The amount of sample you'll need is dependent on the teechnique yu have available to you and your work. Less sensitive requires more sample and vice-versa. Considering your interest in analysing sphingolipid class (Cer, SM, HexCer, LacCer, gangliosides,..) I found that the best solvent system will be the MTBE (Matyash et al., 2008, JLR, 1137) rather than using other solvent systems (Folch, Bligh and Dyer, others). Hope this helps