Pages: 1 2 [3] 4 5 :: one page |
|
Author |
Thread Statistics | Show CCP posts - 14 post(s) |
Ban Doga
|
Posted - 2011.03.24 20:54:00 -
[61]
Originally by: Ishina Fel
Originally by: Ban Doga
Also found the original statement I was referring to:
Originally by: CCP Atlas We only have a single database and it's easier to scale that up than the sol nodes and we're already ahead of the curve in terms of what the DB can deliver. We do cache very aggressively on the server though and consolidating these character node calls onto a half a dozen nodes rather than servicing them throughout the cluster does remove a bit of the DB load since we get more cache hits, but like I said, the DB is not a big issue in this regard today.
http://www.eveonline.com/ingameboard.asp?a=topic&threadID=1371750&page=2#39
That quote is from August 2010... I'm sure it was compeltely true back then. But since then, they released Incursions, activated resource depletion on planets, overhauled the whole inventory system, and did other code improvements... I'm pretty sure that it is especially the latter two things that trouble the database.
Imagine - they just released a blog where they state that they can allow for Jita's maximum population to grow by over 1000 additional people, because the new efficient inventory code allows the node CPU to handle that many more inventory operations per second. But where do all these inventory operations go? Well, they hit the database. And now there's going to be a whole lot more of them in the same amount of time. Not only in Jita, but in every system that saves CPU cycles due to this coding change.
So the very database that ended up sitting around bored because TQ couldn't generate enough requests to saturate it, suddenly had to scramble to keep up, approaching its limits. So an upgrade made sense.
Well August 2010 is not that much time IMO. 7 months won't totally change the landscape. Yes, they released Incursions and a lot of other changes but I don't expect the overall ratio "stress for the CPU" vs "stress for the DB" to change radically in such a short time.
Furthermore there has been no statement that the situation is suddenly reversed and the DB needed an upgrade.
On top of that CCP Valar basically confirmed that an improvement in perceived performance is not probable.
Originally by: Ishina Fel (This is of course pure guesswork, I have no idea what really happened. I only know that more often than not when you improve one part of a complex system, you end up stressing a different part without even meaning to.)
Yes, it is all guesswork we're doing here. Good to keep that in mind. It is also true that there is always a limiting factor and addressing one will usually make something else a limiting factor.
However it is also a good idea to stick to the information actually available and not guess too much. If the last info was "DB is not stressed" and no one said "this has changed" why assume it has changed? As an excercise: try assuming it has changed so that the DB is even less stressed than before - after all no one said this would be the case either.
|
Ishina Fel
Caldari Terra Incognita Intrepid Crossing
|
Posted - 2011.03.24 20:55:00 -
[62]
Originally by: Shandir I suspect they already have to put a *lot* of effort into cooling, although this is an idea for reinforced nodes. If CCP currently is looking into multicore as they cannot push single-core processing as much as they'd like - what is stopping you from taking the highest clock-speed rating CPU commercially available, and then overclock it under heavy cooling for the max performance reinforced nodes?
Probably the fact that they will lose all warranty support from their hardware supplier for anywhere from a single machine to the entire cluster, and depending on the terms of use, might be found guilty of breach of contract?
Then there's stability concerns. You do not run servers outside specifications, period. If you overclocked desktop CPU botches a single calculation a day because it is at its limit, you probably won't even see anything happen at all for months to come. If each of TQ's CPUs botches a single calculation a day, then the server will become borderline inoperable - and CCP will have to foot every single cent of the bill itself.
Never, ever run a production environment that your entire business depends on outside of specifications. That's like building a nuclear power plant in Japan and forgetting the flood walls.
And despite how utterly minimal the chances of the worst case scenario were, we all know how THAT ended up, don't we. - Signature? What signature? |
Tinkanium
|
Posted - 2011.03.24 20:58:00 -
[63]
Originally by: Ishina Fel Imagine - they just released a blog where they state that they can allow for Jita's maximum population to grow by over 1000 additional people, because the new efficient inventory code allows the node CPU to handle that many more inventory operations per second. But where do all these inventory operations go? Well, they hit the database. And now there's going to be a whole lot more of them in the same amount of time. Not only in Jita, but in every system that saves CPU cycles due to this coding change.
So the very database that ended up sitting around bored because TQ couldn't generate enough requests to saturate it, suddenly had to scramble to keep up, approaching its limits. So an upgrade made sense.
