| Pages: 1 2 3 [4] :: one page |
|
|
| Author |
Topic |
Veng3ance
Multiversal Enterprise Inc.
Cry Havoc.
    |
Posted - 2008.10.08 18:07:00 -
[91]
Nice work guys, thanks for the update!
|
Nadezjda Allilujeva
|
Posted - 2008.10.08 19:03:00 -
[92]
whatever I want an cargohold in my ship for mine longbearded mongolian goat's! Please i would be happy!
|
Mack Dorgeans
Camelot Innovations
|
Posted - 2008.10.08 22:13:00 -
[93]
It sure seems like the improvements to the back-end are being done primarily to benefit the minority of players -- 1000 in a fleet battle here, or 1000 in Jita there, or mission runners in specific high-traffic systems, etc.
What happened to all the rah-rah anti-blob sentiment CCP posted over the last couple years? Did you give up and decide more is always better? It sure seems to me that in the end all that back-end work will just encourage players to congregate in over-populated systems more, and engage in more blob warfare until they reach the new limits of the server technology.
EVE has TONS of empty space going largely unused, and instead of opening up the game mechanics to encourage people to spread out and utilize that space that blades are currently doing very little to run, it seems like you're pandering to the big alliances, shopping alts, and mission farmers (some of whom are also isk farmers).
What are you doing for the "little guy" or smaller corp/alliance players?
On an unrelated note, I would like to mention there are other technological hurdles that don't seem to be improving. For instance, when I run the game in windowed mode (premium graphics), I get a significant drop in FPS for no apparent reason, even if I'm not running anything else but the usual background processes (under Vista 64-bit). In fullscreen today I noted FPS in the mid-50s with my current settings while warping. There was no obvious stuttering. As soon as I switch to windowed, that drops by 10-15 points, and obvious stuttering appears constant. When I dock in a Caldari station with those silly spotlights, that number drops another 10 points. Since the fullscreen mode does a pretty poor job alt-tabbing between tasks, I like to run windowed when multitasking, yet even before I run another application or second copy of the client, I'm already seeing a significant performance drop.
Yes, I'll submit this as a bug report, but the fact remains this and other everyday performance issues don't appear to be improving at all while others seem to be improving only enough to encourage players to hit new limits in a never-ending cycle.
 |
Aprudena Gist
GoonFleet
GoonSwarm
|
Posted - 2008.10.08 22:54:00 -
[94]
 Originally by: Mack Dorgeans
It sure seems like the improvements to the back-end are being done primarily to benefit the minority of players -- 1000 in a fleet battle here, or 1000 in Jita there, or mission runners in specific high-traffic systems, etc.
What happened to all the rah-rah anti-blob sentiment CCP posted over the last couple years? Did you give up and decide more is always better? It sure seems to me that in the end all that back-end work will just encourage players to congregate in over-populated systems more, and engage in more blob warfare until they reach the new limits of the server technology.
EVE has TONS of empty space going largely unused, and instead of opening up the game mechanics to encourage people to spread out and utilize that space that blades are currently doing very little to run, it seems like you're pandering to the big alliances, shopping alts, and mission farmers (some of whom are also isk farmers).
What are you doing for the "little guy" or smaller corp/alliance players?
On an unrelated note, I would like to mention there are other technological hurdles that don't seem to be improving. For instance, when I run the game in windowed mode (premium graphics), I get a significant drop in FPS for no apparent reason, even if I'm not running anything else but the usual background processes (under Vista 64-bit). In fullscreen today I noted FPS in the mid-50s with my current settings while warping. There was no obvious stuttering. As soon as I switch to windowed, that drops by 10-15 points, and obvious stuttering appears constant. When I dock in a Caldari station with those silly spotlights, that number drops another 10 points. Since the fullscreen mode does a pretty poor job alt-tabbing between tasks, I like to run windowed when multitasking, yet even before I run another application or second copy of the client, I'm already seeing a significant performance drop.
Yes, I'll submit this as a bug report, but the fact remains this and other everyday performance issues don't appear to be improving at all while others seem to be improving only enough to encourage players to hit new limits in a never-ending cycle.
Your kinda dumb if you think this is CCPs issue with your FPS loss in vista in windowed mode its called Aero it uses the graphics card to display a all your desktop crap turn if off and you'll gain the fps back. People like you waste developers time with issues like these.
|
Mashie Saldana
Sharks With Frickin' Laser Beams
|
Posted - 2008.10.08 23:08:00 -
[95]
Originally by: Del ReyII
I'd say its more like LGA771 Wolfdale running on Intel 5000p using 4x 4GB FBDIMMs. Probably like a IBM BladeCenter HS21 .
