indio wrote:Why a shot in the dark? Because you've got an untested, and untestable (until the investment is made) theory.
How could this turn out to be a "motorised scooter" ? When the shot misses its target, and we are left with a $5000 computer hosting maybe 1 or 2 servers, our IRC chanel, forums and database. It's a Rolls doing the job of a scooter.
What are the technical reasons you do not see this working "as advertised" ? You want to run 7 instances of a NWN modules on a computer with 8 cores.
If you have the technical background to think otherwise, what are your ideas on best solving this ?Make a better informed investment. Spend some time and money on R&D first. Compare the actual NWN2 hosting performance on a core2duo versus the quadcores. Are 4 6300 core2duo machines a better environment for players than 1 dual quad? What difference in actual hosting terms does a 6800 make over a 6300, given their price difference is almost $1000? Given mods over 2GB can now be hosted, what difference will that make? Were you aware they could now be hosted? See if you can hire a quadcore, install the software, and set up a test for us. Know your subject before developing a plan.
Consider also that if we have one server host, when it goes down, all modules go down. It stays down for a week, we are out of business for a week. As it stands, it's a bad call. And it's so often the case with hasty plans. You've just arrived as IA, I'm guessing have never even locally hosted a NWN2 module, and are now planning to spend ALFAs money.
We will give the money. But before you spend ALFAs money, I'm suggesting to you that actually doing some research of your own might help convince people this is a good idea. The onus is on you to show that your plan can work, and convincing us that this investment is actually a good one. As it stands, it's not.
I dont see any answers to the 3 first questions. You're only restating what you allready said. I was asking for reasons as to why.
Modules cannot use more than 2GB ram because nwserver is 32bit. In a 32bit x86 environment, the maximum ammount of memory you can allocate to any single process is 2GB. This is a mathematical limitation that can only be solved if a 64bit version of the serversoftware is released. The developers have stated that this will not happen.
The whole purpose of multiple processors(cores), is scalability. Since each core running a nwserver instance, has access to 2gb of memory, you can scale. Problems would arrise if each of the instances would require massive ammounts of data to be read or written along the PCI bus. With neverwinter nights, this is not the case. Only very little data needs to be sent and recieved over the bus to the NIC and the PCI Express bus to the harddrive(s). Since the memory bus is wide enough to facilitate even more nwserver processes without becoming a bottleneck, it is very realistic to expect scalability of 1:1 to be possible atleast up to 8 instances of nwserver. I say 7 because we need 1 core and 2GB for the system itself (OS) and the SQL server.
Also, with this approach, we get the very valuable benefit of running the SQL server locally. This means that the only bottleneck on game<->db interfacing is the memoryspeeds (measured in nanoseconds). Furthermore, we get the huge advantage of a vault that will be local to each module.
IF instead we would use multiple servers, we would have to run a local SQL database on each of them, and tie those servers up to a "mothership" database elsewhere. This may introduce new problems, and is not efficient. Also, we would have to rely on means of sharing the bic's, wich we know from experience is hard to do efficiently. There is also an added problem with performance, and multiple hosts, running on various links, most beeing in peoples homes, are much less likely to have the sort of uptime you get from centralizing and using a professional serverpark that actually promises 100% uptime on connectivity and power. If we were to host multiple servers in a serverpark, we would have to pay for each one. The price of colocation is set by the size the server(s) take. I am proposing a single 1U blade since it is very cheap to host, meaning low montly cost for hosting ALFA.
With all due respect Indio, i would wory less about the technical aspects of this, and leave it to the members who have the most practical IRL experience and technical theory under their belts.
This discussion is meant to be about how to organize the donations themselves, not discussing scalability theory.
Just to state the obvious: Nobody is talking about spending alot of money on something we are not sure gives us what we want. I would have hoped everyone were able to give me atleast that much credit.
(I personally trust you, just playing devil's advocate here as I said).
Ouch...stop...my sides hurt too much from laughing...no more...bwahahaha