(This is of course pure guesswork, I have no idea what really happened. I only know that more often than not when you improve one part of a complex system, you end up stressing a different part without even meaning to.)]
I was planning to drip sarcasm all over this thread in a post about how I witnessed performance go down in light of hardware increases and code optimization. Thankfully in a rare fit of wisdom I read the entire thread before posting to avoid looking like a total ass. I suspect that Ishina Fel is quite correct. Immediately after the patch dealing with the new inventorying system (designed to nerf the soul-crushing lag caused by missile swarms) I personally witnessed a very noticeable hit to performance when loading new grids, entering systems, initiating jumps, docking/undocking, loading planetary interaction (sometimes a crash, but maybe that's client hardware specific) and most annoyingly, when selecting something from the overview and then trying to click buttons in the Selected Item window. At one point I was even presented with the buttons for a ship after clicking a stargate. Instead of being able to warp to this gate, I was presented with greyed out approach and warpto buttons, as well as a button for enter ship. Interesting choice when it comes to gates.
I think this will fix it. I am looking forward to experiencing gameplay on the new hardware. Especially the backend stuff for all of you server-jockeys. Fewer extended downtimes is better! It's already a rather impressively short downtime when compared to other online games. It's certainly not five-nines but let's face it games don't seem to have a hope in hell of achieving that any time soon.
To end on a funny note in response to the post regarding the use of profits to improve the service which garnered said financial gains, looks like the sandbox is not the only thing that CCP is unique for!
And yes. I am insanely jealous of the shiny new hardware you guys are playing with!
|
Tinkanium
|
Posted - 2011.03.24 21:06:00 -
[64]
Originally by: Ban Doga It is also true that there is always a limiting factor and addressing one will usually make something else a limiting factor.
Yeah man. In fact based on what you yourself said you're already wrong. There's always a limiting factor. If you fix that, and it is no longer the limiting factor, then by the logic that there's always a limiting factor something else must become the limiting factor.
Even in the dev blogs there's always someone. Seriously. Incredibly long post containing nothing but quotes and seemingly intelligent logic that basically amounts to nothing. Please do save us all some time and just read along k?
|
Korerin Mayul
Amarr
|
Posted - 2011.03.24 21:19:00 -
[65]
Resident EVE Storage nerd checking in with some questions, comments and suggestions:
...so you have four 8g ports per server, rolled into 2 trunks of 16g? - sweet!! what is managing the host level trunking? I havent seen this done in anger before *or* you have two 8g ports per server, rolled into 1 trunk of 16g - less sweet, in this case how do you handle link failure and are you still doing two paths to two fabrics?
Might i recommend a cross-grade to Brocade HBSs? ÆRumour has itÆ That they will be first to market with ratified 16gb FSPs / HBAs / Switches û super plasma fast if you want on that shizzell early.
Also, as a shot in the dark, go and take a look at the EMC Greenplum appliance û its a weird use-case but a big sack of ETL capabilities may be useful for something up there. A fully kitted, massively parallel VSP *might* outperform that IBM box, but im only mentioning it so i donÆt sound like an EMC shill :P
Additionally, be really careful with thin provisioning, I speek form bitter, bitter experience that using that stuff in prod will cause many tears and late night clickeyfests.
Finally, if you have any more money spare, you could give it to me, or buy yourselves some Data Domain kit, those boxes can handle 3gb/s sustained throughput per controller. If that doesnÆt back your face right up, i donÆt know what will.
|
Hawk TT
Caldari Bulgarian Experienced Crackers RED Citizens
|
Posted - 2011.03.24 21:27:00 -
[66]
CCP, now you have to look at Intel Xeon X5698 Dual-Core 4.4GHz and X569X @ 4.66GHz...
I suppose those would be nice match for the SOL nodes... ___________________________________ BECKS
|
Akita T
Caldari Navy Volunteer Task Force
|
Posted - 2011.03.24 21:28:00 -
[67]
Always nice to see YARRRrrrrdware p0rn blogs _
Make ISK||Build||React||1k papercuts _
|
Commander TGK
Gallente The Deep Space Armada
|
Posted - 2011.03.24 21:36:00 -
[68]
Now that is our subscriber money at work! Excellent job CCP! Those are gigantic improvements over the old.
|
Vaerah Vahrokha
Minmatar Vahrokh Consulting
|
Posted - 2011.03.24 21:43:00 -
[69]
Quote:
If I had to be told
"you are going to work in the server room"
I swear I would take a sleepingbag, one of those funny inch-thick mattress things, a perpetual coffee machine, a portable shower and never leave. Just being around that kind of awesome is enough
I don't get why a server room would be so exciting.