There are no dual core 3.3GHz Xeon CPUs, the closest is the 3.4GHz Xeon X5272 running the Harpertown core.
I do however stand corrected regarding my first post, there are 16GB memory kits (4x4GB) that will work with P45 based motherboards. So 16GB Wolfdale systems are possible. 
|
Mack Dorgeans
Camelot Innovations
|
Posted - 2008.10.09 01:29:00 -
[96]
Originally by: Aprudena Gist
Originally by: Mack Dorgeans
It sure seems like the improvements to the back-end are being done primarily to benefit the minority of players -- 1000 in a fleet battle here, or 1000 in Jita there, or mission runners in specific high-traffic systems, etc.
What happened to all the rah-rah anti-blob sentiment CCP posted over the last couple years? Did you give up and decide more is always better? It sure seems to me that in the end all that back-end work will just encourage players to congregate in over-populated systems more, and engage in more blob warfare until they reach the new limits of the server technology.
EVE has TONS of empty space going largely unused, and instead of opening up the game mechanics to encourage people to spread out and utilize that space that blades are currently doing very little to run, it seems like you're pandering to the big alliances, shopping alts, and mission farmers (some of whom are also isk farmers).
What are you doing for the "little guy" or smaller corp/alliance players?
On an unrelated note, I would like to mention there are other technological hurdles that don't seem to be improving. For instance, when I run the game in windowed mode (premium graphics), I get a significant drop in FPS for no apparent reason, even if I'm not running anything else but the usual background processes (under Vista 64-bit). In fullscreen today I noted FPS in the mid-50s with my current settings while warping. There was no obvious stuttering. As soon as I switch to windowed, that drops by 10-15 points, and obvious stuttering appears constant. When I dock in a Caldari station with those silly spotlights, that number drops another 10 points. Since the fullscreen mode does a pretty poor job alt-tabbing between tasks, I like to run windowed when multitasking, yet even before I run another application or second copy of the client, I'm already seeing a significant performance drop.
Yes, I'll submit this as a bug report, but the fact remains this and other everyday performance issues don't appear to be improving at all while others seem to be improving only enough to encourage players to hit new limits in a never-ending cycle.
Your kinda dumb if you think this is CCPs issue with your FPS loss in vista in windowed mode its called Aero it uses the graphics card to display a all your desktop crap turn if off and you'll gain the fps back. People like you waste developers time with issues like these.
Hey, thanks for the tip. I appreciate helpful comments. It must be hard being so smart, which explains how rude you are, but then most geniuses have poor social skills.
By the way, turning Aero off didn't improve my FPS (in fact, my numbers were slightly worse), though it did kill most of the stuttering.
Still, this was just one thing that happened to be bothering me at the moment I posted. My point stands that there are other areas which seem to be slow to improve, yet certain fractions of the population seem to be getting the lion's share of attention. I'm all for improving the node performance, but I think encouraging more and more people to congregate in fewer and fewer systems is sort of silly when there's so much empty space.
|
Heptameron
Gallente
Octavian Vanguard
RAZOR Alliance
|
Posted - 2008.10.09 09:42:00 -
[97]
Originally by: Mack Dorgeans
It sure seems like the improvements to the back-end are being done primarily to benefit the minority of players -- 1000 in a fleet battle here, or 1000 in Jita there, or mission runners in specific high-traffic systems, etc.
What happened to all the rah-rah anti-blob sentiment CCP posted over the last couple years? Did you give up and decide more is always better? It sure seems to me that in the end all that back-end work will just encourage players to congregate in over-populated systems more, and engage in more blob warfare until they reach the new limits of the server technology.
EVE has TONS of empty space going largely unused, and instead of opening up the game mechanics to encourage people to spread out and utilize that space that blades are currently doing very little to run, it seems like you're pandering to the big alliances, shopping alts, and mission farmers (some of whom are also isk farmers).
What are you doing for the "little guy" or smaller corp/alliance players?
Kind of self defeating argument there right? 'EVE has TONS of empty space going largely unused' so go use it.
Eve is a game yes, but it's 'character' is one developed by live human beings and if you take a look out of your window at the pretty lights you will see how a major part of human nature is to congregate into large groups. Part of that human nature is conquest, but to acheive that you generally need a bigger stick than the guy you're going to conquer. It's a never ending vicious cycle but unfortunately unless you remap it something like WoW then there is no way around it.
