated a bulk of the load to systems that are not present in nullsec.
The reason remapping is so messy is, unfortunately, quite simple - there's a mountain of code that was not designed with that ability in mind. A team over in Core made solid progress toward cleaning it up last fall, but after a few months of effort they've had to put it on the back burner for other tasks.
Since I'm not on payroll I will have to guess at this...but is there a way to offload the sec status processes to it's own node. If that is possible if could reduce the load caused by the massive wave of sec modifiers that occurred all at once. I don't know what all is tied to the security status so if I am off in left field I apologize.
|
monkfish1234
Caldari Attempted Ninjury 0ccupational Hazzard
|
Posted - 2011.02.17 09:30:00 -
[17]
This isnt really so much aobut low sec battles. (they have their own special issues that need to be refined).
There is need to reinforce 0.0 systems on a regular basis esecially when drake armys come into town.
tbh for now i'd rather see CCP ingnor what happened in Uemon as it was a rare event, and deal with 0.0 which would atleast be slightly easier to do this kind of thing with. ------------------------------------------------
Please resize your signature to the maximum allowed of 400 x 120 pixels.StevieSG
|
|
|
|
Pages: [1] :: one page |
First page | Previous page | Next page | Last page |