Pages: 1 [2] 3 :: one page |
|
Author |
Thread Statistics | Show CCP posts - 9 post(s) |
GRIEV3R
Clan Shadow Wolf Fatal Ascension
16
|
Posted - 2012.02.09 21:36:00 -
[31] - Quote
TDI is one of the most technically impressive things I've ever seen. Moreover, its implementation has been a triumph of QA and customer service. Excellent work, CCP. I am truly impressed. You guys are pros.
I love how the blog just casually comments that "We pause the universe at shutdown." That's so ******* cool! |
Kusanagi Kasuga
Ferocious Felines
12
|
Posted - 2012.02.09 21:38:00 -
[32] - Quote
I am curious. What difference does it make that you pause the universe at downtime - why is it done? |
Salpun
Paramount Commerce Tactical Invader Syndicate
189
|
Posted - 2012.02.09 21:52:00 -
[33] - Quote
Kusanagi Kasuga wrote:I am curious. What difference does it make that you pause the universe at downtime - why is it done? It means that the server has a clean value to save. Lets say you are in a fight when the server shuts down. Between the client and the server the position of your ship is in flux. If you freeze the server you can get a clean save, you can shut down do what you need to do to the DB and start the server up frozen.
This is the same tech they use to remap live nodes but becouse it requires us to relogin, it is it is not used to remap nodes under stress. Once the tech changed so you get kicked back to the character screen during remaps or the screen just freezes and the node gets remaped while you wait. Remaping nodes will be done on the fly which will be better for everyone.
Lastly using it for down time is a good way to test the TIDI frame work. you get a daily test of its fuctinality. |
|
CCP Veritas
C C P C C P Alliance
326
|
Posted - 2012.02.09 21:58:00 -
[34] - Quote
Kusanagi Kasuga wrote:I am curious. What difference does it make that you pause the universe at downtime - why is it done? It takes a fair bit of processing to get a character logged out gracefully, but since we're shutting down anyway none of that matters. By pausing the universe we avoid a large bulk of that useless processing.
If I'm remembering what Atlas told me correctly (he did the work on the shutdown routine), that change alone moved our shutdown time from 2 minutes to 42 seconds. CCP Veritas - Senior Programmer - EVE Software |
|
|
GM Homonoia
Game Masters C C P Alliance
276
|
Posted - 2012.02.09 22:07:00 -
[35] - Quote
CCP Veritas wrote:Kusanagi Kasuga wrote:I am curious. What difference does it make that you pause the universe at downtime - why is it done? It takes a fair bit of processing to get a character logged out gracefully, but since we're shutting down anyway none of that matters. By pausing the universe we avoid a large bulk of that useless processing. If I'm remembering what Atlas told me correctly (he did the work on the shutdown routine), that change alone moved our shutdown time from 2 minutes to 42 seconds.
It also has the happy side effect that everything stops simultaneously. Before this it was possible to get disconnected due to shut down, only to have NPCs destroy your ship even though the server was officially offline already. Other, similar side effects were possible as well. Senior GM Homonoia | Info Group | Senior Game Master |
|
HELIC0N ONE
GoonWaffe Goonswarm Federation
81
|
Posted - 2012.02.09 22:52:00 -
[36] - Quote
GM Homonoia wrote:[quote=CCP Veritas]It also has the happy side effect that everything stops simultaneously. Before this it was possible to get disconnected due to shut down, only to have NPCs destroy your ship even though the server was officially offline already. Other, similar side effects were possible as well.
I believe there were occasions in the past where fleets got disconnected due to downtime whilst fighting at a hostile POS, and then were blown up by it whilst TQ was offline . Funny,if you weren't the one on the losing side of this, not so much if you were. |
Indahmawar Fazmarai
The I and F Taxation Trust
238
|
Posted - 2012.02.09 23:01:00 -
[37] - Quote
Mashie Saldana wrote:(...) Please, read up on how the TQ cluster is built and designed before making silly suggestions.
That could be an interesting read, is it available somewhere?
EVE residents: 5% Wormholes; 8% Lowsec; 20% Nullsec; 67% Highsec. CSM 6: 100% Nullsec residents.
