This is a technical question on setting pricing breakpoints and there is no way to answer it without understanding what you are actually selling and gathering actual usage data. Use this rough approch.

1. Start with the unlimited option ONLY and run that for a few months to get the "natural" demand curve... how much people actually use.
2. Then plot the frequency distribution of number of customers with respect to amount of actual usage. This will likely be a bell curve.
3. Find your average "share" by dividing your load capacity by the number of users. Plot that as a vertical line. You are losing money to the right, making money to the left.
4. Outliers on both ends are bad. Customers who are paying too much can be taken away by the competition. Customers who are paying too little are of course just unprofitable and the more of them you have, the more money you lose.
5. Pick a volume breakpoint line to the left of the breakeven "share" line. You want to give the people to the left of THAT a break. So set the price cut based on the volume to the left. Compute lost revenue for the fixed base (if you like, round to the nearest psychological price point like 14.95 or whatever).
6. Now increase the price on the right of the breakpoint line to make up the losses on the left. You've basically done tax redistribution. You may still be losing money on the right side, but at least things are a little fairer.
7. You'll gain users on the left, lose users on the right.
8. Rinse and repeat. People don't like frequent price changes, so to finish estimating the demand curve, try special discount promotions. Preferably of the "first three months free" variety rather than discounts to the base. That way you don't devalue perceptions of the product. Or do an illegible promo codes thing that users can't really see clearly beyond the promo code they personally get. You can run tests on a group of similar prospects by offering different promo codes and tracking conversion rates with respect to discount levels.