Pages: [1] :: one page |
|
Author |
Thread Statistics | Show CCP posts - 0 post(s) |
Sapporo Jones
|
Posted - 2010.09.04 18:30:00 -
[1]
I'm sorry. I have only ever posted negative feedback to the devs ONCE and that was for the AWFUL 'linux client'. But I come to you now with a plea. Stop, please stop, asking for 24 hour notices before massive fleet engagements. You have the ability to see when a system in a constellation is spiking with activity, why not automate the load balancing when that happens? There should be no reason that in 2010 you need 24 hours notice to support a system for a fleet op. Especially since we can't always tell you what system an op will be happening in if the OpFor moves around a lot. Just setup some script or cron that watches for local spikes and balances server load. It will greatly improve gameplay and prevent all this backlash.
Last time our alliance had an op we had 700 people in a system fully engaged with the enemy and it was a slideshow. Later a plea from the devs came saying that other systems in the constellation were affected and this could have been resolved had we provided you with 24 hours notice to beef out the constellation. Many in our corp and alliance have multiple accounts (myself included). Were you only getting one sub fee per person I could understand the financial involvement this would take seeming insurmountable. However, with the intake you have, you have resources to invest in some simple scripting to prevent this in the future. If there is something I don't understand about this please respond but I don't wanna hear about this 24 hour thing again. Its ridiculous that a fleet has to plan with you when to have an engagement with the enemy.
|
Mike Tug'gah
|
Posted - 2010.09.04 18:34:00 -
[2]
wrong forum dude
|
Herschel Yamamoto
Agent-Orange Nabaal Syndicate
|
Posted - 2010.09.04 18:37:00 -
[3]
They can only replace a node in the middle of a fight by killing it first. So if you want everyone to be booted off the server in the middle of a fleet fight, then you can get your reinforced node. Or you can ask a day in advance and get them to do it on downtime. They're working on this constraint, but it hasn't been fixed yet. === "The data does not support that polished quality sells better than new features" "Once Incarna and Dust are fully implemented, focus will probably shift far more towards improvement" CCP, FTW? |
Sapporo Jones
|
Posted - 2010.09.04 18:43:00 -
[4]
OK. If it is getting worked on I can deal with that, it just seems ludicrous that their concept of load balancing involves killing an entire node. In my RL experience in load balancing it can happen without killing a server. That sounds more like a setup issue or topology issue more than a constraint in their system. Hell, I would even call that a bug of showstopper proportions.
|
Tres Farmer
Gallente Federation Intelligence Service
|
Posted - 2010.09.04 19:06:00 -
[5]
Edited by: Tres Farmer on 04/09/2010 19:08:27
Originally by: Sapporo Jones OK. If it is getting worked on I can deal with that, it just seems ludicrous that their concept of load balancing involves killing an entire node. In my RL experience in load balancing it can happen without killing a server. That sounds more like a setup issue or topology issue more than a constraint in their system. Hell, I would even call that a bug of showstopper proportions.
I suggest you start reading the latest Dev Blogs on fighting lag, especially the one from CCP Warlock The Long Lag and CCP Atlas Fixing Lag: Character Nodes. The comment-threads linked there are also a source of information about whats going on.
|
Akita T
Caldari Caldari Navy Volunteer Task Force
|
Posted - 2010.09.04 19:14:00 -
[6]
Originally by: Sapporo Jones You have the ability to see when a system in a constellation is spiking with activity, why not automate the load balancing when that happens?
Because the current server-side coding REQUIRES a cluster downtime in order to move solar systems around on different physical nodes. They're allegedly "working on it" (the dynamic load balancing). Have been, allegedly, for the past year the very least, if not two years. They're not there yet. _
Beginner's ISK making guide | Manufacturer's helper | All about reacting _
|
Alara IonStorm
Agent-Orange
|
Posted - 2010.09.04 19:17:00 -
[7]
Originally by: Herschel Yamamoto They can only replace a node in the middle of a fight by killing it first. So if you want everyone to be booted off the server in the middle of a fleet fight, then you can get your reinforced node. Or you can ask a day in advance and get them to do it on downtime. They're working on this constraint, but it hasn't been fixed yet.
