Global Network simulation cancellation question


#1

If I have one of your Global Network simulations running and I cancel it and shut down the client, does the client save the work that has already been done in the simulation and pass off the rest of the work to another node on the Global Network? Or is all the work up the point of cancelling it lost?


#2

The work for the chunk sent to you would be lost. The glonet would send that chunk to another client the.


#3

Out of curiosity, is a “chunk” a small portion of a larger job (i.e. a 15 minute portion of a 3 hour sim)? Or is a “chunk” considered an entire simulation which could potentially take several hours?

Are there any plans to include “estimated time until completion” functionality into the client?

I use and pay for AMR. I love it and want to help progress the cause. But I also take care of my computer equipment, meaning I don’t leave it running for extended periods of time for no reason.

My concern is that the client could be working on a chunk, where it takes 10 more minutes to complete. In that case, I would just wait out the 10 minutes before shutting it down. Another example would be where the chunk could take another 8 hours. In that case I wouldn’t mine leaving it running overnight. What I want to avoid is having a chunk running, that will complete in a few more minutes, and then the client gets no more work and I have an idle machine running throughout the night.

I know the Global Network thing is your new baby, but I suggest that, at some point in the future, some user feedback be integrated into the client. This will help the user decide how to manage the client on their end.


#4

A “chunk” is a batch of between 5 and 100 “setups”. So, it will run up to 100 simulations in one go. Depending on the speed of the client, any single chunk shouldn’t take more than a few minutes.

Right now we are working on the “clustering” for the new gearing strategies. Basically, a bunch of code we have written to decide which strategies to even calculate. That clustering itself needs to run a lot of simulations, which we are sending to the global network as fast as we can finish them, since we need to seed the logic with some “hints” for each spec to make the problem even approachable. Once those are done, we will calculate all the gearing strategies.

We also continue to add more boss scripts based on ToS bosses.

So, once the clustering is done in the next few days… there will be simulations running on the global network almost constantly, maybe even 24/7.