Rob Jansen
Rob Jansen
> Since Linux only seems to do this for tcp, it would be a no-op for non-tcp sockets. Or maybe we would implement this same behaviour anyways for udp sockets,...
I agree that jitter is a reasonable thing to want in a network simulator. We had it before the great Rust migration, but the model sucked so we never implemented...
> Do you have any preference whether we do this initially or as a follow up task? Keeping it simple to start with would be fine with me. I'm not...
> Would it be theoretically possible to integrate static or dynamic linking allowing users to provide their own rust functions that can be compiled/loaded into shadow Theoretically, sure. But I...
I don't think anything was merged yet to fix it... @stevenengler was this a one-off occurrence or does this test fail for you persistently in your environment?
> From some quick testing, this appears to be a kernel bug. Or at least the behaviour changed, since it seems nobody (including the kernel people) actually knows how edge-triggered...
Thanks for the post. Are these graphs created using data that was generated from within Shadow? To clarify, here is our current approach (mirrored in both directions): ``` Sender Interface...
>> But can we allow the process to send as fast as it wants without any form of back pressure? In Shadow's runtime model, wouldn't eventually result in ENOMEM? >...
I think you are correct in that perhaps eventdns does not use `getaddrinfo()`. Have you been able to dig into eventdns to see what is is actually doing here? Does...