open All Channels
seplocked EVE General Discussion
blankseplocked SuperNodes
 
This thread is older than 90 days and has been locked due to inactivity.


 
Author Topic

Lengram
Posted - 2011.03.31 21:57:00 - [1]
 

not sure if this is the right place, but how many supernodes do you guys have for reinforcing systems for big fights?

CCP Yokai

Posted - 2011.03.31 23:23:00 - [2]
 

We can adjust how many are available every day at Down Time. So I guess the answer is... how many do you need?

Lengram
Posted - 2011.03.31 23:25:00 - [3]
 

Originally by: CCP Yokai
We can adjust how many are available every day at Down Time. So I guess the answer is... how many do you need?


over 9000?

Merouk Baas
Gallente
Posted - 2011.03.31 23:30:00 - [4]
 

Do you guys have any sort of predictive algorythms for where fights are likely to take place? I imagine something can be extrapolated from the (static) truesec map + previous day's fleet fights + sov. reinforcement timers + chokepoints and approach lanes.

Or is it solely based on players notifying you?

Lengram
Posted - 2011.03.31 23:39:00 - [5]
 

Originally by: Merouk Baas
Do you guys have any sort of predictive algorythms for where fights are likely to take place? I imagine something can be extrapolated from the (static) truesec map + previous day's fleet fights + sov. reinforcement timers + chokepoints and approach lanes.

Or is it solely based on players notifying you?


if you watch the eve keynote vid from this years fanfest, you will see that the answer is yes, they do

and that one of the devs (i think it was yokai, but cant remember for certian) bought a $50k server just before fanfest started

Steve Thomas
Minmatar
Sebiestor Tribe
Posted - 2011.04.01 00:56:00 - [6]
 

Edited by: Steve Thomas on 01/04/2011 01:10:08
Edited by: Steve Thomas on 01/04/2011 01:08:56
Edited by: Steve Thomas on 01/04/2011 00:57:03
Edited by: Steve Thomas on 01/04/2011 00:56:27

they try to keep all of the servers the type as mutch as posible for a variety of reasons. that said, they aparently are Testing a new server box that will probably be inserted the next "medium" update they do live. Note this is NOT the Fanfest Frankensever but a diferent box.

currently the core on TQ is

(TQ currently)each server set(there are dozens, I honestly dont know how many)
2 x 8 Core Intel x7560 processors at 2.26GHz

32 x 16GB strips of RAM for a total of 512GB DDR3 RAM

4 x Gigabit Ethernet cards

2 x Dual-port 8Gb/s Fiberchannel cards

4 x 15K 300GB SAS drives internal

9 x 300 GB SSD's drives RAID 5 attached via IBM V7000

36 x 600 GB SAS drives RAID 10 attached via IBM V7000

"Frankenserver" at fanfest was a unknown mark that was (I think) 12 or 16 core 2.33Gh that is on Sisi and MAY be being sent over to the main servers sometime next month or three.

Intel is suposedly sending to them to benchtest a (I think) 16 core 2.666GH base chipset

Lengram
Posted - 2011.04.01 01:19:00 - [7]
 

Originally by: Steve Thomas
that said, they aparently are Testing a new server box that will probably be inserted the next "medium" update they do live. Note this is NOT the Fanfest Frankensever but a diferent box.


i thought the frankenserver was the one they towed on stage with a donkey riding it. then they have a new super secert server from ibm and intel they hasnt been delivered yet (i think it was due to be installed/deleviered 4 weeks after fanfest, which would mean they dont get it for another 3 weeks as of this post.) yokai can correct me if im wrong (or any other devs working on/with the server hardware)

Sidrat Flush
Caldari
Eve Industrial Corp
Posted - 2011.04.01 01:25:00 - [8]
 

pics or it didn't happen.

Hardware **** is the best **** you know you want to post pics devs, why wait for a dev blog!!

