Pages: 1 2 3 4 [5] 6 :: one page |
|
Author |
Thread Statistics | Show CCP posts - 13 post(s) |
|
CCP Claw
|
Posted - 2009.02.17 18:56:00 -
[121]
Originally by: MotherMoon
sisi?
For various different things actually - don't forget that these are still super awesome blades, just not quite as super awesome as the newer ones; there's always a use for lots of super awesome blades :)
|
|
Lakia
|
Posted - 2009.02.17 19:31:00 -
[122]
I would like to see how much power you will be saving using the RamSan-500 vs the DS4800. |
Finnroth
Caldari The Guardian Agency Guardian Federation
|
Posted - 2009.02.17 20:04:00 -
[123]
I bet they're collecting spare parts for the comming emo-vamp mmo
|
Chuy Fatt
Homicidal Minds
|
Posted - 2009.02.17 21:34:00 -
[124]
Cisco 7606-S is not a bad idea. Its always good to look into Network lag as well Thanks
|
Desiree Mabuto
Caldari Jagdkommando RAZOR Alliance
|
Posted - 2009.02.17 22:00:00 -
[125]
Edited by: Desiree Mabuto on 17/02/2009 22:01:38
Originally by: CCP Mindstar Oh yes we have plans to go to 2008 Windows / SQL also..... all in good time
If it doesnt take too long to answer - why do you use Windows, not Unix operating systems? It's not like "eww guys, you need to change that asap, that's uncool" I am just asking as a matter of personal ( technical ) interest as I am running a own quite busy network on unix systems
/me is jealously looking at the RamSAN500, wished to have one but not our budget
|
evefour
|
Posted - 2009.02.17 22:06:00 -
[126]
Originally by: Cergorach A question about the Ramsan 400 vs. 500 performance.
Here: http://www.superssd.com/products_sub.htm It looks like that the 400 is faster then the 500, Is that true, or am I missing something? Have you guys start nagging for a Ramsan 5000 ? *grins*
Is it necessary to upgrade the other 104 nodes to new hardware (in the short term)? And please post how the nehlams perform (and when you upgrade the Jita server again with that new proc ;-)
RamSan-400 CAPACITY, 32 to 128 GB RAM RamSan-500 CAPACITY, 1 to 2 TB Flash It's probably the storage capacity that played a decisive role.
|
Darth Sith
Genbuku. Daisho Syndicate
|
Posted - 2009.02.17 23:30:00 -
[127]
Originally by: Lord Fitz Edited by: Lord Fitz on 17/02/2009 12:29:36
'cooling' Is not a problem in a datacenter, they are built to handle rows upon rows of blade servers. Cooling a single mainframe if one exists on the other hand, is far more difficult.
.
Actually it is the exact opposite m8. It is the rows of med / high density systems that cause the most headache in the datacenter.
|
Ki Sol
|
Posted - 2009.02.18 00:00:00 -
[128]
First let me say I think EVE is an AWSOME game. However I have some major concerns about the choice of upgrade hardware and it's abilities. No one wants lag or downtime but if I had to choose I would tolorate a little downtime to avoid lag during the game. Lag=Pod Killed!
Let's look at Specs of your current hardware.
RamSan-400: I/Os Per Second 400,000 Capacity 32-128 GB Bandwidth 3000 MB (3GB) per second
...and now your new hardware.
RamSan-500
I/Os Per Second 100,000 Capacity 1-2 TB of Flash RAID 16-64 GB of DDR Cache Bandwidth 2 GB per second
I know flash drives are fast (internally to the raid) and doing backups and updates MAY be faster but if you are going to reduce your Input/Output from 400k to 100k and your bandwidth from 3GB to 2GB, How is this going to REDUCE lag?!
2TB of storage, big woop! If you can not get the data in and out of the box any faster then you are not reducing lag, you're installing more lag.