EVE demographics vs CSM demographics, nothing to worry about...-á |
|
CCP Phantom
C C P C C P Alliance
971
|
Posted - 2012.02.09 23:28:00 -
[38] - Quote
Indahmawar Fazmarai wrote:Mashie Saldana wrote:(...) Please, read up on how the TQ cluster is built and designed before making silly suggestions. That could be an interesting read, is it available somewhere? A good start would be the EVElopedia article here, followed by this devblog about upgrades. Also don't miss the Fanfest 2011 presentation about TQ's new toys. It is quite interesting stuff.
And on top of all this we now have the fantastic Time Dilation! CCP Phantom - German Community Coordinator |
|
Matuk Grymwal
Lapse Of Sanity Exhale.
2
|
Posted - 2012.02.10 01:16:00 -
[39] - Quote
Just wondering, how seriously have you looked at using virtualisation and live migration for better load balancing? As you say porting a software load at the application level over to a different server is nigh on impossible unless you design it from the ground up to support that. However shuffling an entire VM over to a different (beefier) server is much easier with the live migration support available in most commercial hypervisors. So on an overloaded node you could up the TIDI factor, possibly even briefly pausing the node like you do at shutdown, live migrate to a reinforced node, dynamically up the RAM/VCPU allocation on the VM and then unlock TIDI to run as normal.
-Mat |
Karsa Egivand
Sebiestor Tribe Minmatar Republic
8
|
Posted - 2012.02.10 01:24:00 -
[40] - Quote
Quote:..., which pauses the universe while shutting down.
Completely aside from the technical issues, this simply sounds awesome. |
|
|
CCP Veritas
C C P C C P Alliance
336
|
Posted - 2012.02.10 09:49:00 -
[41] - Quote
Matuk Grymwal wrote:Just wondering, how seriously have you looked at using virtualisation and live migration for better load balancing? As you say porting a software load at the application level over to a different server is nigh on impossible unless you design it from the ground up to support that. However shuffling an entire VM over to a different (beefier) server is much easier with the live migration support available in most commercial hypervisors. So on an overloaded node you could up the TIDI factor, possibly even briefly pausing the node like you do at shutdown, live migrate to a reinforced node, dynamically up the RAM/VCPU allocation on the VM and then unlock TIDI to run as normal. I never said it was "nigh on impossible", we have plans and such they're just not easy ;)
The fine folks over in Operations have been running experiments with virtualization in order to get a better idea how the Eve server specifically reacts to it. For instance, Sisi is currently virtualized. If it works out and we can migrate smoothly, that would allow us to take *groups* of solar systems and put them on beefier hardware. It doesn't give us the ability to isolate a specific solar system though, which is what the software-level migration (non-destructive live remapping, we call it) would give us.
As always, a step in the right direction is always good even if it doesn't go *all* the way, so here's hoping the tests go well! CCP Veritas - Senior Programmer - EVE Software |
|
smaster
BLOOM. 0ccupational Hazzard
16
|
Posted - 2012.02.10 09:56:00 -
[42] - Quote
Time Delation has shown us players esp one thing, that your hardware is indeed undersized and not capable of running the users playing the game.
CCP U NEED MORE SERVERS FOR RUNNING THE GAME
When u fly in vale of the silent, and 4 regions further in branch there is a fight, and u get stuck in 90% time delation and it takes u 20 minutes to travel 3 jumps, YOU NEED MORE SERVERS FOR 0.0 |
thecunning mrfox
Darwin's response
9
|
Posted - 2012.02.10 10:19:00 -
[43] - Quote
Karsa Egivand wrote:Quote:..., which pauses the universe while shutting down. Completely aside from the technical issues, this simply sounds awesome.
Always makes me think of dark city http://www.youtube.com/watch?v=S5ZSDCvUwN8 |
TR4D3R4LT
Pator Tech School Minmatar Republic
28
|
Posted - 2012.02.10 11:59:00 -
[44] - Quote
CCP Veritas wrote:BeanBagKing wrote:Does this mean you're fixing bracket lag?!? Work has been done along those lines. I can't say if any individual will call it "fixed" with what's been done, but it's possible.
You decided to remove brackets alltogether right? I hope not...
Still credit where credit is due, I've enjoyed TiDi in couple fights thus far and they've certainly made my fight more responsive. Overall slower, but responsive.