There are plenty of smaller alliances in game that exist and survive quite nicely in 0.0 space, ok maybe paying a premium to do it but hey... it's supposed to be a cost/risk kind of game unlike some others where even if my elven underpants of stealth get damaged i never lose them.
|
Heptameron
Gallente
Octavian Vanguard
RAZOR Alliance
|
Posted - 2008.10.09 09:46:00 -
[98]
Hmmm damn and forgot my original point... the whole 'tell a GM 24 hours before a fight starts' is pretty dire. Maybe we should also drop in exact numbers/shipe types/loadouts/drones overview of tactics to be used bla bla...
|
Rho'varo
Minmatar
Diversified Operational Services
|
Posted - 2008.10.09 10:25:00 -
[99]
Originally by: CCP Mindstar
Gentlemen!
Thanks. Also, I'm looking forward to seeing what improvements are going to be announced for/to women!
 |
|
CCP Mindstar
|
Posted - 2008.10.09 11:29:00 -
[100]
Originally by: Mashie Saldana
Originally by: Del ReyII
I'd say its more like LGA771 Wolfdale running on Intel 5000p using 4x 4GB FBDIMMs. Probably like a IBM BladeCenter HS21 .
There are no dual core 3.3GHz Xeon CPUs, the closest is the 3.4GHz Xeon X5272 running the Harpertown core.
I do however stand corrected regarding my first post, there are 16GB memory kits (4x4GB) that will work with P45 based motherboards. So 16GB Wolfdale systems are possible.
Hey guys,
Just to clear this up, there are indeed such things as dual core 3.3ghz Xeon Processors, and the IBM blades that we are using are capable of 16gb memory out of the box. 
-- |
|
|
Giuno Dea
|
Posted - 2008.10.09 12:29:00 -
[101]
Maybe you should buy more of those Xeons, equip them all with 8 or 16Gb ram, then overclock them to at least 4Ghz, with some liquid cooling which isn't too hard to do. That would, without a great expense, make things alot better.
|
Hawk TT
Caldari
Bulgarian Experienced Crackers
Circle-Of-Two
|
Posted - 2008.10.09 19:25:00 -
[102]
Originally by: CCP Mindstar
Originally by: Mashie Saldana
Originally by: Del ReyII
I'd say its more like LGA771 Wolfdale running on Intel 5000p using 4x 4GB FBDIMMs. Probably like a IBM BladeCenter HS21 .
There are no dual core 3.3GHz Xeon CPUs, the closest is the 3.4GHz Xeon X5272 running the Harpertown core.
I do however stand corrected regarding my first post, there are 16GB memory kits (4x4GB) that will work with P45 based motherboards. So 16GB Wolfdale systems are possible.
Hey guys,
Just to clear this up, there are indeed such things as dual core 3.3ghz Xeon Processors, and the IBM blades that we are using are capable of 16gb memory out of the box.
Ok, I don't want to break my NDA, but you could expect this anyway - IBM will release an updated HS21 blade (dubbed HS22) in early Q1 2009.
The main features will be PCIe 2.0, DDR3 (higher freq.), support for 1600MHz FSB, more DIMM slots than on the HS21. Of course, even now you have HS21 XM (8 DIMMs vs. 4 DIMMs on the normal HS21). Only CCP (and God) knows if the EVE server code would benefit from having 32GB of RAM (vs. 16GB). What about the FSB/Memory having considerably higher bandwidht? You have the Intel X5272 3.4GHz 80W part which has 1600MHz FSB. The DDR3 RAM increases the memory latency, but allows for higher memory frequencies (bandwidth) and lower power consumption. I know that latency is King, but IMHO this should be tested.
Also, the combination of PCIe 2.0 chipset, Intel IOAT enabled Intel Gigabit controllers (the latest, not yet to be found in IBM servers) + latest Windows Server 2003 patches, provide CONSIDERABLY lower CPU utilization, thus freeing up CPU cycles for other tasks. This week we are running tests with Intel 5400 chipset (PCIe 2.0 / 1600MHz FSB) and some OEM Intel NICs and the new Brocade 8Gbps FC HBAs (yes, Brocade finally decided to go to the market with HBAs). So far, wonderful results!
