Well, it's all about the overall throughput that most are concerned with. Finding the bottlenecks that produce the most issues is a daunting task. I seriously doubt, especially at this stage of the game, that there is any connection whatsovever between the base code being directly tied to any single core registers. My belief on the matter is that they just claim one sim per core as a division unit per server. In other words, 2 cores, two simulators as a generic way of conducting what resource usage is provided each. But I serously doubt that the base code is tied to any certain processor core. I'm not saying that it's impossible, but I am saying that I doubt it is the situation you witness when Linden speaks of a core per simulator.
I have, just for a simple visual test loaded upwards of 45 back to back open simulators on as little as 780M of ram before the blank simulators would not function on linux from a remote location. This is by no means a serious way of testing, but it does show that a requirement may be obtained as to what the true resource usage and requirements are. At least in a broader sense. Detail can become apparent via the testing. However, serious detailed reference to what resources are being used would of course take time and a set of testing scripts written for the purpose, and dependant on the platform. There are some generic scripts that would be of use to monitor some of the usage.
It's trial and error, but Linden should have this down by now as to what hardware resources are required per simulator and limits to scripting per simulator is paramount even with scripting limits imposed. You don't have to have low leval statistics to have a fundamentally sound perspective on what an end user sees and experiences. And certainly the end user is not required to be all knowlegable of usage when they have no access to verify the claim. Even so, Linden's product is a managed service and end users are not required for the over usage even if it be the case. However, with no hard limits imposed on the various areas of concerned, it certainly isn't within Linden's best interest to babble blog about over usage.
Personally, I believe Linden believes they under estimated how successful the product would be at that price setting, and decided to do whatever it took to make more from it.
1 Core, 2 Core, 3Core, Give me more!
Tue, 11/11/2008 - 23:44 — Anonymous (not verified)Well, it's all about the overall throughput that most are concerned with. Finding the bottlenecks that produce the most issues is a daunting task. I seriously doubt, especially at this stage of the game, that there is any connection whatsovever between the base code being directly tied to any single core registers. My belief on the matter is that they just claim one sim per core as a division unit per server. In other words, 2 cores, two simulators as a generic way of conducting what resource usage is provided each. But I serously doubt that the base code is tied to any certain processor core. I'm not saying that it's impossible, but I am saying that I doubt it is the situation you witness when Linden speaks of a core per simulator.
I have, just for a simple visual test loaded upwards of 45 back to back open simulators on as little as 780M of ram before the blank simulators would not function on linux from a remote location. This is by no means a serious way of testing, but it does show that a requirement may be obtained as to what the true resource usage and requirements are. At least in a broader sense. Detail can become apparent via the testing. However, serious detailed reference to what resources are being used would of course take time and a set of testing scripts written for the purpose, and dependant on the platform. There are some generic scripts that would be of use to monitor some of the usage.
It's trial and error, but Linden should have this down by now as to what hardware resources are required per simulator and limits to scripting per simulator is paramount even with scripting limits imposed. You don't have to have low leval statistics to have a fundamentally sound perspective on what an end user sees and experiences. And certainly the end user is not required to be all knowlegable of usage when they have no access to verify the claim. Even so, Linden's product is a managed service and end users are not required for the over usage even if it be the case. However, with no hard limits imposed on the various areas of concerned, it certainly isn't within Linden's best interest to babble blog about over usage.
Personally, I believe Linden believes they under estimated how successful the product would be at that price setting, and decided to do whatever it took to make more from it.