I have worked for 6 years next and inside a datacenter with loads of fiber optic connections, super SANs, 7-8 different OSes and databases, web servers...
... after the first days of amusement at retina scans and similar, it gets old quick.
Best use? We stored good seasoned wine under the modular floor tiles, the air conditioning kept them at optimal temperature.
Another time I worked in and out of a controlled air (I don't know the exact English word) datacenter and the only thing that got me was how boring is to constantly dress the white suit and do that cleaning shower.
Auditing | Research | 3rd Party | Collateral Holding | EvE RL Charity |
Kalia Masaer
Amarr Border Defense Consortium
|
Posted - 2011.03.24 21:45:00 -
[70]
While I greatly appreciate the effort and cost of the hardware being installed to reduce lag, it will not be capable of staying ahead of players who will simply bring bigger blobs until the new hardware has reached its limit and crashes just like they currently do, it will just take more players.
As important though is a way to make it so simply bringing a bigger blob is not the only way to win. It is sadly where I feel we are being let down. The battle to make the software and hardware more capable seems to have people working head over heels to do the impossible but really it is a pointless gestures as ever bit of gain made is gone within weeks of as more people come to fights.
I'd really like to see some development effort spent on moving forward with a rework of SOV. Sad as it maybe the old tower spamming sov method was in many ways better at spreading out battles. CCP needs to give the guys fighting lag a break by making it so we the players do not keep taxing the system to the limit, which we will do so long as more is better.
|
|
Justin Cody
Caldari Instant Annihilation Everto Rex Regis
|
Posted - 2011.03.24 22:11:00 -
[71]
so much beautiful YARRRDWARE!!! Can I haz one server? Remind people that profit is the difference between revenue and expense. This makes you look smart. Scott Adams
|
Patient 2428190
DEGRREE'Fo'FREE Internet Business School
|
Posted - 2011.03.24 22:14:00 -
[72]
Well I'll be damned, a tech blog that isn't "HURRRR WE THINK ITS WURKIN LOL"
Very nice, very sexy ...Then when you stopped to think about it. All you really said was Lalala. |
Svizac Marmotov
Caldari
|
Posted - 2011.03.24 22:50:00 -
[73]
Is Tranquility going to become self-aware?
|
ModeratedToSilence
|
Posted - 2011.03.24 23:05:00 -
[74]
Originally by: CCP Yokai
Originally by: Marconus Orion I'm not very tech savy so for us laymen non-super geeks, please using a car reference, what is the comparison?
We going from a Fiesta to a Mustang?
We had a Ferrari 1985 288 GTO
We have a Ferrari 2011 599 GTO
Both great... both fast... the new one is just magic.
make sense?
Attempting to align computer technology development with automotive development is stupidity. Computer hardware tends to double in its basic attributes every 18-24 months. If we had the same development in motor vehicles we would not be playing internet spaceships - we would be in real life spaceships.
|
Prince aikka
Grand River Trades Intrepid Crossing
|
Posted - 2011.03.24 23:27:00 -
[75]
I got wood reading that. *starts printer for a bathroom copy*
|
Mihara Shiharu
|
Posted - 2011.03.24 23:46:00 -
[76]
Tech **** indeed!!!!!!
|
GRIEV3R
Gallente Galactic Defence Consortium STR8NGE BREW
|
Posted - 2011.03.24 23:48:00 -
[77]
Awesome dev blog. My nerd boner is the size of an Avatar. CCP wins like Charlie Sheen.
|
Karajishi
Caldari Caldari Provisions
|
Posted - 2011.03.25 00:04:00 -
[78]
Great job on the upgrades, looks sweeeet! Loving these blogs to, a great insight into how you guys do things. Keep it up CCP
|
Dawn Harbinger
|
Posted - 2011.03.25 00:50:00 -
[79]
If I understood half the things in this blog I'd have a much cooler job that I do now ;)
Sounds epic though, hail our robot overlords!!
WINNING!
|
LeeBaak
|
Posted - 2011.03.25 01:00:00 -
[80]
Originally by: Andrea Griffin So, how much did this all cost (roughly)?