About Nehalem - check out this engineering-sample benchamrks: http://www.anandtech.com/showdoc.aspx?i=3326&p=5
2.93GHz for an engeneering sample is not that bad...Frequency is not always that important. Having the memory controller integrated + 3-channel DDR3 direct-attach memory + considerably faster and bigger caches do some interesting things.
Of course, CCP could not always wait for the latest & greatest, but if you plan to upgrade the whole cluster, you'd better invest in something really future proof. The new Nehalem platform will be Socket compatible with the 32nm derivatives, so CCP could expect much longer life for the hardware ;-).
___________________________________
Science & Diplomacy Manager @ BECKS
The OSS |
bowk
Republic Military School
|
Posted - 2008.10.10 04:04:00 -
[103]
Originally by: CCP Mindstar
SOL Blades - These are the workhorses of Tranquility and are the primary focus of our ongoing work. The cluster is divided across 90 - 100 SOL blades which run 2 nodes each.
* Node - a single EVE server process. This is the lowest level of granularity within the cluster.
Dedicated SOL blade - These are SOL blades that we dedicate to one system only. Systems such as Jita, Motsu and Saila reside on these. They run two nodes like any other SOL blade, however the second node is idle and does not load any solar systems.
With the above do you mean that a "normal" "node", i.e. single eve server thread, runs on average 5000 (=total solarsystems in eve?)/100 / 2 (=2nodes per blade) =25 solarsystems ?
And how are these 25 solarsystems then selected ? like all the solarsystems of up to 3 constellations which are adjacent to each other .. or random solarsystems ? or a combination ?
|
Megan Maynard
Minmatar
17th Minmatar Tactical Wing
|
Posted - 2008.10.10 21:36:00 -
[104]
I'm not sure where the improvements are....
All I know is that every since stackless IO the FW areas between minny and amarr space have been terribly laggy.
It isn't just me, multiple people have noticed this. (Amarr pilots as well.) |
DrAtomic
League of Disgruntled Fast Food Employees
|
Posted - 2008.10.11 11:31:00 -
[105]
Originally by: CCP Lingorm
Originally by: CCP Mindstar
Originally by: DrAtomic
Aaaaanyway, did your wife move to Iceland as well?
While we did "meet" in a bar, I am not entirely sure I remember getting married.... I'll have to double check on that one I think.. Where do you get your information?? 
I think he got the In Crowds mixed.
Yes, my wife, did move to Iceland with me.
Ugh, lol, yes I did mix up the two articles my bad... :)
- - -
Originally by: CCP Wrangler
If you can understand our goal, disagree with our solution and offer a solution that is equal or better your opinion has a better chance of being heard...
|
Darth Sith
Industrials Gone Wild
|
Posted - 2008.10.12 04:00:00 -
[106]
Originally by: CCP Mindstar
Originally by: mechtech
Shouldn't you wait for Nehalem to come out?
Those processors are about 50% faster in server based applications.
This is an interesting question, and one we have talked with our suppliers at length about. Essentially, it comes down to the fact that the Nehalem will not be running at speeds that will be useful for EVE until late next year. It is true that Nehalem makes some good advances in for server applications, however the nature of the current EVE process being mostly single threaded means that going with higher clock speeds and less cores gives us a much greater performance gain.
In the future we may be able to better utilise the Nehalem (or some subsequent technology), however for what EVE needs at this point, we really can't find something that tops the Wolfdale.
Actually, then you haven't recieved the full briefing m8. I work for one of your main suppliers and there is an interesting "feature" in Nahalem which is not really discussed that much that specifically targets your low thread count issue. Mail me ingame if you want some detail . .I cannot discuss it in open forums.
Your signature is too large. Please resize it to a maximum of 400 x 120 with the file size not exceeding 24000 bytes. If you would like further details please mail mods@ccpgames.com - Mitnal |
Alexander Ronay
Crushed Ambitions
|
Posted - 2008.10.12 05:29:00 -
[107]
Edited by: Alexander Ronay on 12/10/2008 05:43:52
Edited by: Alexander Ronay on 12/10/2008 05:30:03
Originally by: Mack Dorgeans
....stuff....
...On an unrelated note, I would like to mention there are other technological hurdles that don't seem to be improving. For instance, when I run the game in windowed mode (premium graphics), I get a significant drop in FPS for no apparent reason, even if I'm not running anything else but the usual background processes (under Vista 64-bit). In fullscreen today I noted FPS in the mid-50s with my current settings while warping. There was no obvious stuttering. As soon as I switch to windowed, that drops by 10-15 points, and obvious stuttering appears constant. When I dock in a Caldari station with those silly spotlights, that number drops another 10 points. Since the fullscreen mode does a pretty poor job alt-tabbing between tasks, I like to run windowed when multitasking, yet even before I run another application or second copy of the client, I'm already seeing a significant performance drop.