General Discussion is my territory!
Get your A$$ back to CAOD!
Sir.
-- Tactical Responder who is Organized and a Leading-edge Linguist |
Stick Cult
|
Posted - 2010.09.04 20:48:00 -
[8]
Originally by: Akita T
Originally by: Sapporo Jones You have the ability to see when a system in a constellation is spiking with activity, why not automate the load balancing when that happens?
Because the current server-side coding REQUIRES a cluster downtime in order to move solar systems around on different physical nodes. They're allegedly "working on it" (the dynamic load balancing). Have been, allegedly, for the past year the very least, if not two years. They're not there yet.
And Warlock said it "Will take a long time." Cmon Warlock, weren't you supposed to be the smart one?!
Anyway, yea, they can't just move *this* system to a new node whenever they want to, without kicking everyone off it first, so they have to do it at downtime.
Originally by: CCP Tuxford my bad. Rest assured I'm being ridiculed by my co-workers.
|
Bill Ny
Gallente
|
Posted - 2010.09.04 20:52:00 -
[9]
fact of the matter is, when they first coded EVE, the software architecture was not designed with dynamic load balancing in mind. The Devs didn't even consider it some odd 7-8 years ago.
In fact, they've built so much on top of a restricted foundation that to replace the foundation to allow dynamic load balancing would require EXTENSIVE rewriting of lower level code in the game system.
Laymans terms: You can build a house on a good enough foundation, but when you keep adding to the house, the foundation wont be enough, and by the time you finish putting in the fifth floor expansion, you realize that you need to redo the foundation since it's not good enough for the size of the house now.
Only time and great patience with the Devs will solve this issue. You gotta give them some credit, a code re factoring of that magnitude is astronomically hard to pull off. I am also the Science Guy.
I am not inclined to tell you why ice "cracks." I am entitled to enjoy my goddamn meal. |
Spurty
Caldari D00M. Northern Coalition.
|
Posted - 2010.09.04 22:15:00 -
[10]
In real life I recently virtualized a quality lab for a software company.
Enterprise level tools allow me to move load around with a Max of 5 second pause from a loaded server to an unloaded one.
This technology can do this today for any Intel based operating system.
Its sick how easy it is.
Perhaps worth sending a CCP dude to vm ware office for a demo?
There's one in Cambridge mass. Not far from where I work. Would buy you a beer :+)
|
|
Aessoroz
|
Posted - 2010.09.04 22:44:00 -
[11]
Edited by: Aessoroz on 04/09/2010 22:45:19 Edited by: Aessoroz on 04/09/2010 22:44:24 Edited by: Aessoroz on 04/09/2010 22:44:07 Yea issue is, this isn't like web server load balancing where you just have more processes on another server just handle the request like normal due to shared files/scripts. There's the issue that the node keeps things stored in memory as well which requires shutting down the entire node to save the basic states.
|
Camios
Minmatar Insurgent New Eden Tribe
|
Posted - 2010.09.04 23:11:00 -
[12]
Well, things like offensives, pos sieges, system siege, sovereignty battles and so on are planned well in advance by the attacker side.
So the at least attacker knows it well before.
Filling such a form requires just 5 minutes and gives you a better fleet fight experience, while remapping a solarsystem on a node requires that all the people in system is booted and logs back in.
Some problems can be related to traps to titans, or other things that call a lot of people from around the cluster without any possible early notice.
|
Spurty
Caldari D00M. Northern Coalition.
|
Posted - 2010.09.04 23:16:00 -
[13]
Edited by: Spurty on 04/09/2010 23:16:23
Originally by: Aessoroz Edited by: Aessoroz on 04/09/2010 22:51:48
Yea issue is, this isn't like web server load balancing where you just have more processes on another server just handle the request like normal due to shared files/scripts. There's the issue that the node keeps things stored in memory as well which requires shutting down the entire node to save the states before they can start doing anything.
You wut?
Go read up on Xen. Actually you seem lazy, so ..
