At FamilyTreeDNA, two introductory levels of Y-chromosome STR testing are offered to new project members:  37-marker and 67-marker tests.  Every level of testing tells you something, but I have standardized my projects on 67 markers because that level allows for confident estimations of relatedness in virtually all cases.  Advanced testing to 111 or more markers is available to refine relationships (e.g., for distinguishing branches of the same family).

STR testing measures relationship in a time frame of hundreds of years, making the distance to the MRCA (Most Recent Common Ancestor) sufficiently close to be genealogically useful.  And "useful" is an understatement.  STR testing is a powerful tool for determing whether people do or do not share a recent common ancestor, and it is the foundation for Y-DNA surname projects.

The haplogroup (sometimes abbreviated, "Hg") is a measure of deep ancestry.  The MRCA may be thousands of years in the past, which places the individual's ancestry in a paleoanthropological time frame.  The chronological appearance of SNP mutations has been used to order the branching of the "Y-DNA Haplotree," and a SNP test will determine a male's position on the tree.

As each individual has just one evolutionary path on their patrilineal line, so their SNP-based haplogroup should correlate with their STR-based haplotype, and it does.  If the haplotype is reasonably common (or similar to one that is), it can be used to deduce the basic Y-DNA haplogroup, without the added expense of SNP testing.  If the haplotype does not unequivocally indicate the haplogroup, FamilyTreeDNA will do a "backbone" (basic) SNP test without charge to securely determine the basic haplogroup.  Deep SNP testing would still be needed to determine the haplogroup subclade, and if your haplotype is rare, I recommend deep SNP testing simply as an aid to researchers because their research on your haplogroup ultimately tells you more about yourself.

A decade ago, I would have said SNP testing was merely "interesting" in its ability to correlate haplogroups and their subclades with the paths of human migration over the millenia, and I did not press my project members to be SNP tested.  I considered it of minimal use to the genealogist, beyond the fact that being in different haplogroups totally rules out having a near common ancestor (in genealogical time) — and because haplogroups strongly separate families, they are a convenient way for me, as project administrator, to organize surname project member results. 

At this writing (2012), the number of SNPs found has greatly increased, and the rate of their discovery is increasing due, in part, to FTDNA's "Walk through the Y" program.  We are reaching the point where they are of use to the genealogist, especially in the common haplogroups, as the Y-DNA haplotree turns into a finely divided bush.  For that reason, I now urge that Y-DNA project members,  especially those who are Haplogroup R1b1a2 or I1 — the two most common haplogroups in western Europe — be deep SNP tested and that they continue to test new SNPs as they are discovered.

As the number of people STR tested increases, SNP testing will become important in separating "borderline" STR matches that are merely coincidental, that is, not indicative of real relationship, from ones that do indicate real relationship.  The problem of ambiguous coincidental matches, particularly in Haplogroup R1b1a2, is going to increase as more people are tested.  Thankfully, deep SNP testing gives us a method for detecting them.

The bottom line is that, while SNP testing is not a requirement for participation is a Y-DNA STR surname project, I encourage it because, while I once thought deep SNP testing was merely "interesting," it has now become actually useful to the genealogist.