Make assumptions and crunch equities. This is not a feel spot, but a math spot. Crunch numbers and build your intuition that way. When modeling it's useful to start at the extremes and see what happens there. So let's start with something simple and worst case'ish for you.

For example:

- UTG: AA$nt, AKK$R$nt$ds,$R$R$R$R$np$ds
- SB: AA$nt, AKK$R$nt$ds,$R$R$R$R$np$ds
- BB: $0g$ds,$1g$ds,$2g$ds

And assume only CO folds. Your equity against these very premium ranges is 22.44% in the main pot. So losing a little there (your stackoff threshold is a little lower than 25% because of the dead money from calling the first raise, but it's not a big mistake to ignore that).

In the side pot with SB and BB you have 31.21%, so losing a little there as well (since you need 33.33% to break even 3-way). And in the other (small) side pot HU vs BB, you're 58.22% and winning a little (50% is the break even point).

OK, this indicates that the worst case isn't so bad, but you're not making money there. You might find ranges that are even more worst case, but you get the point. 

So next step is to make more realistic ranges (it's a safe bet that we can keep the assumption that nobody except CO is in a folding mood). People are looser than the ranges above, and you should do better.

Crunch your equity in the main pot and two side pots, and calculate your expected return from each of them. If the sum of those returns is larger than your current stack (use the stack you have after calling the raise), you can stack off. But you should call and let someone else do the shoving, because you want as many weak hands in there with you as possible, because your hand is nutty and performs well all-in multiway against loose ranges.

Subtract the rake from the pot for extra EV accuracy points.