Also any change for blog that updates how TQ looks atm, how many dedicated supernodes, what they have inside, how many normal nodes, how many systems per average are hosted on them, future plans hardware/software wise etc. You fellas know stuff like that is p0rn to some of us and making blogs about server things is easiest way to let us know that you care. Nobody needs constant updates but once a year general outline "this is what we have now, this is what we plan to get during this year and this is SoonGäó" would go long way. |
Gevlin
EXPCS Corp SpaceMonkey's Alliance
98
|
Posted - 2012.02.10 12:02:00 -
[45] - Quote
I know in the old say when a side was about to loose, including their Super Caps they would perposly over load the node to make it crash. Then there would he a long file of people petitioning to get their shinny toy replaced.
How has the customer services been since this implementations.
have the Customer Services been able to replace few ships resulting from large fights? I agree with several people: CCP needs to focus most of eve's recources on FIS, but the development of WIS still needs to continue, just as a slower and more efficient pace. In eve I wish to be more than just a machine. |
Pinky Denmark
The Cursed Navy Important Internet Spaceship League
86
|
Posted - 2012.02.10 13:22:00 -
[46] - Quote
Indeed it seems to be very nice and I like it...
Question: Being caught up in time dilation will your skills train slower and in effect result in you training slower than people outside the time dilation? What about manufacture time and research time? Are those affected as well?
Pinky |
Crucis Cassiopeiae
EvE-COM
841
|
Posted - 2012.02.10 15:13:00 -
[47] - Quote
One question about TiDi: Is it possible to put in game some space distortion/anomaly that appear in the space that's under time dilation? This way players will have a feeling like its game feature, especially in systems that are on same node an system where fight is. And add message from our ship computer, something like: "Ships sensors discovered time flux anomaly generated by unknown source. Ship computer cant predict time variables in this conditions."
There would be few pluses of this: 1. From annoying thing that ppl don't know why it is happening you make new game feature. In all space related stories there was some sort of time flux anomaly so its all ok from the side of the space story. 2. You add some nice new graphic of space distortion/anomaly in the space. 3. You add one new thing that ppl cant predict in the game (good thing). 4. You give ppl a story behind all of this.
Just a thought. |
John McCreedy
Eve Defence Force Fatal Ascension
101
|
Posted - 2012.02.10 15:27:00 -
[48] - Quote
CCP Phantom wrote:I want my 45 minutes back!!!
Well you can joke about it but it's incredibly frustrating when you're chasing a target and get traffic a 90 second control and then you're hit with 50% TiDi on top of that. 3 Minutes is the difference between a fight and blueballs. |
Lyron-Baktos
Selective Pressure Rote Kapelle
48
|
Posted - 2012.02.10 15:54:00 -
[49] - Quote
TiDi sucks when you are not in the big fight. I lost a scout ship a while back because right when I tackled a ship and my fleet was coming up behind me to support me, they got hit in TiDi and didn't make it to me in time. On holiday. -áIn some other world. Where the music of the radio was a labyrinth of sonorous colours. To a bright centre of absolute convicton. -áWhere the dripping patchouli was more than scent. -á It was a sun |
Hakaru Ishiwara
Republic Military School Minmatar Republic
73
|
Posted - 2012.02.10 16:00:00 -
[50] - Quote
CCP Veritas wrote:Mr Bigwinky wrote:I hear there have been some issues.
One day our wormhole was time dilated, only our wormhole and with only 5 people in it at that time.
Care to shed some light on the possible reason as to why that would happen? Wormholes, being nice semi-population-capped things, group up very nicely. Your average wormhole-housing node is going to have a few hundred wormhole systems on it. While that sounds nifty from a technical resource conservation standpoint, that is **** from an immersion and game play perspective.
I remember the days that you could tell that a 100-pilot, Sniper BS gang was moving in or near your region. The game / node would show performance stress and that was fine. You knew that you were in a null-sec and that big fleets were a part of the environment. [Yeah, now it is 500 - 1000 pilot fleets.]
W-space is / was / should be a totally different environment. Despite the settling down (aka farming) of some w-space systems, w-space is big, mysterious and above all else: remote.
Having a game mechanic indicate that you are on a node with hundreds of other like-systems ruins that image IMHO.
Are we not paying enough cash to support the infrastructure? Has CCP made the choice to go the cheapest possible route, provisioning marginal service while pocketing the rest?
What can be done from a business perspective to correct this? 284 Characters 284 Characters 284 Characters 284 Characters 284 Characters 284 Characters 284 Characters 284 Characters 284 Characters 284 Characters 284 Characters 284 Characters 284 Characters 284 Characters 284 Characters 284 Characters 284 Characters 284 Characters 284286 |
|
Scrapyard Bob
EVE University Ivy League
676
|
Posted - 2012.02.10 16:13:00 -
[51] - Quote
Hakaru Ishiwara wrote: What can be done from a business perspective to correct this?
