10 units, i.e 1uL is usually enough to digest 1ug of DNA.

For a total volume of 20ul you would use 2ul 10x buffer. If you want to digest 1 ul of DNA with a concentration of 5ug/ul, you should make a 1:10 dilution of your DNA and then use 2ul of your 500ng/ul DNA. For the restriction enzymes you would have to look up the units/ul (one unit is enough to cut one microgram of DNA per hour). Usually 1ul of RE is more than enough for your digestion. The rest is water.

As I remember, 1 u definition is the amount of enzyme sufficient to digest 1 ug of DNA (in most cases phage lambda) in 50 ul during 1 h. However your real situation can differ from the conditions used by enzyme manufacturer. For example, SalI prefers linear DNA and its activity towards a super-coiled plasmid can be much lower than indicated on the tube. Excess of enzyme may lead to star activity (due to high enzyme concentration (>20u per ug of DNA) and glycerol (>5%), which is a component of storage buffer (50%), however this is a problem only in case of certain enzymes such as EcoRI. I have never met this problem. If you are going to use you DNA as a cloning vector, some over-digestion might be required to ensure absence of partially digested DNA, which may produce empty clones after ligation.  

while it is true that 1u is often equivalent to 1ug of dna that definition relies on the average distribution of 4 or 6 cutters in either lambda or human genomic dna and 6 cutters cut rathed infrequently so generally both of the above answers are correct. Bear in mind though that if your target is ocr amplified dna then every molecule may contain the cut site which might make it 1 cut site every 200 bases ( or the size of your amplimer) in which case more enzyme would be needed. Due to star activity it is often better to cut for longer rather than use more enzyme which means more glycerol so increases star activity,  Also some enzymes need 2 recognition sites to pull the target dna together in order to cut it so pcr product may cut badly if it only has one cut site as cutting then requires 2 target molecules to come together for good cutting which happens less frequently than finding the second cut site on the same target