In my opinion you should buy more RamSan-400's or something with better bandwidth and I/O speeds and use the cluster of them to support your online database. Then you could still use your RamSan-500 to backup your entire cluster. Also by updating your database software you should be able to do these backups in realtime so you would only require down time to do server upgrades.
Please reconsider this "upgrade". As a hardware tech myself I understand there is a balance between storage capacity, speed and price. With a single 2TB storage raid running on only 300 watts of power you save money and space. But these "savings" sacrifice throughput speed which equates to LAG. As a growing MMORPG with the retail release of your game coming soon the last thing you want to do is sacrifice I/O speed and Bandwidth to save a few bucks up front.
Technical Specs: RamSan-400 RamSan-500
|
Matthew
Caldari BloodStar Technologies
|
Posted - 2009.02.18 00:10:00 -
[129]
Originally by: Ki Sol I know flash drives are fast (internally to the raid) and doing backups and updates MAY be faster but if you are going to reduce your Input/Output from 400k to 100k and your bandwidth from 3GB to 2GB, How is this going to REDUCE lag?!
Try reading the blog again. The existing two RAMSAN 400 units that they already have are still there. These will continue to do what they are already doing, holding the busiest tables on the fastest possible storage.
The data being moved onto the new RAMSAN 500 is currently being held on a DS4800 fibre channel disk array, which is considerably slower than the RAMSAN 500.
Yes, it would be even faster still if they put it all onto RAMSAN 400's, but you'd need 16 400's to offer the same storage capacity as the 500. Which is unlikely to be practical in terms of cost or datacentre space, at least until Eve grows enough to actually need that approach. ------- There is no magic Wand of Fixing, and it is not powered by forum whines. |
Annatar
The Galactic Empire Executive Outcomes
|
Posted - 2009.02.18 00:54:00 -
[130]
A Couple of Pictures of the new Stuff would be very cool.
-------------------------------------------- Never argue with an Idoit, they will drag you down to their Level and beat you with experience. |
|
Tasmanian Devil
|
Posted - 2009.02.18 01:13:00 -
[131]
Quote: Our network engineers are busily finalizing the plans on some Cisco 7606-S routers which will be placed as the primary routing points within our game server network, and they are also revising our Internet peering strategy to expose Tranquility to our customers in the least number of hops around the globe. Lower ping times mean faster pew pew!
Does this mean that the "Socket Closed" and frequent disconnects will be fixed?
|
Lord Fitz
Project Amargosa
|
Posted - 2009.02.18 02:00:00 -
[132]
Edited by: Lord Fitz on 18/02/2009 02:01:10
Originally by: Ephidaurus lol, you are sure you know what I'm talking about ? A Mainframe is never a "single" system. Have a look at IBM System z10 But if you like you can go for a Parallel Sysplex as well. However, point is that nearly 90% of all banking stuff is done by IBM Mainframes. So I guess that should be enough to show that its not a "play" system. IBM itself even says that MTBF (Mean Time Between Failure) is about 40 years. You wont get a better one on large scale computing. And btw. IBM is living only because of that point, at least in the Mainframe section.
Yes, I know, I've played with a nice $1,000,000 IBM mainframe used for banking. Great throughput but only when doing very simple calculations. 40 Year MTBF is nice, but that's the MTBF, not a guarantee, when they fail it takes several weeks to get a new one. And again, the throughput is nothing compared to the existing eve cluster, they don't need more throughput from one machine, at the cost of the whole cluster.
Originally by: Gamer4liff Forgive my ignorance but how did the server's database get to only be 1 TB? That seems rather small to me, given the enormous amount of stuff just per character. Bookmarks, items, transaction history, item data, that sort of thing.
Lots of optimisation ?
It does seem small, I recently worked on a poorly written app with just 20,000 users, had 1 TB of data. Even so being poorly written was still less complex than Eve, so I'm guessing things like eve-mails are not part of the main database on the RAMSAN ?
|
Chainsaw Plankton
IDLE GUNS IDLE EMPIRE
|
Posted - 2009.02.18 02:05:00 -
[133]
Originally by: CCP Claw
Originally by: MotherMoon
sisi?
