Phenotypic variance = Genetic Variance + Nongenetic Variance, where

Genetic Variance = Additive Genetic Variance + Dominance Variance + Epistatic Variance

You can read more on this in Falconer's book. 

Ok... Since we talk about Heritability H2), we need to know that H2 has two parameters that specify its sense... The total genetic variance and the additive genetic variance.

if we know that

1- phenotypic variance = total genetic variance + non-genetic variance

2- total genetic variance = additive genetic variance + dominant genetic variance+ epestatic genetic variance + interaction between all previous genetic variances individually and altogether.

3- non-genetic variance = variances due to environmental factors + Error.

4- heritability H2= Total Genetic variance / phenotypic variance. And by then called broad sense heritability.

5- heritability H2= only the additive genetic variance/ phenotypic variance, and by then H2 will be called narrow sense heritability.

now I think the relation between additive and phenotypic variances is clear based on heritability H2 

Thanks Ahmed for your comprehensive answer

Don't know if you are still looking for answers, but since your question was the difference between VP and VA, I thought I'd clarify.  Both answers so far are completely correct in what they have said, but it appears that they did not actually address the question.  VP, or phenotypic variance, is simply the observed, measured variance in the trait of interest.  For instance, if you are studying milk yield, then the observed variation in milk yield among the cows is your VP.  As the other responders mention correctly, this variance is the sum of the variation due to genetic reasons (VG) and the variation due to environmental reasons (VE), and the ratio VG/VP gives the proportion of observed variation (VP) that can be attributed to genetic reasons (as opposed to the environment).  This is called heritability in the broad sense, because it is a rather crude measure that includes reasons for the genetic variation that are not necessarily passed on to the next generation.

Why are any two individuals different in their genetics  (let's only consider two loci for the sake of explanation)?  There are three reasons. 1) The alleles at the corresponding loci are different.  For instance, individual 1 may have A1 and A2 at the A locus and B1 and B2 at the B locus, whereas individual 2 may have A3 and A4 at the A locus and B3 and B4 at the B locus.  (Note of course, that we are referring to diploid individuals).  The variation caused by these differences is called the Additive Genetic Variance (VA).  2) The second reason is the interaction between the alleles at each locus (complete dominance, partial dominance, co-dominance, etc.).  Thus, B1 may be completely dominant over B2, whereas B3 and B4 may be co-dominant.  Therefore, these interactions between the alleles at each locus also add to the genetic variance between the two individuals.  This is called the Domiance Variance or VD.  3) The third component of VG is the variation between individuals as a result of the interaction among loci.  For instance, the alleles at the B locus may be epistatic over those at the A locus, such that only the alleles at the B locus matter and those at the A locus are suppressed.  The genetic variation due to these reasons is termed Epistatic variation, or VI.  Furthermore, VG = VA + VD + VI.

Heritability in the narrow sense is given by the ratio VA/VP.  Why?  What is special about VA that it has been singled out in this definition?  The reason becomes clear when you realize that what's passed on to the next generation are only the alleles and neither the dominance interaction nor the epistatic interaction.  These have to be formed newly each generation depending on which alleles are inherited.  For instance, if the two individuals in the above example mated, it is possible that some offspring might have alleles A1 and A3 at the A locus and B2 and B4 at the B locus.  Clearly these are new combinations that were not seen in either parent, and so the dominance and epistatic interactions will be new.  However, barring denovo mutations (or gene-flow), the only alleles that the next generation can inherit are those available in the parents.  

As a side note, you can easily see that greater the VA in a given population, greater the diversity in it, and thus greater the narrow-sense heritability (healthier the population).

As if my answer was not long-winded enough, I feel the need to add more to it!  (Also, this has been a rather slow day, what with the holiday mood and all :-)).  Anyway, I feel the need to add to my answer because I missed the last part of the question, where you mention "realized heritability".  This term is used to refer, not to the ratio VG/VP or VA/VP that all of us have been harping about.  In other words, it is not a direct or definitional measure of heritability, in which you are measuring heritability as a function of VG or VA, as the case may be.  Rather, it is what is realized - after the fact (breeding).  By definition, realized heritability is given by R/s, where R is the response to selection, and s is the Selection Differential (also called Selection Intensity).  Let me explain.

Continuing with the example of milk yield, imagine a herd of cows whose milk yield you have measured for each cow.  The mean or average yield from this population is, let us say, a value X.  Now let us say you select only the highest yielding cows from this herd or population (say 10% of the best) for breeding.  Hence the term "Selection".  The average milk yield of these chosen cows is, let us say, X'.  You then allow the next generation to grow and then again measure the milk yield from all the females in this generation.  Now, if the alleles for the milk yield genes in the chosen parents were largely responsible for their high milk yield (that is, if the other cows - the ones that were not chosen - had low milk yield largely because of the "bad" alleles at these loci), then it means that VA is relatively high (and thus, the heritability in the narrow sense) is high.  Assuming that the environment remains the same for all the animals and for both generations, we then expect the average milk yield for the new generation (let us call this value Y) to be higher than that of the parent generation (X).   The difference (Y - X) is termed the "Response to selection" (R) as it is what has been achieved because you selected only certain parents for breeding.  Furthermore, the smaller the percentage of the chosen best parents (or larger the value of (X' - X)), the greater is the expected difference between Y and X.  That is why (X" - X) is called the Selection Intensity (s), because it refers to how far away from the mean of the entire population the mean of the selected parents is.

Now, it should be clear that the response to selection R is a function of two quantities - s, the intensity of selection (that is, which parents you select for breeding), and h2, the heritability of the trait (the relative  value of VA). That is, R = h2s.  This equation is used to determine the heritability: h2 = R/s.  Hence the term, Realized Heritability.  Of course, while it depends on the relative amount of VA, the two terms, Realized Heritability and VA are not generally used in the same context.  

I have attached a figure, which makes it easier to visualize R and s.

Just thought I'd complete my thoughts here.  Now I feel better!