RKS Server Info as requested...

PY004

Administrator
Staff member
Administrator
Moderator
Joined
Jun 11, 2010
Messages
375
Reaction score
28
Location
New York City
Some people have been sending me e-mails asking for the specs to the RKS Servers. Instead of having to reply to them all, I have decided to post the specs here hoping that the number of e-mails sent to me would be decreased.

MSC servers hosted by RKS is currently located in... Virginia (Daedalus), Texas(Prometheus), and United Kingdoms (Beliskner). Yes, the servers are named after Stargate ships. There are two further RKS servers that are currently not hosting MSC servers in VA (Odyssey and Orion). Those servers are used for auxillary services such as web server, voice server, VM, etc. Total monthly cost for operations is approx. 1300USD.

VA Server (Daedalus)
Dual Intel Xeon E5520 Processors (16 Threads total)
12GB DDR3 Triple Channel RAM
1Gbps Full Duplex Dedicated Uplink to the Internet
Windows Web Server 2008 R2
*Oddysey has the same specs as Daedalus except for Windows Server 2008 R2 Standard, Dual 1TB HDD RAID1.
*Orion has the same specs as Prometheus except for CentOS5 32Bit, Dual 1TB HDD RAID1.

TX Server (Prometheus)
Intel Core 2 Quad Q9400 (4 Cores)
4GB Dual Channel RAM
100Mbps Full Duplex Dedicated Uplink to the Internet
Windows Web Server 2008 R2

UK Server (Beliskner)
Intel Q9400 (4 Cores)
4GB Dual Channel RAM
100Mbps Full Duplex Dedicated Uplink to the Internet
Windows Web Server 2008 R2
 

cartman-2000

New Adventurer
Blades of Urdual
Joined
Jun 6, 2010
Messages
160
Reaction score
0
Age
38
Location
Beaverton, Oregon
I knew there was a difference in hardware from the uk/central servers form the eastern ones, this mod likes high clock rate and large cache cpu's, so the core 2 ones out perform your xeon one for msc servers.
 
  • Thread starter
  • Admin
  • #3

PY004

Administrator
Staff member
Administrator
Moderator
Joined
Jun 11, 2010
Messages
375
Reaction score
28
Location
New York City
Yes that was expected as the server was mostly single threaded. For some weird reasons, the process tends to use the first available logical core on the server. Changing affinities on the process results in a server crash (or server not starting up if set in the service properties). So when multiple servers on the same box is loaded with people, it competes for resources. Also TurboBoost is usually active most of the time except during times when processor intensive apps are run on it. so the clock speed should be 2.53GHz but for some reason, MSC doesnt seem to benefit much from it.
 

Thothie

Administrator
Staff member
Administrator
Moderator
MSC Archivist
Joined
Apr 8, 2005
Messages
16,342
Reaction score
326
Location
lost
Odd... Might be something with the HLDS and Windows Server. I offload the HLDS to core #4 on XP with no issues (on a Q6700). :\ For a time I was running both the regular and newb servers on cores #3 & #4, respectively.
 

cartman-2000

New Adventurer
Blades of Urdual
Joined
Jun 6, 2010
Messages
160
Reaction score
0
Age
38
Location
Beaverton, Oregon
You can try to use the -noaff flag, without it it tends to lock all servers to a single core(to the core the first server started on.), with it servers can run on any core, something I think they added to the steam works update a while back. Also, is it possible to disable HT on the east server, a server shouldn't really be running on a logical core(for best performance.).
 

Thothie

Administrator
Staff member
Administrator
Moderator
MSC Archivist
Joined
Apr 8, 2005
Messages
16,342
Reaction score
326
Location
lost
Techinical thing I forgot... I'm using Prio and Taskassign to assign cores, so it maybe overriding the HLDS.
 

cartman-2000

New Adventurer
Blades of Urdual
Joined
Jun 6, 2010
Messages
160
Reaction score
0
Age
38
Location
Beaverton, Oregon
I use serverdoc for running servers, but it can't force a core unless I use the -noaff flag.

I think the reason why the added the affinity lock back in the steam works update is supposedly it can get better performance when it's not floating between cores, but there implementation is bugged for multiple servers, once one server is started all others get forced to that core.
 
  • Thread starter
  • Admin
  • #8

PY004

Administrator
Staff member
Administrator
Moderator
Joined
Jun 11, 2010
Messages
375
Reaction score
28
Location
New York City
The East server's workload consists of software that primarily benefits from HT which is why it's on at the moment. I found out from migrating from a E5420 that HT has improved significantly from the previous implementations (or the software itself has been updated to work better with it). It hurts single threaded performance, however, with careful loading of the server, you can get better overall processing power at the expense of a minor performance drop for single programs.

Currently I am running my servers through TCAdmin (makes administrating the servers easier since my leaders can have access to the server without me granting them RDP) and I can set the priority there or manually through task manager. I have manually tried launching hlds but still the same problem. I will try the noaff.
 
Top