Convince those 10-20% of subscription losses since last summer to start subscribing again.
(The bleeding stopped in December, once they revealed that Crucible was going to focus on the core EVE experiences with lots of bug fixes. But a lot of folks are still waiting to see whether CCP can continue to do things right, or whether they'll go back to not listening to feedback and just implementing whatever the hell they want, damn the consequences.) |
Vincent Athena
V.I.C.E. Comic Mischief
431
|
Posted - 2012.02.10 19:35:00 -
[52] - Quote
Pinky Denmark wrote:Indeed it seems to be very nice and I like it...
Question: Being caught up in time dilation will your skills train slower and in effect result in you training slower than people outside the time dilation? What about manufacture time and research time? Are those affected as well?
Pinky
As I recall, skills, industry, and the market are not slowed by TiDi. I am running for the CSM https://forums.eveonline.com/default.aspx?g=posts&m=779668#post779668 |
Saucy Okanata
GoonWaffe Goonswarm Federation
0
|
Posted - 2012.02.10 19:36:00 -
[53] - Quote
Time dilation is amazing. It makes the funnest part of large fleet battles last even longer, watching people blow up. How can you go wrong? Now instead of floating in space for an hour waiting, and then a lagged out fight that happens in the blink of an eye, you actually get to spend a good amount of time in fleets actively fighting. Also, since it makes it easier to reship and get back to the fight in time the large fleet battles have become really interesting, they've become an event rather than just a quick show down.
I'm really looking to forward to Crucible 1.3 with the overview and bracket update. Even with time dilation it's hard to get your client to be responsive in large fleet battles, even with shutting off all the graphics and brackets. Things are certainly looking up for fleet battles. |
Ranger 1
Ranger Corp
952
|
Posted - 2012.02.10 20:58:00 -
[54] - Quote
Hakaru Ishiwara wrote:CCP Veritas wrote:Mr Bigwinky wrote:I hear there have been some issues.
One day our wormhole was time dilated, only our wormhole and with only 5 people in it at that time.
Care to shed some light on the possible reason as to why that would happen? Wormholes, being nice semi-population-capped things, group up very nicely. Your average wormhole-housing node is going to have a few hundred wormhole systems on it. While that sounds nifty from a technical resource conservation standpoint, that is **** from an immersion and game play perspective. I remember the days that you could tell that a 100-pilot, Sniper BS gang was moving in or near your region. The game / node would show performance stress and that was fine. You knew that you were in a null-sec and that big fleets were a part of the environment. [Yeah, now it is 500 - 1000 pilot fleets.] W-space is / was / should be a totally different environment. Despite the settling down (aka farming) of some w-space systems, w-space is big, mysterious and above all else: remote. Having a game mechanic indicate that you are on a node with hundreds of other like-systems ruins that image IMHO. Are we not paying enough cash to support the infrastructure? Has CCP made the choice to go the cheapest possible route, provisioning marginal service while pocketing the rest? What can be done from a business perspective to correct this?
With all of the strange effects possible in WH system, a rare time anomally should serve to deepen your immersion.
I would not prefer nodes to be in proximity to each other as you describe above... it would be foolish in the extreme and only magnify the stress the node would be under as ships converge on a contested system.
I'm not sure where you comment about "not paying enough cash to support the infrastructure" is coming from. Right now the infrastructure is larger and healthier than it has been at any time in the past.... and I hardly think you will find many players to support you in saying the CCP is purposefully providing "marginal service while pocketing the rest". They are allowed to actually make a profit and pay their own salaries you know.
A solution to these issues isn't as simple as "buy more hardware"... indeed, that would be a great way to try (futilely) to AVOID dealing with these issues. When I check troll in the dictionary, it has a photo shopped picture of you standing somewhere in the vicinity of a point.