Steve Thomas
Minmatar
Sebiestor Tribe
Posted - 2011.04.01 01:46:00 - [9]
 

Originally by: Lengram
Originally by: Steve Thomas
that said, they aparently are Testing a new server box that will probably be inserted the next "medium" update they do live. Note this is NOT the Fanfest Frankensever but a diferent box.


i thought the frankenserver was the one they towed on stage with a donkey riding it. then they have a new super secert server from ibm and intel they hasnt been delivered yet (i think it was due to be installed/deleviered 4 weeks after fanfest, which would mean they dont get it for another 3 weeks as of this post.) yokai can correct me if im wrong (or any other devs working on/with the server hardware)
the new server box may just be a standard Box, or it may be "clone" of Franken thats being sent to Tranq, or Franken itself. (this was brought up in 2 diferent places) Franken may be the one put into Sisi recently, Im not sure the only definit thing I know is that the one that they had on stage with the Donky on it was that that one is not in london for whatever reason.

Originally by: Sidrat Flush
pics or it didn't happen.

Hardware **** is the best **** you know you want to post pics devs, why wait for a dev blog!!
Frankenserver is on one of the fanfest vidioes,

Lord XSiV
Amarr
Kodar Innovations
Posted - 2011.04.01 02:01:00 - [10]
 

Actually CCP has already proven beyond any doubt time and time again that they have no clue when it comes to hardware.

Relying upon IBM to provide them advice which consists of substandard equipment (IBM is considered junk at the enterprise level these days) at over inflated prices is a recipe for disaster. It has been proven time and time again that IBM does not care for the interests of their client. The saying that you couldn't get fired for hiring IBM has been refuted since 1999.

Proof you ask? Well take a look at the Matrix Online for one. Very excellent example of IBM's best work and how seriously they consider the entertainment industry.

More importantly, just look at Eve itself. It can't handle anywhere near the fleet fights CCP advertises so you can definitely say 'the proof is in the pudding'....

Once again it really comes down to the the intellectual capability of the staff working at CCP. After having let go any of the decent developers years ago leaving just a bunch of low end content developers and helpdesk qualified network personnel, what else would you expect? Why else does CCP toss in all these 'distractions' such as the CSM (which the playerbase has obviously rejected based upon the turnout), useless content that caters to WoW players and of course the pathetic 'upgrades' that are buggy with promises of 'fixes' coming.

It will be nice once they do finally sell out. Unfortunately that means that the CCP owners have to come to grips with the fact that CCP isn't worth 16 times what they are asking for it.


Evelgrivion
Gunpoint Diplomacy
Posted - 2011.04.01 02:06:00 - [11]
 

Edited by: Evelgrivion on 01/04/2011 02:06:30
Originally by: Steve Thomas
"Frankenserver" at fanfest was a unknown mark that was (I think) 12 or 16 core 2.33Gh that is on Sisi and MAY be being sent over to the main servers sometime next month or three.

Intel is suposedly sending to them to benchtest a (I think) 16 core 2.666GH base chipset



Just 2.66 GHz? That's the lowest end four-core, four-processor implementation of the Sandy Bridge (32nm) Xeon that I know of in their code books. For $50,000, I was expecting a 3.9 GHz model. Confused

Steve Thomas
Minmatar
Sebiestor Tribe
Posted - 2011.04.01 02:07:00 - [12]
 

http://www.youtube.com/ccpgames#p/c/9923468C02F5604D/2/ibb1M0Ty7es

13 min mark is where he talks about WHAT he thinks Frankensever can do when installed.

Lord XSiV
Amarr
Kodar Innovations
Posted - 2011.04.01 03:12:00 - [13]
 

Originally by: Steve Thomas
http://www.youtube.com/ccpgames#p/c/9923468C02F5604D/2/ibb1M0Ty7es

13 min mark is where he talks about WHAT he thinks Frankensever can do when installed.


Key word there is 'thinks'....