Quote: Xen's support for virtual machine live migration from one host to another allows workload balancing and the avoidance of downtime.
in short, you are plain wrong about that idea.
Nothing needs to be shutdown at all. This was the point I was making. At worst case (100% load) users experience a 5 second freeze and then everything seems to just be working nice again.
This is a max of 5 seconds (gg burn in tests on VMs).
Might be lower on the sort of hardware CCP could afford.
|
FreddyMac
|
Posted - 2010.09.05 01:47:00 -
[14]
Right, because if it was just that simple, they wouldnt have just done it to shut everyone up lol. How anyone that knows jack about the codebase can act like they know what they are talking about here will never cease to amaze me.....
|
Intense Thinker
Minmatar
|
Posted - 2010.09.05 02:21:00 -
[15]
Originally by: Alara IonStorm
Originally by: Herschel Yamamoto They can only replace a node in the middle of a fight by killing it first. So if you want everyone to be booted off the server in the middle of a fleet fight, then you can get your reinforced node. Or you can ask a day in advance and get them to do it on downtime. They're working on this constraint, but it hasn't been fixed yet.
General Discussion is my territory!
Get your A$$ back to CAOD!
Sir.
Better give CCP 24 hours notice to reinforce the forum node Signature locked for editing a moderator's warning. Zymurgist |
De'Veldrin
Minmatar CareBears on Fire The Obsidian Legion
|
Posted - 2010.09.05 02:25:00 -
[16]
Originally by: Alara IonStorm
Originally by: Herschel Yamamoto They can only replace a node in the middle of a fight by killing it first. So if you want everyone to be booted off the server in the middle of a fleet fight, then you can get your reinforced node. Or you can ask a day in advance and get them to do it on downtime. They're working on this constraint, but it hasn't been fixed yet.
General Discussion is my territory!
Get your A$$ back to CAOD!
Sir.
Someone is going to be scrubbing the plasma conduits with a toothbrush, I think. --Vel
|
Stick Cult
|
Posted - 2010.09.05 02:56:00 -
[17]
Originally by: FreddyMac Right, because if it was just that simple, they wouldnt have just done it to shut everyone up lol. How anyone that knows jack about the codebase can act like they know what they are talking about here will never cease to amaze me.....
This.
Originally by: CCP Tuxford my bad. Rest assured I'm being ridiculed by my co-workers.
|
Abdiel Kavash
Caldari Paladin Order Fidelas Constans
|
Posted - 2010.09.05 03:44:00 -
[18]
There is a notice of at least 24 hours before every important fight available to everyone - and that is the reinforcement timer. A fleet does not just reinforce an IHUB for the lulz, you can count on large battles happening around the time it comes out. I don't know how many nodes are available or how much strain would it put on the servers to potentially reinforce a couple of nodes unnecessarily. But I have seen an entire constellation reinforced at the same time, so probably this won't be an issue.
Maybe it's not worthy reinforcing the nodes for every reinforced POS; I am not sure if it would be possible to make it happen only for R32 and R64 moons. ___________ EVE is dying! Now for real! |
Bill Ny
Gallente
|
Posted - 2010.09.05 03:51:00 -
[19]
Originally by: FreddyMac Right, because if it was just that simple, they wouldnt have just done it to shut everyone up lol. How anyone that knows jack about the codebase can act like they know what they are talking about here will never cease to amaze me.....
Frankly, I agree. Only reason I posted my explanation was because I remember reading a Dev post about it explaining it literally the same way.
Honestly, EVE is far too complex for any of us to figure out in a forum thread. And hey, if you can end lag, why don't you go work for CCP?
If you do it, we'll all mail cookies to you! I am also the Science Guy.
I am not inclined to tell you why ice "cracks." I am entitled to enjoy my goddamn meal. |
Zeba
Minmatar Honourable East India Trading Company
|
Posted - 2010.09.05 04:42:00 -
[20]
Simple guide to node reinforcing bliss. If your fc/alliance head is smart enough to organise a massive op to put something on a timer then they should be smart enough to file a node reinforcement petition at the proper time if said op was succesful or shows signs of being a success. If they didn't do it then they are incompetant or are trying to play the lag card to snafu the upcoming fight when the timer runs out for whatever reasons they might have.
