Terebinthia wrote on Mar 12
th, 2013 at 3:35pm:
. I just can't grasp why they can't mirror live...
Fucking hardware cost...
I have what I think is an accurate idea of infrastructure involved in the Live server....
Lets just say that :
- 1 Oracle Cluster ( 2 powerful nodes at least )
- 1 EMC² Symetrix ( to hold the several terabytes of data )
- 1 SAN Infrastructure ( to connect the Symetrix to the Oracle Cluster and the Game Cluster )
- 1 Load Balancer for the Oracle Cluster
- 1 dedicated back end network infrastructure ( Gigagyte net, probably fiber)
- 1 Game Cluster ( around 100/200 blades, probably more )
- 1 Billing/accounting/authentication cluster ( several nodes .... 5/6 as it's common infrastructure )
- 1 Load Balancer infrastructure for the game server.
- 1 Front End Network ( common to all the servers )
Should be IMFO representative of what we call one single game server. ( say G-Land )
The number of blades in thee Game Cluster can be dynamically allocated to another server if needed... Eventually the Symetrix can be shared if they went for a big one, but it add complexity at the SAN and back end network.
not many companies can waste several hundred thousand dollars ( or euros ) for a test system.
And the few that can ( and do ) are usually way bigger than Turbine. ( lets say : in Telco : ATT, Verizon, BT, FT, DT, Telia, Vodafone, ... Ericsson, Alcatel-Lucent, Nokia-Siemens, Huawei, ZTE, ... HP, IBM, Oracle, SAP, ... )