Professionals 'know'

Big difference - guess which one CCP has.


Steve Thomas
Minmatar
Sebiestor Tribe
Posted - 2011.04.01 04:34:00 - [14]
 

Pff riiiiigt.

thats why they install things like black boxes, because the proffesionals who make and maintain aircraft and navigation system all know that nothing will ever go wrong with them and they need to show its always pilot error or freak bouts of extream weather.

Idiots like you are the reason why we have runaway jury verdics in some cases.

Akita T
Caldari Navy Volunteer Task Force
Posted - 2011.04.01 04:47:00 - [15]
 

Edited by: Akita T on 01/04/2011 07:26:28
Originally by: CCP Yokai
We can adjust how many are available every day at Down Time. So I guess the answer is... how many do you need?

7,881 minus however many Jove systems there might be.
Half of that would be nice too.
Tell you what : make it 3,500 and we're happy.
Very Happy

Take top 1,500 busiest systems, put them each on their own supernode. In fact, make the top 200 be super-supernodes.
Then group remaining systems (only between 2 to 4 per group, preferably 3 tops, never 5 or more) on a "combined average load" roughly equal between all groups basis and put them on the remaining 2,000 nodes.
ugh


Yes, yes, that'll never happen, way too expensive.
You'd probably manage to code "on the fly" system node-jumping before that's even remotely affordable.
... is what I used to be tempted to say.

...

Hmm...

...

ALTHOUGH, for the 2000 other "lowball" nodes... never say never, I hear they're making processors with as many as 12 threads each for which dual-processor configurations are possible (at ~3300$/blade just for the CPU)... so that would be 24 nodes per blade... they also support up to 288 GB, so only if you can afford to also get enough RAM into them to handle everything (hmm, 2 GB per node should be enough, right ? so make that two kits of 3x8GB ECC RAM for 48 GB on each blade, or about ~1200$ minimum), plus whatever the blade itself costs, plus blah blah blah, call it 6k$/blade, and you'd need about 80 of those in your racks...
Laughing Only half a million dollars worth of hardware to not have more than 4 systems per node thread ! Laughing
Hell, make that 250 of those puppies (for ~1.5 mil dollars) and you have EACH solar system handled by its own thread... yeah, you're still getting two systems sharing a core, but c'mon...

Ban Doga
Posted - 2011.04.01 06:26:00 - [16]
 

Originally by: CCP Yokai
We can adjust how many are available every day at Down Time. So I guess the answer is... how many do you need?


I know you like to make jokes but that was really a bad one.

Someone should file a couple hundred fleet fight notifications and test your promise of "we will honor every request for reinforced nodes".

Evelgrivion
Gunpoint Diplomacy
Posted - 2011.04.01 06:48:00 - [17]
 

Edited by: Evelgrivion on 01/04/2011 06:50:22
Turns out you know more when you actually watch the source keynote. Who knew? Wink

Regarding the super server at $50,000 ("after discounts... *******s." Laughing), I've got a sneaking suspicion that I was on the right track with Sandy Bridge, but with the wrong process. Given Intel's progress, the CPUs inside might be Ivy Bridge, which is a 22nmExclamation version of the Sandy Bridge architecture, with a 20% performance advantage over its 32nm Counterpart. With two process advantages and two architecture advantages over the current server nodes, this makes the claimed capabilities of the new hardware sound reasonably accurate.

Super Server? Do want.
Please visit your user settings to enable images.

Rasz Lin
Caldari
Uitraan Diversified Holdings Incorporated
Posted - 2011.04.01 12:50:00 - [18]
 

"supernode" doesnt exist
When you request reinforced node all this does is give your sol system dedicated thread on a multiprocessor node, so instead of 2-5-10 solar systems running on one thread you have 'only' one. Nothing super about that, in fact it SUCKS because solar system code just like EVE client is SINGLE THREADED and DOESNT SCALE FOR SHIAT.

