Multiple Tor relays vs one bigger relay on one connection

Question

Today I got an idea, that I have multiple (5) computers, which I can have running all the time, so why not to make them all relays, right? Well for one, I have limited connection (50mbit up/down). So I guess I would surely lose my Guard status, if I ran those 5 computers on that line. Is there any reasoning as to why would I consider it?

1. server with Xeon 4/4 3.2GHz
2. desktop with i7 4/8 2.4GHz
3. desktop with i7 2/4 2.6GHz
4. desktop generally slow, I don't know the exact CPU
5. desktop generally very slow, I don't know the exact CPU

I am currently running a relay on the server. With regard to the CPU speed, which I don't really think is the limit, nonetheless, if I had those first 3 computers running all the time as relays, is that a good idea for some reason than having all bandwidth used by the server?

Answer 1

You can only run two relays per IP address (as I understand it, anything published after that will overwrite one of the older ones).

Tor doesn't currently handle multithreading crypto operations, as such it's semi-common practice for some operators to run two tor instances on one server in cases where the CPU throughput isn't sufficient to saturate the available bandwidth, each tor instance would utilise one CPU core.

In your case, you'd need to see if a single core running tor is sufficient to use up the bandwidth you wish to provide to the network (AESNI?) and if not you might consider running a second tor instance or a relay on a second device. Running on 5 systems would require you to have 3 distinct IP addresses to operate the relays from.

If you chose the multiple relays path you should look at setting up a MyFamily directive in your torrc to explicitly state that all the relays belong together.