ACI 318 (1999 & newer)and IBC2000 both have sections on anchorage to concrete from which you can calculate the required anchor rod embedment depth.
I assume you have a moment type conenction, so that the rods are not equally loaded. In this case, figure out your worst case rod, and design for that embedment depth for all rods. Now if you have overlapping failure cones, you will have to reduce the capacity based on the reduction in projected surface area. So you might have to iterate a few times to get the correct answer. If you need to prestress the rods or have a significant ductility requirement, make sure you have enough rod length unbonded to provide for the required ductility. This may mean either extra long embedment depth, or connecting the rods to the columns some distance above the base plate.
For analyzing the stress at each rod, assume a neutral axis thru the middle of your round column, preferrably thru two of the rods. Now, assume a linear strain distribution. Using this, you can calculate the strain at each rod in tension based on its distance to the neutral axis. Once you have the strain, you can figure out the stress, and rod area required. This would be similar to a deep concrete beam with multiple layers of reinforcing steel. I actually had to do this analysis during a concrete exam back in university for a round beam with bars spaced around the perimeter. Kinda took us all by surprise, but the same method holds as for rectangular beams.