For various different things actually - don't forget that these are still super awesome blades, just not quite as super awesome as the newer ones; there's always a use for lots of super awesome blades :)
I want one, can I has?
|
Komen
Gallente Aliastra
|
Posted - 2009.02.18 04:01:00 -
[134]
I'd say thank you for all you've done over the years to improve Eve's hardware, but frankly i'm downright jealous of how pwn all this hardware is.
Curse you, CCP! Cuuuuurse yooooou!
And now, my sig:
|
Taisuke Black
|
Posted - 2009.02.18 04:52:00 -
[135]
It's great that you guys are improving hardware. I have request:
Taru and its surrounding systems.
I've been L4 missioning here for a while, and it's a very busy area, what with Traroh the L4 Q20 agent.
The lag is getting ATROCIOUS. Wrecks spawn up to 90 km away from where the ship died, missiles do no damage at 500m because the position of the enemy ship is wrong. I can salvage from a wreck 50km away if I was in range of its proper position and I start within 10 seconds of its death.
I'm inclined to blame my wireless router, but even when I plug straight into it the result doesn't change. I'll be upgrading the thing soon, I think, but I'm pretty certain the majority of lag is on your end, since it's happened to other people missioning with me.
Thank you for your unending efforts to fight these things.
|
|
CCP Mindstar
|
Posted - 2009.02.18 12:45:00 -
[136]
Originally by: Ki Sol I know flash drives are fast (internally to the raid) and doing backups and updates MAY be faster but if you are going to reduce your Input/Output from 400k to 100k and your bandwidth from 3GB to 2GB, How is this going to REDUCE lag?!
We are adding have added the RamSan500 alongside our two existing RamSan400's. The 500 is replacing a fiber channel disk array, and our usage of the 400s will remain unchanged.
Instead of using 2Tb of plain old hard disks, we now have 2Tb of Flash Raid -- |
|
huang lee
|
Posted - 2009.02.18 15:19:00 -
[137]
hey im currently studying for a cert III in I.T (aussie for technical school) and i still have a lot to learn about this stuff... however eventually i will and hope to work for CCP one day in the future the stuff you guys are talking sounds really exciting btw
|
Lou Tiang
|
Posted - 2009.02.18 17:30:00 -
[138]
Edited by: Lou Tiang on 18/02/2009 17:30:41 We need pictures! Good job keep it up!
Is it me or is undocking and jumpst a little faster? Or maybe it's just wishful thinking.
|
sukmanobov
Minmatar Republic Military School
|
Posted - 2009.02.18 19:45:00 -
[139]
Originally by: Hugh Ruka Edited by: Hugh Ruka on 16/02/2009 16:46:23
Originally by: Washell Olivaw
Originally by: Hugh Ruka Even the RAMsan specs are low. You should upgrade to an enterprise class storage solution that can grow with the load.
But meh. You are running on Windows after all ...
Compare random I/Os per second on the RAMsan400 and 500 vs enterprise class storage in the same size range. Then come back and say you're unimpressed.
I guess for the small size of the EVE database, enterprise class arrays are not financialy viable.
1.1TB isn't large
|
Tillmen
LFC Executive Outcomes
|
Posted - 2009.02.18 20:31:00 -
[140]
Originally by: CCP Mindstar
Originally by: Ki Sol I know flash drives are fast (internally to the raid) and doing backups and updates MAY be faster but if you are going to reduce your Input/Output from 400k to 100k and your bandwidth from 3GB to 2GB, How is this going to REDUCE lag?!
We are adding have added the RamSan500 alongside our two existing RamSan400's. The 500 is replacing a fiber channel disk array, and our usage of the 400s will remain unchanged.