Curious about this, too.
Also, since you won't be needing all of that "old junk" anymore, feel free to ship it my way...it's so out-dated...but I'm sure I can find a use for it.
|
|
Jason Edwards
Autistic Sharks Test Alliance Please Ignore
|
Posted - 2011.03.25 01:19:00 -
[81]
Meanwhile my most important production server has 256mb ram, freebsd 4, 2 core xeon, 80 gig hd.
------------------------ To make a megathron from scratch, you must first invent the eve universe.
|
MidKnight Reign
|
Posted - 2011.03.25 01:21:00 -
[82]
Edited by: MidKnight Reign on 25/03/2011 01:24:05 Edited by: MidKnight Reign on 25/03/2011 01:23:22 Edited by: MidKnight Reign on 25/03/2011 01:22:22 Win upgrade is win! Also hardcore tech **** is hardcore. (Kosher in a dev blog but not on the forum? Oh, I see how it is...)
Congratulations on subduing the dreaded disk I/O bottleneck (for now), as well as royally beating the snot out of the approaching CPU bottleneck.
Remember kids: More threads > more cycles Would you rather do a few tasks simultaneously at 3Ghz or many tasks at 2Ghz? Or rather, would you rather have a few crazy fast turrets or a crazy number of decently fast turrets?
|
Lev Aeris
b.b.k Fidelas Constans
|
Posted - 2011.03.25 01:58:00 -
[83]
Will this new hardware make Team Gridlock more successful?
Cool blog. Thanks for sharing your new tool specs with us.
|
Wollari
Phoenix Industries
|
Posted - 2011.03.25 02:13:00 -
[84]
Can I've your old hardware, pleeeeease :-)
|
freshspree
|
Posted - 2011.03.25 02:17:00 -
[85]
I lold. Anyway, let's see the difference when I log on today.
|
Naomi Wildfire
Amarr Men Who Stare At Gates
|
Posted - 2011.03.25 02:23:00 -
[86]
Can i have your old suff? :(
|
ivar R'dhak
Minmatar
|
Posted - 2011.03.25 03:31:00 -
[87]
Edited by: ivar R''dhak on 25/03/2011 03:31:23 Niiice. Yardware p÷rn. Ooh .. I¦m getting all tingly, and, uh ..
THERE ya go. ______________ Mal-¦Appears we got here just in a nick of time. What does that make us?¦ Zoe-`Big damn heroes, sir.` Mal-¦Aint we just.¦ |
Essenz
|
Posted - 2011.03.25 03:44:00 -
[88]
Only 2 words @ CCP: HELL YEAH! |
Dr BattleSmith
PAX Interstellar Services
|
Posted - 2011.03.25 04:39:00 -
[89]
Quote: One day we woke up and our super cool database servers... well... weren't super cool anymore.
ROFL! They actually once considered MS SQL to be "super cool"? LOL
|
Sith LordXXX
|
Posted - 2011.03.25 05:28:00 -
[90]
Edited by: Sith LordXXX on 25/03/2011 05:35:33 Edited by: Sith LordXXX on 25/03/2011 05:30:49 Opteron 6174's would have been a much better upgrade as apose to 8 fake cores in the intel and intel costing twice as much going for $2400, the AMD's have 12 real cores giving far more processing power then a intel chip with only 8 real cores with a marketing gimmic of 8 fake cores.
16 threads don't equal twice the processing power. The performance boost 2 threads give is fake. Like smoke and mirrors you can't get power from something that isn't there. As apose to 12 real physical cores with 12 threads. Your getting real world performance for half the cost. Means you could have used 4 socket boards to get 48 real cores of power for the same price as a dual socket intel. Intel doesn't have quad socket boards on newegg, amd does, for 12 real core cpus. Thats more then 50% the power then a intel system. Intel is a joke. AMD beats intel in every way in the server area its physical fact.
If anybody thinks 16 cores with 16 fake cores can beat even 24 real cores they get a for being derp, and 48 real cores impossible intel can even compete with, if you think so, you get a double... Also price per performance ratio amd wins by 50%. Thats madness. AMD is the real smart choice for real processing power. Wow intresting intel's 8 core cpu's run at 130w and amd's 12 core cpu's run at 115w. Not only does amd have more power, they use less power consumption with more cores.
|
|
|
|
|
Pages: 1 2 [3] 4 5 :: one page |
First page | Previous page | Next page | Last page |