Also, I can kill you with my brain. |
Morgals
Sturm Reich Sturmgrenadier Syndicate
11
|
Posted - 2012.02.10 21:37:00 -
[55] - Quote
Could lore this away. Be cool to have time dilation explained as to many warp drives messing up time and space. In space animation would be neat as well. Looking for a mature, adult gaming community that has been active in EvE since 2004?Recruitment is open! Come join our public channel and get to know us. SGHQ-PUBLIC [url]http://sgeve.dai-coar.com/[/url] |
Ranger 1
Ranger Corp
952
|
Posted - 2012.02.10 22:16:00 -
[56] - Quote
I'm sure this has been explored, but to me the more logical way to attempt (as a baby step perhaps) to beef up a node under stress would be to simply shut down and move any systems on the node that are currently empty.
Then as more systems on the node empty out (due to people moving aournd) you repeat the pause/shut down/ move routine to get the stressed node down to the minimum number of systems possible... possibly even down to only having the overloaded system being the only one left on the node.
I realize this is somewhat the opposite of moving the single overloaded system off the node, but the end result would be (reasonably close) to the same. When I check troll in the dictionary, it has a photo shopped picture of you standing somewhere in the vicinity of a point.
Also, I can kill you with my brain. |
Salpun
Paramount Commerce Tactical Invader Syndicate
189
|
Posted - 2012.02.10 22:26:00 -
[57] - Quote
Ranger 1 wrote:I'm sure this has been explored, but to me the more logical way to attempt (as a baby step perhaps) to beef up a node under stress would be to simply shut down and move any systems on the node that are currently empty.
Then as more systems on the node empty out (due to people moving aournd) you repeat the pause/shut down/ move routine to get the stressed node down to the minimum number of systems possible... possibly even down to only having the overloaded system being the only one left on the node.
I realize this is somewhat the opposite of moving the single overloaded system off the node, but the end result would be (reasonably close) to the same. Its all about the server tools required and how long it takes to write and QA. Right now more easy and faster wins can be found in the client optimizations route.
When they get those fixed hopefully the planing work will already be done so its a straight forword path to writing and QA any code that is needed.
CCP from hard experience knows how many bugs get created from "simple" low level changes. The Client fixes are one example. On sisi the client is still fairly buggy but the results of the changes are major wins. The issue is finding the edge cases and the only way to do that is to let players test it. As was sayed above the next step is already in testing. If it gets cleared to deploy to TQ then we are one step closer.
Your ideas match mine but we just have to give the devs some time. |
boeboe joe
SQUINGEL Nulli Tertius
1
|
Posted - 2012.02.11 03:48:00 -
[58] - Quote
CCP Veritas wrote: Dev blog win, with some winning graphs. Also Team Gridlock is deploying win in Crucible 1.2.
I think something needs to be added to make this official, Picard. Credit to CCP Soundwave for the pic.
|
Alain Kinsella
94
|
Posted - 2012.02.11 07:27:00 -
[59] - Quote
John McCreedy wrote:The only criticism of TiDi I have is (as mentioned in the blog) jumping through gates, specifically the fact that you still get traffic control and then on top of that, TiDi kicks in meaning your traffic control takes even longer. Would scrapping Traffic control and relying solely on TiDi when jumping through gates cause more problems than it would be worth or is TC just a redundant system that's not been taken out due to time constraints?
The problem with the 'jump lag' is that's its a handoff between two different sol node processes. Your client must open a new connection to the destination node, destination node tells the source node 'Hi, I've got this one', the source node disconnects, and the destination node tells the client to start loading environment. That 'hand off' has to be done within less than a second, *and* without any delays due to system load.
I know this problem very well, from some early work on reverse engineering the Second Life protocol (their case is very extreme, because you can have a vehicle crossing multiple 'systems' within a second as well!). That was back in 2006 or so, and last I heard the problem is still not solved over there.
From another angle, I think CCP does have the edge on this problem. Apparently windows systems have a facility to move an existing open socket between processes - even across computers. The downside, when/if implemented, is that Eve will no longer work in Wine/Linux as this feature has not been re-created there yet (it broke Uru's client in Wine when something similar was implemented there). Mac should be OK though.
[If Veritas or another Dev could pipe up at some point with confirmation or details, it'd be appreciated. ] I may have come here from Myst Online, but that does not make me any less bloodthirsty than the average Eve player.
Just more subtle.
|
Mioelnir
Cataclysm Enterprises Ev0ke
49
|
Posted - 2012.02.11 12:36:00 -
[60] - Quote
CCP Veritas wrote:42 seconds I see what you did there. |
|
|
|
|
Pages: 1 [2] 3 :: one page |
First page | Previous page | Next page | Last page |