Instead of using 2Tb of plain old hard disks, we now have 2Tb of Flash Raid
Do me a favor and never buy IBM storage again. They are awful compared to what is out there.
|
|
Tiger's Spirit
|
Posted - 2009.02.18 20:46:00 -
[141]
The lag reduced ? I think not, or the new ramsan will used for the new wormholes ? The systems still laggy.
|
Regat Kozovv
Caldari Deep Core Mining Inc.
|
Posted - 2009.02.18 20:52:00 -
[142]
Originally by: Matthew
Originally by: Ki Sol I know flash drives are fast (internally to the raid) and doing backups and updates MAY be faster but if you are going to reduce your Input/Output from 400k to 100k and your bandwidth from 3GB to 2GB, How is this going to REDUCE lag?!
Try reading the blog again. The existing two RAMSAN 400 units that they already have are still there. These will continue to do what they are already doing, holding the busiest tables on the fastest possible storage.
The data being moved onto the new RAMSAN 500 is currently being held on a DS4800 fibre channel disk array, which is considerably slower than the RAMSAN 500.
Yes, it would be even faster still if they put it all onto RAMSAN 400's, but you'd need 16 400's to offer the same storage capacity as the 500. Which is unlikely to be practical in terms of cost or datacentre space, at least until Eve grows enough to actually need that approach.
Right. And someone correct me if I'm wrong, but it was my understanding from the CCP presentations that not all tables in the database are accessed as frequently as others (character data vs. market data). So while yes, the 500 is slower, the bulk data stored on it isn't under the same demands as the high transaction data that resides on the 400s. You could certainly get 400s for everything, but this wouldn't be reasonable from a cost/benefit standpoint, as Matthew states.
|
Tillmen
LFC Executive Outcomes
|
Posted - 2009.02.18 21:06:00 -
[143]
Edited by: Tillmen on 18/02/2009 21:11:51
Originally by: Ki Sol First let me say I think EVE is an AWSOME game. However I have some major concerns about the choice of upgrade hardware and it's abilities. No one wants lag or downtime but if I had to choose I would tolorate a little downtime to avoid lag during the game. Lag=Pod Killed!
Let's look at Specs of your current hardware.
RamSan-400: I/Os Per Second 400,000 Capacity 32-128 GB Bandwidth 3000 MB (3GB) per second
...and now your new hardware.
RamSan-500
I/Os Per Second 100,000 Capacity 1-2 TB of Flash RAID 16-64 GB of DDR Cache Bandwidth 2 GB per second
I know flash drives are fast (internally to the raid) and doing backups and updates MAY be faster but if you are going to reduce your Input/Output from 400k to 100k and your bandwidth from 3GB to 2GB, How is this going to REDUCE lag?!
2TB of storage, big woop! If you can not get the data in and out of the box any faster then you are not reducing lag, you're installing more lag.
In my opinion you should buy more RamSan-400's or something with better bandwidth and I/O speeds and use the cluster of them to support your online database. Then you could still use your RamSan-500 to backup your entire cluster. Also by updating your database software you should be able to do these backups in realtime so you would only require down time to do server upgrades.
Please reconsider this "upgrade". As a hardware tech myself I understand there is a balance between storage capacity, speed and price. With a single 2TB storage raid running on only 300 watts of power you save money and space. But these "savings" sacrifice throughput speed which equates to LAG. As a growing MMORPG with the retail release of your game coming soon the last thing you want to do is sacrifice I/O speed and Bandwidth to save a few bucks up front.
Technical Specs: RamSan-400 RamSan-500
You are making some basic mistakes when it comes to storage Ki Sol.
First - The bandwidth of 3GB/sec is from the cache to the flash. When the hosts are accessing the data, they are accessing it from cache, which is faster then the flash. SANS use prefetch algorithms which *should* have the data tracks that will be used, in the cache before they are needed. There will be no need to "wait" for the data to be pulled from the flash to the cache. Also if the same data/file/table is being changed repeatedly, it is done so in cache and not on the flash, hence the data does not need to traverse the backend (3GB/sec speed) of the SAN. Random/non-sequential reads will more then likely have to traverse the backend but then that data will be in cache for use, provided it is not been destaged from cache.
