Ever tried resource pools? This would solve your problem.
You can create a pool, say 2G.
Put all 4 VM's in it. What's the likelihood that ALL of your VM's NEED 1G all at once? Probably very low..
In a larger server farm, more VM's = lower likelihood that ALL of them are going to be using their full compliment of RAM allocation, pools help fix this problem, and a pool is dynamic.
The memory size that you set is a limit. A VM cannot go over that limit. Since you can't dynamically allocate memory while the VM is running, you can simply allocate as you did enough RAM and then use pools / user limits to keep the memory at CURRENT usage, not future usage.
You have to manage your ESX server, look at the performance chart, and keep an eye on what EACH VM is doing. ESX requires some management, you can't expect to just set it and walk away. Any good system requires management.
You can use tools like ESX charter to plot your data and export it to a PDF, since ALL VM's have different requirements, its safe to assume their memory limitations will be different as well.
IF VM1 has 1G physical allocation, yet the avg memory never goes above 512, you are wasting resources. You should be able to determine how much a given VM is USING and watch the peaks to ensure adequate memory, not fluff.
This why I ALWAYS start low on my memory, and wait for people to complain it's not enough. In 6 months, I have only had to increase memory from 512 to 768 on just 2 VM. You may even find less is more, try setting your physical values a little lower.
4 VM's @ 1G each is your host limit. 4 VM's @ 768 is 3072 leaving approx 1G for 1 more VM . . . . . . Managment is the key.