Originally by: CCP Oveur My milkshake brings all the boys to the yard.
Originally by: Ryhss There is no paranoia in Eve, everyone is out to get you....
|
|
Jesslyn Daggererux
Interstellar Brotherhood of Gravediggers The 0rphanage
|
Posted - 2010.09.05 05:18:00 -
[21]
Originally by: Stick Cult
Originally by: FreddyMac Right, because if it was just that simple, they wouldnt have just done it to shut everyone up lol. How anyone that knows jack about the codebase can act like they know what they are talking about here will never cease to amaze me.....
This.
that NAMBLA, standing up and giving attention to your right to love. Also, we inspect things in an unbiased fasion. |
sue denim
|
Posted - 2010.09.05 05:21:00 -
[22]
stop raging, you as well as everyone else whose ever been on a "reinforced node" knows that it does absolutely nothing to help.
|
Doctor Ungabungas
Caldari GoonWaffe Goonswarm Federation
|
Posted - 2010.09.05 06:10:00 -
[23]
Originally by: Herschel Yamamoto They can only replace a node in the middle of a fight by killing it first. So if you want everyone to be booted off the server in the middle of a fleet fight, then you can get your reinforced node. Or you can ask a day in advance and get them to do it on downtime. They're working on this constraint, but it hasn't been fixed yet.
In the 'real world', we've already solved this problem. Unfortunately CCP would rather use their own 'virtualisation solution' rather than using one designed by experts.
|
Ranger 1
Amarr Dynaverse Corporation Sodalitas XX
|
Posted - 2010.09.05 06:26:00 -
[24]
Originally by: Doctor Ungabungas
Originally by: Herschel Yamamoto They can only replace a node in the middle of a fight by killing it first. So if you want everyone to be booted off the server in the middle of a fleet fight, then you can get your reinforced node. Or you can ask a day in advance and get them to do it on downtime. They're working on this constraint, but it hasn't been fixed yet.
In the 'real world', we've already solved this problem. Unfortunately CCP would rather use their own 'virtualisation solution' rather than using one designed by experts.
Heh.
So drop them a line and tell them how simple it is to sort the problem out. I have no doubt you will be amply compensated.
Somehow, I doubt that's going to happen. ===== If you go to Za'Ha'Dum I will gank you. |
Doctor Ungabungas
Caldari GoonWaffe Goonswarm Federation
|
Posted - 2010.09.05 06:43:00 -
[25]
Originally by: Ranger 1 So drop them a line and tell them how simple it is to sort the problem out. I have no doubt you will be amply compensated.
Somehow, I doubt that's going to happen.
I doubt it too. I'm not saying CCP are stupid. I'm saying they are cheap.
|
Aessoroz
|
Posted - 2010.09.05 06:46:00 -
[26]
Originally by: Zeba Simple guide to node reinforcing bliss. If your fc/alliance head is smart enough to organise a massive op to put something on a timer then they should be smart enough to file a node reinforcement petition at the proper time if said op was succesful or shows signs of being a success. If they didn't do it then they are incompetant or are trying to play the lag card to snafu the upcoming fight when the timer runs out for whatever reasons they might have.
Strategically speaking, the smart FC would get into the system first and not request reinforcement.
|
Spurty
Caldari D00M. Northern Coalition.
|
Posted - 2010.09.05 11:23:00 -
[27]
Their reasons for not going this way are their own.
I have heard horror stories which didn't turn out to be very true.
It may be a layering issue though. An extra layer of work.
I give up though. Just use 56000 skill point scout, find 500 man gate camp (wish I was making that up) go no further and shoot the odd straggler.
Then get called a blubber by that 500 man fleet when they get bored and come break up an 8 man camp.
Sigh
Sooner see rats gain bomb launching technology
|
|
|
|
Pages: [1] :: one page |
First page | Previous page | Next page | Last page |