Research statistical packages for simulation

[mmacfadden]

We are going to want to be able to model user behavior, network bandwidth and latency, etc. We will probably want to use normal distributions for this. We should find a good pattern for sampling from a statistical distribution.

[NoraCodes]

I don't know that we need anything special for this; rand::distributions has a wide selection, including Normal and Gaussian.

[mmacfadden]

Hi thanks. Yes I am aware of those packages. This is not just about the distribution. This is about how we build profiles for describing things like user behavior and network performance using these. Some design thought is needed to decide what parts of these concepts we need to model and to what fidelity. Also which distributions are most appropriate for which concepts, etc.

[NoraCodes]

Oh, I see. That makes more sense. I know there are packages around for data mocking but it might make sense to roll out own depending on exactly what we need.

…

On Sat, Mar 4, 2017, 15:16 Michael MacFadden ***@***.***> wrote: Hi thanks. Yes I am aware of those packages. This is not just about the distribution. This is about how we build profiles for describing things like user behavior and network performance using these. Some design thought is needed to decide what parts of these concepts we need to model and to what fidelity. Also which distributions are most appropriate for which concepts, etc. — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#11 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ADI0_WC4TBQBTqyiikgAswepdVv2IjDkks5ridStgaJpZM4MTM7J> .

[mmacfadden]

Yeah at this point I am not sure what we want to "use". Totally focused on what we need to simulate and then thinking what kinds of statistics we need.

[dnorman]

@mmacfadden @SilverWingedSeraph I'm hoping it'll become a little clearer what we need to do once I've completed the upgrades to the network simulator which @mmacfadden and I had discussed the other night. I have substantially more clarity about what we'll need to support the user behavior simulator. Should be another couple days on that. Working on other core stuff tonight :)