EVE General Discussion

BIAB is taking a service and moving it outside of the solar system service, much like how chat or the market is unaffected by the tidi in a system. Effectively if you undocked or jumped into a system in tidi, the math to work out your stats used to get queued with other action in that system, but now it does not. While that helps, it doesn't fix the issue (and isn't actually multithreading, it's relocating a process) because every action taken in that system by players is still in a single queue relying on a single CPU to process them. A modern game server would have spread the bulk of those actions across multiple CPU cores so that multiple complex calculations could occur simultaneously.

This is why the new sov system takes you into multiple systems, because it spreads players across multiple nodes and thus across multiple CPUs. If they want to be able to run large battles without tidi however then the only real solution is to allow system actions to be spread across CPU cores, i.e. multithreading.

Ed: Oh, and yes, you are mistaken.

No I am not, since in a few posts you have proven to be very limited in knowledge of parallel systems. Taking a process to outside of another process is the DEFINITION OF PARALELISM. Since the BIAB is running on a different machine and therefore in a different CPU, by DEFINITION it is running in a different thread since not a single architecture of computers that is used outside academics have a single thread run at same time on different CPUs.

The simple basics of operating systems for you.. a process can have one or more threads, 2 process will have at least 2 threads.. therefore.. multi thread. Now, since it is clearly established that you have no clue what multi threaded really means, maybe you should stop trying to say to CCP how to solve anything?

Good for you? Though more important than a degree is working experience. Most of us that work as developers do, even new starters with very limited knowledge.

The thing is, we already know what the problem is here, as CCP have shared this in the past, and the fact that it would add overhead is irrelevant. Even if they were just offloading some of the math to other cores and still processing the actual events on a single CPU, that would be better than what they currently do which is wait more than 10 times as long to process the events.

But we know what happens in MMORPG servers. There is wide ranging research into it, and even information given out by CCP. If they threaded the servers and reduced performance that would simply mean they wrote it wrong. I don't even think CCP would suggest that they wouldn't benefit, they simply aren't willing to dive that heavily into their legacy code.

No, you really are. And I didn't say parallelism, you did. I said multithreading, which can be a form of parallelism but isn't the same thing. Kinda like how an apple is a fruit but fruit isn't just apples.

See, what you are talking about is how they took the old attribute code that existed within the solar system service and they added it to a new BIAB service which operates as a separate process on the server and thus can run on it's own core. That reduces the amount of work inside the solar system service. There's limited scope to do that however as a lot of what bottlenecks the solar system service can't be handed over to a separate service (at least not easily) as it requires granular control over how that service would operate and a lot of shared data.

Taking the actual solar system service however and allowing it to use threads to distribute its workload across multiple cores would allow it to do some of the calculations simultaneously with the main service thread only needing to worry about the events as a whole. Say for example 10 people fire lasers at a target, each of those damage calculations could be done together on different cores and the service thread would take the results of those and apply them to the target.

So if I open a copy of notepad, then I open another copy of notepad, have I multithreaded my notepad? The answer is no, those are two separate processes. Did you actually pay attention when you got that degree?

*nerd rage intensifies.

Are you aware that one process can have multiple threads, and that parallelism by multi-threading a single process and parallelism by creating multiple processes are not the same thing? Running a single-threaded process twice or running two single-threaded processes that communicate with each other does not make the processes multi-threaded. The bulk of the work of the solar system process cannot be split into multiple processes, and so the only way to increase throughput without increasing the speed of a single CPU core is by multi-threading the single process and distributing the load across multiple cores.

See, here you seem to be confused. At no point did I claim you could have a 0 thread process, and you're going on about shared address space and multiple computers as if that's part of the conversation. It's not. I get the feeling that you think the solar system service uses all of the cores of a single CPU and that we're talking about distributing the workload onto another machine. We are not, we are talking about how the solar system services uses one of many cores on a single CPU, and how by utilising more cores the single process could get more work done in less time even with the overhead.

What I find amazing here is that you have the audacity to insult me as if I'm the one that is confusing the terminology. Perhaps you should go back and re-read the posts. You seem to have no understanding of why "parallelism" and "multi-threading" are not directly interchangeable. I call bullshit on you having a degree if you can't even work that out. Screenshot or it never happened.

The latest TQ devblog hints at them being able to direct resources at will ... soon ...
... thanks to virtual environments that allow to sidestep pythons problematic use of multithreading.

Don't forget ... it's still python with C, or C++.
Doesn't make things easier.

Depending on what you are doing you'll already be distributed over multiple services. Undocking does a lot on BIAB now, market are on a service, so is chat, etc. The problem service that causes the lag is SOL, the solar system service. This only has a finite amount of resource that can be used (up to one maxed out CPU core) at which point it's at it's limit. If they know in advance, they can make sure the target solar system is on one of the machine with really beastly CPUs, but they can't move the node with a fight already occurring as everyone will get booted. Then the best they can do is move other solar systems off of the node and crank up the processor a bit. That can go wrong though, like in the battle of Z9PP. In theory though, the new virtual environments will allow them to move or expand a running node without booting people, but we'll have to see how that works out in practice once it's all rolled out.

Now I'm curious... You should start a threadnaught on your chosen subject.

I am so looking forward to this.

Greetings Pilot!

I would like to invite you to apply to www.ccpgames.com/jobs if you believe you are capable of creating a multi threaded experience For Tranquility that CCP could use to remove Tidi, since you believe it to be so easy. CCP Could really use your Expertise and assistance on such a grand task!