Second - More IOPS does not mean less lag/speed or more lag/speed for that matter. You need to look at the cluster size in order to make a comparison between the 2 SANS. For example - The 400 does 400k IOPS, now if the cluster size of the storage is 4k the SAN can move 1600k of data per second (400,000 x 4k = 1,600,000k). The 500 does 100k IOPS, now if the cluster size is 16k, it will move 1600k of data per second (100,000 x 16k = 1,600,000). Both SANS have different max IOPS but are moving the same amount of data per second.
Third - When looking to draw performance parallels, you need to look at the entire fabric from HBA to Storage Subsystems. This includes the design of the fabric (number of switches/ISL's/port speeds), the redundancy (multiple paths to the storage) as well as the layout of the SAN itself (RAID config, number of controllers, number of RAID groups, number of luns, how those luns are spread across the RAID groups and how the RAID groups are spread across the backend controllers).
|
Furion Riddick
|
Posted - 2009.02.18 22:32:00 -
[144]
Once you have exhausted your performance on SQL Server you can then promote to ORACLE 11g RAC with your hardware, Infiniband for clustering the RAC systems and have Oracle reside on your RamSan. You can cluster up to 100 nodes on a database with TAF (transparent Application Failover), Load Balance and online backups using RMAN. With Load Balancing your users will not know which node there on and the performance with Active/Active...adnauseum will be tremendous.
Your up time (if designed appropriatley) will be 99.99% and to include you can do rolling upgrades with GRID and have your servers be managed by grid by manageing node/system resource utilization to include the EVE ONLINE game.
Interested?
|
The Mule2
|
Posted - 2009.02.18 22:51:00 -
[145]
Edited by: The Mule2 on 18/02/2009 22:51:19 Glad you didn't go for an Intel based SSD solution???
I bet EVE is pretty hard on the flash memory.
http://arstechnica.com/news/2009/02/sector-remap-fragmentation-slowing-intel-x25-m-ssds.ars
|
Night Doc
Orekaria
|
Posted - 2009.02.19 07:43:00 -
[146]
Seems to me that the querys are now almost instant but the time to wait for the query results have increased.
ex: you ask for prices about a item. you get one or two second pause (seems more like a hold state) and then you almost instantly get the results.
- Target analysis - Fit EVE to screen - EFT setup sort |
Naga Tokiba
|
Posted - 2009.02.19 12:06:00 -
[147]
Did I read somewhere:
Shorter daily downtimes
But that dont seem to apply to today I guess
|
|
CCP Valar
|
Posted - 2009.02.19 12:52:00 -
[148]
Originally by: Naga Tokiba Did I read somewhere:
Shorter daily downtimes
But that dont seem to apply to today I guess
We are still doing some changes. Today I was moving disks between disk arrays.
We currently have 6 enclosures, but only need 3 for backups and such after we moved the data to the RAMSAN 500. The array that houses the "quorum", the database for the Windows clustering service was partially on the enclosures we are removing, so I had to fix that this downtime.
Tomorrow, I'm moving the backups from the current array to the former primary data enclosure and after that we can turn off the 3 enclosures we are retiring.
---- Virtual World Database Administrator Operations department CCP Games |
|
Naga Tokiba
|
Posted - 2009.02.19 13:20:00 -
[149]
Excelent - yous people are really doing a great job.
I guess we now have all new and fast hardware ready for next expansion - just can't wait.
|
Jekuvu
|
Posted - 2009.02.19 22:22:00 -
[150]
I'm waiting to see eve on top 500 http://www.top500.org/lists/2008/11. I know you guys would be close at 832 cores. There are clusters with 960 cores there (although they are 4.6Ghz apparently). So you are looking at needing another 40 odd servers to get on that list.
PS. How awesome would I have to be to get an internship with you guys?
|
|
|
|
|
Pages: 1 2 3 4 [5] 6 :: one page |
First page | Previous page | Next page | Last page |