Yes, I'll submit this as a bug report, but the fact remains this and other everyday performance issues don't appear to be improving at all while others seem to be improving only enough to encourage players to hit new limits in a never-ending cycle.
This has nothing to do with the eve client and everything to do with the way windows itself works, along with the way graphics are rendered by your hardware. In full screen mode, the Eve client is the only thing being drawn on the screen. In windowed mode, your computer is constantly redrawing the eve client, your desktop, taskbar, icons, and running apps, etc etc. There are also issues concerning window focus, overlays, and so on. Running a graphics intensive app like EVE in a window is always going to take a performance hit when compared to running it in fullscreen.
Anywho, really enjoyed the dev blog. We need more like these  And I've noticed questions about virtualization being raised in this thread. I'd be very interested in hearing whether or not that is a viable option for EVE.
|
IR Scoutar
Caldari
State War Academy
|
Posted - 2008.10.14 13:35:00 -
[108]
Originally by: devblog
The EVE Server application itself (also known as a node) is primarily a CPU intensive process. Due to the nature of the Stackless Python programming methodology chosen for EVE Online, the python component of each node is a single thread, which means it can only ever utilize 1 CPU core at a time
i had to throw up a little when i read this ... then i started laughing quite loudly
it explains so much
|
UnDeRBaLaNcE
The Dead Parrot Shoppe Inc.
|
Posted - 2008.10.15 08:25:00 -
[109]
Expensive stuff, I sent inquiries as a fake business to each maker on price quotes to see how much roughly they cost for old and new. I know the Ramsans 400 are roughly 90,000 USD each. Those server blades are about... 12,000-20,000 USD each. As being in networking related career, im always looking into the future, super computers, with the stackless IO well help CCP push to a super computer.
|
Drake Draconis
Minmatar
|
Posted - 2008.10.15 20:17:00 -
[110]
They should start using clusters and give the nodes the ability to pick up idle CPU timing.
They would only have 1 CPU at anyone time but when things get a little heavy... pick up the second one... and so forth.
Good luck figuring out how to get there tho. (I've seen it done... but EVE is a different ball game)
|
|
UnDeRBaLaNcE
The Dead Parrot Shoppe Inc.
|
Posted - 2008.10.15 23:20:00 -
[111]
So i said i got a quote... here is a ramsans 500 price for a mid level size memory.
GroupSKUQuantityPrice
RamSan 500D-RAMSAN-500-2000 32GB DDR Cache 2 FC-77 (Dual 4-Gb Fibre Channel Link)."$218,000.00
|
Kweel Nakashyn
Minmatar
Aeden
|
Posted - 2008.10.18 11:00:00 -
[112]
+/- 38000 IOPS was the number my ex-bank client pushed.
gj :)
Fetchez la vache !
|
bowk
Republic Military School
|
Posted - 2008.10.18 11:11:00 -
[113]
Originally by: UnDeRBaLaNcE
. Those server blades are about... 12,000-20,000 USD each. As being in networking related career, im always looking into the future, super computers, with the stackless IO well help CCP push to a super computer.
nope on the server blades price.
A blade chassis goes around 20000 USD, with good discount, blades go around, 2000-4000 USD, with good discount.
Standard you have around 8 blades per blade chassis. (16 blades per blade chassis is now also possible..)
There is a reason, that companies goes for blades.. and thats because its so cheap compared to "standard" server prices.
|
Ivan Davis
Minmatar
Luna Free State
|
Posted - 2008.10.19 19:34:00 -
[114]
Originally by: Craleo
Did you guys ever test VMWare for the nodes? You can take a four core node, install VMWare, install 1 instance of windows, and create one virtual CPU for that server. Then assign all four cores to that one virtual server, which only sees 1 CPU.
Would that work?
Nope.
VMWare doesn't work that way -- it can't "magically multithread" single-threaded code. Doing what you describe would result in three almost 100% idle cores (and only "almost" idle" because of VMWare overhead), and no benefit.
Using VMotion to live-migrate a node to a higher-end blade would be nice, but you can't VMotion from AMD to Intel or vice-versa, so that really wouldn't help right now.
