Hi,

re. # Questions & Answers * Q: zimmer family hitting max descriptor size https://gitlab.torproject.org/tpo/core/tor/-/issues/40837 Any news on a possible solution to this ongoing issue?

tl;dr: more powerful relays rather than large numbers of them.

*Please note I have very little idea what I'm talking about here and there's probably something I don't know about or just don't understand.

If a family has a very large number of exit relays but the capacity of each one is not that great couldn't they just spend the same money on making existing relays more powerful rather than having more of them?

If we're using VPSs then the limits applicable to single cores don't seem to apply the way they do to physical servers. I run a single non-exit relay but I get more traffic through it by having more RAM and multiple vCPUs. The VPS seems to act like it has one large, powerful core rather than the 5 smaller one's it 'actually' has. So if 'descriptors' refer to the number of relays not their capacity wouldn't this work?

Side note: to get more bang for your buck you can ask your VPS provider to swap some storage for more vCPUs, that's what I did and it certainly seems to allow me to run more traffic through my one relay. It doesn't need that much storage anyway.

So you get the same amount of traffic through fewer, larger relays.

cheers,