Lyta Jhonson
Posted - 2011.04.01 13:24:00 - [19]
 

When I first read about this I thought: that's what eve needs - officially overclocked server processor with only 2 out of 6 cores active.

CCP Yokai

Posted - 2011.04.01 20:31:00 - [20]
 

Originally by: Steve Thomas
Edited by: Steve Thomas on 01/04/2011 01:10:08
Edited by: Steve Thomas on 01/04/2011 01:08:56
Edited by: Steve Thomas on 01/04/2011 00:57:03
Edited by: Steve Thomas on 01/04/2011 00:56:27

they try to keep all of the servers the type as mutch as posible for a variety of reasons. that said, they aparently are Testing a new server box that will probably be inserted the next "medium" update they do live. Note this is NOT the Fanfest Frankensever but a diferent box.

currently the core on TQ is

(TQ currently)each server set(there are dozens, I honestly dont know how many)
2 x 8 Core Intel x7560 processors at 2.26GHz

32 x 16GB strips of RAM for a total of 512GB DDR3 RAM

4 x Gigabit Ethernet cards

2 x Dual-port 8Gb/s Fiberchannel cards

4 x 15K 300GB SAS drives internal

9 x 300 GB SSD's drives RAID 5 attached via IBM V7000

36 x 600 GB SAS drives RAID 10 attached via IBM V7000

"Frankenserver" at fanfest was a unknown mark that was (I think) 12 or 16 core 2.33Gh that is on Sisi and MAY be being sent over to the main servers sometime next month or three.

Intel is suposedly sending to them to benchtest a (I think) 16 core 2.666GH base chipset



The Frankenserver... not in or ever going to be in production... is 4.44GHz by 6 cores with 2100MHz RAM and 2 x TCP Offload NIC's. It was purely for testing a proof of concept.

The current TQ blades are 3.33GHz on a much older chip, 800MHz RAM... and on TCP offload.

We should be able to double almost any fleet fight with the new hardware that I am not allowed to talk about the spec's on yet. But safe to say... it meet's or exceeds the Frankenserver's capabilities all without any liquid cooling in our nice clean datacenter ;)

Love Always,
CCP Yokai


CCP Yokai

Posted - 2011.04.01 20:33:00 - [21]
 

Originally by: Rasz Lin
"supernode" doesnt exist
When you request reinforced node all this does is give your sol system dedicated thread on a multiprocessor node, so instead of 2-5-10 solar systems running on one thread you have 'only' one. Nothing super about that, in fact it SUCKS because solar system code just like EVE client is SINGLE THREADED and DOESNT SCALE FOR SHIAT.


There is a supernode today... it is the special server we used on Jita for the Jita 2000+ test. It is TQ-SOL-099 and is used for remapping the biggest battles right now. When the new system goes in place it will replace the current supernode and will get a new cool name :)

Dendel Hypermach
Posted - 2011.04.27 16:44:00 - [22]
 

supposedly the $50k IBM server is going in shortly. can't wait to hear what it is.

hey, @ccpgames, whats the most exotic efforts you've taken in building frankensevers?

Dendel Hypermach
Posted - 2011.05.18 01:19:00 - [23]
 

yo guys, still hoping for an update on what exactly the frankserver is. more than understandable if the CPU is still "under wraps," but there was a lot of conjecture it'd be the dual core 4.4GHZ Intel X5698 which was quietly put into the wild mid-march.

Lady Spank
Amarr
In Praise Of Shadows
Posted - 2011.05.18 01:36:00 - [24]
 

Requesting a supernode for every single system every single day.

Dendel Hypermach
Posted - 2011.06.07 15:47:00 - [25]
 

CCP, are you able to update us on the supernode you rolled out at fanfest yet?


 

This thread is older than 90 days and has been locked due to inactivity.


 


The new forums are live

Please adjust your bookmarks to https://forums.eveonline.com

These forums are archived and read-only