What it would help for is if a small subset of nodes had the same arch type (for VMotion) as the rest, but much higher per-core performance, and more memory. Then when a system got loaded, it could automatically, and LIVE, migrate to a higher-end node. That sort of thing could be quite useful if the cluster ends up in a never-ending rolling upgrade cycle. |
Hugh Ruka
Exploratio et Industria Morispatia
|
Posted - 2008.10.21 07:02:00 -
[115]
Originally by: bowk
Originally by: UnDeRBaLaNcE
. Those server blades are about... 12,000-20,000 USD each. As being in networking related career, im always looking into the future, super computers, with the stackless IO well help CCP push to a super computer.
nope on the server blades price.
A blade chassis goes around 20000 USD, with good discount, blades go around, 2000-4000 USD, with good discount.
Standard you have around 8 blades per blade chassis. (16 blades per blade chassis is now also possible..)
There is a reason, that companies goes for blades.. and thats because its so cheap compared to "standard" server prices.
depends on the datacenter ... high computing density also goes hand in hand with high thermal and power density ... so you can't cram many high performance blades into one enclosure without high costs for cooling and power :-)
and blades offer no redundancy without a virtualisation/migration software layer. then you get the added liability of a single enclosure (SPOF).
blades are nice for web hosting centers, where you can cram lots of them into racks as they don't experience much load. but any demanding application will lower your computing density because of thermal and power :-)
the best thing still is a trucluster system with memory channel and other things ... hard to beat in many ways ...
CCPs largest bottleneck is the single process system handling ... there is no granularity to offload different grids to different CPUs. So once a grid takes up one CPU to 100% (a large fleet battle) the remaining system won't help in any way basicaly having idle CPUs or in the other case all the other systems are crammed on the remaining CPUs on the blade.
it also seems that the infiniband thing will only have a marginal impact on performance without an architectural review/rewrite to a multithreaded system. It will allow CCP to migrate busy systems on the fly to free servers, but there will be still a limit of the singlethreaded nature of their code.
--- SIG ---
CSM: your support is needed ! |
bowk
Republic Military School
|
Posted - 2008.10.22 18:10:00 -
[116]
Originally by: Hugh Ruka
depends on the datacenter ... high computing density also goes hand in hand with high thermal and power density ... so you can't cram many high performance blades into one enclosure without high costs for cooling and power :-)
and blades offer no redundancy without a virtualisation/migration software layer. then you get the added liability of a single enclosure (SPOF).
blades are nice for web hosting centers, where you can cram lots of them into racks as they don't experience much load. but any demanding application will lower your computing density because of thermal and power :-)
the best thing still is a trucluster system with memory channel and other things ... hard to beat in many ways ...
CCPs largest bottleneck is the single process system handling ... there is no granularity to offload different grids to different CPUs. So once a grid takes up one CPU to 100% (a large fleet battle) the remaining system won't help in any way basicaly having idle CPUs or in the other case all the other systems are crammed on the remaining CPUs on the blade.
it also seems that the infiniband thing will only have a marginal impact on performance without an architectural review/rewrite to a multithreaded system. It will allow CCP to migrate busy systems on the fly to free servers, but there will be still a limit of the singlethreaded nature of their code.
Oh noes, a tru64 guy. ;)
But yes, a "scale up infrastructure", where all processes run on 1 server, has a lot of advantages, over what ccp now has, i.e. a "scale out architecture", where processes run on a lot of different servers.
However due to the fact that ccp choose from the beginning, for the windows 32bit OS, left them with no choice then to go for a x86 architecture.
And the x86 architecture, meant in the beginning no more then 8/16? cpu per server, so way to low to be able to run "eve". Thus the going for the blade architecture.
For the rest, I also think that infiniband will not gain them much, I would certainly go for the probably much cheaper ethernet equivalent.
PS. Blade chassis, come now indeed with 16 half height blades instead of 8 half height blades.
PPS. I wouldnt worry about power/ventilation consumption with blades. Power certainly isnt a problem, and for ventilation, enough ventilation will get the job done. I think they also now have racks with watercooling.
PPPS. Hopefully ccp will stay far away from virtualisation software like vmware or the microsoft / linux equivalent thing. The features look nice, but the "layers" that this add to the "solution", greatly increase the complexity of maintaining everything and therefor is better to be left alone.
|
|
| Pages: 1 2 3 [4] :: one page |
| First page | Previous page | Next page | Last page |